Split deployment: Fedora desktops, SL servers
Nico's reply to my question prompted by his response to a list question. It was originally private because it was off-topic. But, truth is, I am interested to hear other folks perspectives on doing a "split" deployment: Fedora desktops + SL7 servers, or some other combo. I do sysadmin part time so having a common deployment reduces the workload. However, the flip side is that there will always be outliers that need bleeding edge stuff or situations where required software cannot be installed or built on an older OS without a lot of work. Same questions I posed to Nico. They are in the forwarded message below. Btw, I want to publicly thank Nico for his responses to the list. I've found his responses to be very insightful and helpful. Forwarded Message Subject: Re: Firefox and Thunderbird unpatched question Date: Tue, 26 Apr 2016 07:35:21 -0400 From: Nico Kadel-Garcia <nka...@gmail.com> To: Ken Teh <t...@anl.gov> I'll follow up privately since you did, but respond to the group, please. On Tue, Apr 26, 2016 at 6:51 AM, Ken Teh <t...@anl.gov> wrote: Can you expound a little more on using Fedora for desktops? Is this something you are doing? Not as a matter of course, but when developers or I need bleeding edge versions of perl or python modules, and sometimes even of various graphical toolkits, it's often easier to provide them on Fedora. I've started prep work on migrating from SL6 to 7 and I'm wondering if it's better to do a split deployment like you have hinted: Fedora desktops and SL7 servers. Depends. If your developers or clients need bleeding edge Firefox, Eclipse, or other tools, use what works. Fedora is effectively the development area for RHEL releases, and thus for SL. One important requirement would be an auto-update between versions of Fedora. Does this work well or are they usually major problems with the auto update? It's not recommended, and I've not been doing it on a long-term basis.It worked the last few times I tried it. I did occasionally have issues with RPM's I built myself, or installed from 3rd party sources, to get them out of the dependency lists and allow the upgrades of other components. And the switch from SysV init scripts to systemd just oh, lordie, that can be painful. Most of my users have to write code for their work. Right now we have a common home system across machines so they can work on any machine including remotely via ssh servers. Are they runtime compatibility issues in such a deployment? Would deploying a software collections installation possibly minimize this problem? Lord, yes. The "software collections" can help reduce the version conflict, but it takes extra work to activate those, and they're never complete backports of all the new tools. I wound up building roughly 160 RPM's for SL 6, using the python 2.7 software collection, to port the python based "Airflow" tool to a stable environment, and it only took 100 for SL 7. Fedora would have been much smaller, since many of them were backports of Fedora packages. Thanks in advance for your time. I want to say I have several of your messages tagged. I've found them very helpful. Good! Pass it on.
Re: fglrx problem
I have not had this problem but the last two Radeon-fglrx installations have been problematic. Most times reboots would result in X failing and the system dropping to runlevel 3. When it succeeded, windows would ghost when dragged across the desktop. I have decided to give up on radeon/fglrx. I fixed one of them with an nvidia 710 and I just ordered another one to fix the other. On 03/31/2016 01:18 PM, Stephen Isard wrote: I have a Dell Optiplex 980 with a Radeon HD 6450 card which was working fine with the elrepo fglrx-x11-drv and kmod-fglrx packages until I recently upgraded to kernel 2.6.32-573.12.1.el6.x86_64 and fglrx versions 15.12-1. Now when I move a window on the screen, the content isn't erased from the old location, leaving a trail of ghosts behind. This happens with more than one window manager (fvwm, icewm, fluxbox), so I think its an X problem, rather than a window manager bug. The guys at elrepo were very responsive, but they didn't have the problem on their own machines, and it's not really their software - it is closed source from ATI - and they have run out of guesses for what to try. I have tried installing ATI's own rpm and that behaves the same way. I've filed a bug on the unofficial ATI bugzilla, but don't have great hopes there. I'm wondering whether anyone here has ever seen symptoms like these. Using the radeon driver gets me a mostly usable system, but there is some stuff that says it won't work without glx. Stephen Isard
fglrx problems
I installed a Radeon 5450 card in an SL6.7 machine in the hopes of getting a higher resolution video. The on-board could only handle 1280x1024. But I've been unable to get it to work. At first I thought I bought the wrong card - a Radeon 6430. So, I ordered a 5450 after looking through the list of supported chipsets reported in the Xorg.0.log file. The 6430 was not listed even though it did list 6000 series. The 5450 was explicitly listed. But the net effect is the same. The Xorg.0.log reports no screens found. It apparently starts with a [KMS] drm report modesetting isn't supported Then, starts to go through a series of fallbacks, deleting each Screen 0 because there was no matching config section. I would appreciate any help you can provide. Thanks.
Re: a year later - CERN move to Centos - what are we doing?
The kickstart disk partitioning tool is even dumber than they new GUI tool, only useful for "one-size-fits-all" cases where you also do not mind accidentally deleting the contents of all disks. (yes, open the machine, disconnect disks, install, reconnect disks, close the machine, thanks, but no thanks). I've started using the disk-by-id option in my kickstarts to target specific disks in disk/fs related directives. A tip I got from Nico who suggested I use the %pre section to configure the disks exactly the way I want to. Works really well. My %pre even contains commands to reset the LVM on re-installs. Many thanks, Nico! Ken
Re: [SCIENTIFIC-LINUX-USERS] LLVM and OCAML
What about libstdc++? Does the epel clang come with its own headers and library? I'm still using sl6x and it "lacks" c++-11 support. I'm wondering if clang is ready to use or whether I have to go through a lot of preprocessor contortions to enable/disable features for clang. Thanks. On 09/08/2015 08:33 AM, Pat Riehecky wrote: I generally utilize LLVM/clang from EPEL. If you need a newer version than is published there, you may be able to get some forward progress by getting the EPEL sourcerpm to build first. Pat On 09/06/2015 10:34 AM, Keith Smith wrote: I'm trying to build LLVM & Clang from sources on SL7. I have installed ocaml, but I still get the following two errors when running ../llvm/configure configure: WARNING: --enable-bindings=ocaml specified, but ctypes is not installed configure: WARNING: --enable-bindings=ocaml specified, but OUnit 2 is not installed. Tests will not run I can't seem to find anything on how to correct these errors. One recommendation for the 'types' error was to run opam, but opam is not in the SL7 or EL distros. I can't find anything on OUnit 2. Thank you for your assistance. I will be posting to the LLVM mailing list as well. Keith Smith
Re: Bizarre bug
I wonder if the loopback shutdown is a red herring. The files nis switch around seems more like a clue. Perhaps some outstanding RPC after the network shutdowns (ethx's are down before the loop, no?) I'd try playing around the shutting down these bits manually in various permutations leaving the system running to see if it hangs. I have SL6.x systems running NIS without problems but then I did not reverse the files nis in nsswitch.conf.
Re: Bizarre bug
Just out of curiosity, why *do* you switch them around? Are you overriding the password/group/etc, content? My NIS maps only contain content that is local to the cluster. Leaves the system accounts, etc, untouched. On 03/03/2015 01:33 PM, P. Larry Nelson wrote: Hi Ken, On 3/3/15 1:06 PM, Ken Teh wrote: I wonder if the loopback shutdown is a red herring. The files nis switch around seems more like a clue. Perhaps some outstanding RPC after the network shutdowns (ethx's are down before the loop, no?) Correct - all other net interfaces go down first. I'd try playing around the shutting down these bits manually in various permutations leaving the system running to see if it hangs. I have SL6.x systems running NIS without problems but then I did not reverse the files nis in nsswitch.conf. Actually, I think that is really the problem (but the why may just have to wait until another day or just file it away in my big file cabinet of unsolved linux weirdities). I took a look at a sampling of some older SL5.x nodes that have been up and running for years and see that they all have 'files nis' order, all apparently without any problems or complications. Thing is, I have in my notes on bringing up a new node, Don't forget to edit nsswitch.conf with 'nis files' order. But the problem is I just don't remember why I wrote that! I'd like to think that my notes are things I figured out once so I don't have to revisit the issue every time. Apparently I didn't figure it out well enough. So, I guess I'll just return to the default order of 'files nis' and forget the whole thing and get some sleep. :-) But it's still a weird bug, which bugs me.. Thanks! - Larry
Re: Ye old Thorn Symbol
And for the vim users, it's Ctrl-k t h in insert mode or ctrl-l T H for the uppercase version of thorn. To expand on Brett's gnome-terminal, it also works the same way in firefox when you want to add a Unicode character in a HTML form, for example. I suspect ctrl-shift-u applies to all gnome apps. On 09/29/2014 05:18 PM, Brett Viren wrote: ToddAndMargo toddandma...@zoho.com writes: I just discovered the discontinued thorn letter. In this day and age, thorn is a critical character as it allows one to form a symmetric smiley-tongue-sticker-outer-face :þ You can enter it using it's code point (0xFE). In a gnome-terminal (or similar): Ctrl-Shift-U f e ENTER In emacs: C-x 8 ENTER f e ENTER -Brett.
Re: How do you speed up rsync?
How big are the 2 directories? I carry a USB stick with about 15GB of data that I synchronize between 2 computers. Usually, it takes about 2 minutes to sync them. Sometimes I sync it to a netbook and it's noticeably slower. Clearly, hardware. But, usually there are not a lot of changes. If there are large changes, like new downloads, then it is much slower. But, definitely, not on the order of 3 hours. On 07/11/2014 02:58 PM, ToddAndMargo wrote: Hi All, I have a bash script for synchronizing a flashing drive (target) with my hard drive (source) I take to customer sites (with a read only switch so I don't spread viruses). I currently rsync 11 different directories. Each sync line looks like this: rsync -rv --delete $MyCDsSource/Linux $MyCDsTarget/.; sync; sync Problem: it is slow -- takes three hours. To help the speed issue, I upgraded from USB 2 to USB 3. Backup went from 3 hr-15 min to 3 hr-5 min. It is almost faster to wipe the stick and rewrite it. Anyone know of a way to speed up rsync? Many thanks, -T
Re: How do you speed up rsync?
My USB stick is formatted in EXT4. I wonder if that makes a difference. I don't recall if rsync checks file timestamps or just the checksum.
Re: Any 7 rumors?
Good grief! These guys just cannot leave well enough alone. On the bright side, this will probably extend the end-of-life for RHEL6x. Rpms of updated tools in /usr/local!!! Rant over... On 05/15/2014 07:59 PM, Nico Kadel-Garcia wrote: There are enough significant layout differences, especially the wholesale switch to systemd and the replacement of /bin with a symlink to /usr/bin that it's going to create a lot of cross compatibility and software porting issues. I'm not looking forward to that part. I'm also afraid to see that I've not yet seen a single reason to *want* it, other than updated libraries for third party software such as perl modules.
java vulnerability
What's the status of the java package that's installed on SL6x? java-1.6.0-openjdk. Is it vulnerable to this java security flaw that made the national news this week? Cyber is advising us to remove it but a lot of packages depend on it. The biggie is LibreOffice. Thanks!
Re: mk2fs hangs at writing superblocks...
Check dmesg. I see a lot of task hung kernel messages with SL6x. First encountered it trying to do a mkfs on a large raid array. Systems run but these messages still show up. Let us know. On 01/04/2013 04:38 AM, Jean-Michel Barbet wrote: On 01/04/2013 11:18 AM, Jean-Michel Barbet wrote: Next test is trying with a more recent SL5 (5.8). Same with SL5.8 DVD... JM
Re: cernlib for SL 6x
Hi Valery, Thanks for the info. I am confused about the 2006-34/2006-35 numbers. The pattern is that g77 packages are 2006-34 while the ones not labeled g77 are 2006-35. Except for paw and packlib. There seems to be a 2006-34 packages without the g77 label and a third 2006-35 package labeled gfortran. Since SL6 comes only with gfortran I assume that the ones not labeled g77 are built with gfortran but the fact that there is a paw and packlib with no compiler label but a 2006-34 version number and a separate paw and packlib with a gfortran 2006-35 version... Do you know what the difference is? Which one should I install? Thanks. Ken On 11/21/2012 11:39 AM, Valery Mitsyn wrote: On Wed, 21 Nov 2012, Ken Teh wrote: What do folks do about installing the cern program libraries for SL6? I see that they only have pre-built binaries for SL5. Are you building them from source or is there an semi-official repo you can get them from? it's in official epel repo, they split it for many rpms. # yum -d 1 list all \*2006-\* Available Packages cernlib.i6862006-35.el6 epel cernlib.x86_64 2006-35.el6 epel cernlib-devel.i686 2006-35.el6 epel cernlib-devel.x86_642006-35.el6 epel cernlib-g77.i6862006-34.el6 epel cernlib-g77.x86_64 2006-34.el6 epel cernlib-g77-devel.i686 2006-34.el6 epel cernlib-g77-devel.x86_642006-34.el6 epel cernlib-g77-static.x86_64 2006-34.el6 epel cernlib-g77-utils.x86_642006-34.el6 epel cernlib-packlib.x86_64 2006-34.el6 epel cernlib-packlib-g77.x86_64 2006-34.el6 epel cernlib-packlib-gfortran.x86_64 2006-35.el6 epel cernlib-static.x86_64 2006-35.el6 epel cernlib-utils.x86_642006-35.el6 epel geant321.x86_64 2006-35.el6 epel geant321-g77.x86_64 2006-34.el6 epel kuipc.x86_642006-35.el6 epel kuipc-g77.x86_642006-34.el6 epel patchy.x86_64 2006-34.el6 epel patchy-g77.x86_64 2006-34.el6 epel patchy-gfortran.x86_64 2006-35.el6 epel paw.x86_64 2006-34.el6 epel paw-g77.x86_64 2006-34.el6 epel paw-gfortran.x86_64 2006-35.el6 epel Thanks!
cernlib for SL 6x
What do folks do about installing the cern program libraries for SL6? I see that they only have pre-built binaries for SL5. Are you building them from source or is there an semi-official repo you can get them from? Thanks!
Re: task blocked for more than 120 seconds
Thanks Nico! Anyone else want to comment? I'd like to hear people's opinions about such issues. When one is not 100% involved in such efforts, it helps to have input from people who are or have dealt with similar issues. On a related note, this is probably the last disk array I want to put together myself. I discovered to my exasperation that I needed enterprise class drives because of the TLER effect. The next disk array I buy will be a ready-made appliance. On 11/14/2012 08:58 PM, Nico Kadel-Garcia wrote: On Wed, Nov 14, 2012 at 12:37 PM, Stephen John Smoogen smo...@gmail.com wrote: On 14 November 2012 10:20, Ken Teh t...@anl.gov wrote: The common thread is I/O to a MegaRAID raid5 device. Which is cause for concern since the primary function of both machines where I've encountered this problem is file-serving. Perhaps I am just unlucky and have 2 bad MegaRAID cards in a row. I'm trying to understand this better, figure out if I am doing something wrong. Well there are a couple of issues this could be: 1) You are asking more than the MegaRaid is meant to do... it may be running out of cache, or other resources. My experience with MegaRAID has been *horrible*. Poor driver compatibility, awkward and destructive firmware, and deceitful specifications only start the list of horrible failures. Their best technological use is as doorstops. 2) The megaraid is still rebuilding its array beneath and you are hitting a locking problem because it hasn't finished what it needs to do before you ask it to do something else (really sort of #1). Most of the time you will need to install the proprietary Megaraid tools to see what is going on under the disks to find out. See above: good luck getting those tools working! Every hour you spend waiting on those things to come to their senses, or trying to debug them, is an hour wasted on problems that may not ever be solved by your efforts. I encourage you to replace them with a better quality manufacturer: Adaptec makes very solid, not too expensive controllers, and Rocketport remains the cream of the crop.
Re: ext4 stride and stripe-width for a hardware RAID5
Well, I tried 3 scenarios with the stride/stripe-width settings: (1) None, mkfs.ext4 with defaults. (2) Using the -E option to set the stride/stripe-width to match the disk array configuration. (3) Using LVM with defaults. There was no difference in writing to the disk. On 11/06/2012 11:21 AM, Ken Teh wrote: I'm wondering if anyone has tried using stride and stripe-width options when creating an ext4 filesystem on a hardware RAID5 array. Does it improve the performance of the array? Should you create the filesystem directly on the partition? What happens if you create it on a logical volume that is created on the physical partition? The LVM volume was created without options, ie, defaults. I know LVM supports stride and stripes as well but I dont know how they map to a physical RAID device so I've never bothered with it.
task blocked for more than 120 seconds
I've recently been encountering this problem trying to stand up a large RAID 5 disk server. My first encounter was when I was doing write speed tests. I thought I had solved this problem by letting the megaraid card complete a slow init of the volume before trying to create a linux filesystem on it and re-doing my speed measurements. But I have just now encountered it again on a new RAID 5 volume which I also let complete a slow init over the weekend. I was in fact trying to do a pvcreate on the volume when it hung. Can anyone shed some light? I see posts for it but everything I read suggests it's been taken care of.
Re: task blocked for more than 120 seconds
The common thread is I/O to a MegaRAID raid5 device. Which is cause for concern since the primary function of both machines where I've encountered this problem is file-serving. Perhaps I am just unlucky and have 2 bad MegaRAID cards in a row. I'm trying to understand this better, figure out if I am doing something wrong. My procedure is create a RAID 5 volume on the megaraid, do a slow init. Reboot the system into Linux, write a single large partition with parted, then put one or more logical volumes on the drive. The hung problem has cropped up under the following situations: (1) pvcreate on the disk (2) mkfs.ext4 on the volumes created on the disk (3) writes to the filesystem on the disk It's happened on 2 fileservers each with a megaraid. On 11/14/2012 10:19 AM, Jamie Duncan wrote: is there a specific bug/bugs you're referring to? a hung task means that a process is sitting on a core waiting on a specific bit of I/O for 120 seconds. Not the length of the entire process, mind you, which depends on countless inputs and outputs to complete, but something on the other side isn't answering for a very long time. It usually means an unhealthy system at some level. On Wed, Nov 14, 2012 at 11:04 AM, Ken Teh t...@anl.gov mailto:t...@anl.gov wrote: I've recently been encountering this problem trying to stand up a large RAID 5 disk server. My first encounter was when I was doing write speed tests. I thought I had solved this problem by letting the megaraid card complete a slow init of the volume before trying to create a linux filesystem on it and re-doing my speed measurements. But I have just now encountered it again on a new RAID 5 volume which I also let complete a slow init over the weekend. I was in fact trying to do a pvcreate on the volume when it hung. Can anyone shed some light? I see posts for it but everything I read suggests it's been taken care of. -- Thanks, Jamie Duncan 804.571.0458
ext4 stride and stripe-width for a hardware RAID5
I'm wondering if anyone has tried using stride and stripe-width options when creating an ext4 filesystem on a hardware RAID5 array. Does it improve the performance of the array? Should you create the filesystem directly on the partition? What happens if you create it on a logical volume that is created on the physical partition? The LVM volume was created without options, ie, defaults. I know LVM supports stride and stripes as well but I dont know how they map to a physical RAID device so I've never bothered with it.
Re: mt and LTO
Thanks Nico for your reply. Especially the historical context. I always find I understand something better, know how to make choices, etc., with this information. Glad to know I'm still cool. Retro cool, perhaps...but still cool. :) On 11/01/2012 04:30 AM, Nico Kadel-Garcia wrote: On Thu, Nov 1, 2012 at 3:27 AM, FSG WD Andre Paetzold paetz...@fsg-ship.de wrote: On Wed, Oct 31, 2012 at 13:27:49 -0500, Ken Teh wrote: Do you still use mt and tar for backups on LTO drives or is that so uncool compared to bacula? Does that mean I'm uncool if I confess I still do? ;) mt and tar are built into most open source tape management solutions. I ported Amanda to Solaris long, long, long ago, and it still works well. It's also available with commercial support at www.zmanda.org Yes, we even still use 'find | cpio' ... ;) But nowadays, when the files become bigger than 8GB, we must switch our scripts to 'tar' To increase the performance of the tape-drive, we use 'tar --blocking-factor=2048' to write with Blocksize of 1MB, but with our new HP LTO5 Tape Blade (connected via SAS) we got input/output errors, so we must switch to Blocksize of 512KB... :( On our HP LTO5 MSL4048 (connected via FibreChannel) it is still possible to use 1MB Blocksize. Oh, getting down into the guts! I still remember my lamentations that the Exabyte drives had no EOT marker, the end-of-tape marker used to show where writing had stopped, which was usually two EOF or end-of-file markers in a row. I went nuts writing wrap-arounds to find the end of the tape to write the next dump on a tape I'd just inserted The problem was that I had a closet full of old magtape with irreplaceable research data, the tape drives the oldest tapes were compatible with were going away, and I needed the material on new media where it would be stored. Enter the Amanda software, and a chunk of time adapting it to the Exabytes. It very effectively uses tar (and now, star for SELinux settings as well) to write the tapes, and *the technology is stable*. Unlike many commercial solutions, any idiot 10 years from now can still read the tapes even without an expensive sotware package that may not run on any OS they own. These days, I first dump to disk with rsnapshot, because large, consumer grade disk is so cheap now. That lets me keep the last few days or weeks of snapshots and make them available to users without having to give them tape access, they can just access their NFS shares from the cheap storage. And I can run the tape against *that*. Do you still need mt if you use a package like bacula? We need 'mt' to rewind the tape in case of an error, to retry it, and at the end of the backup to eject the tape from drive, so the operators only have to grab the tape or the see there was an error, if the tape isn't ejected... Binigo. Thanks! So long, André -- FLENSBURGER SCHIFFBAU-GESELLSCHAFT mbH Co. KG Batteriestrasse 52, 24939 Flensburg Sitz der Gesellschaft / Place of business : Flensburg Geschaeftsfuehrer / CEO : Peter Sierk Handelregister / Commercial register : Amtsgericht Flensburg, HRA 3140 Steuer Nr. / Tax number: 1528040009 USt.-Id-Nr. / VAT no : DE 134633705 Amtsgericht / District court : Flensburg HRB 2036 The information contained in this email message may be privileged and confidential. If you have received this communication in error, please notify the sender immediately by telephoning (+49-461-49400) and return this message to the above address. Thank you.
mt and LTO
Do you still use mt and tar for backups on LTO drives or is that so uncool compared to bacula? Does that mean I'm uncool if I confess I still do? ;) Do you still need mt if you use a package like bacula? Thanks!
disk recommendations
I've run into problems trying to use desktop disks in a RAID array with a MegaRAID 9260-8i. I built 2 previous systems with desktop disks and did not have any problems but I've been unable to get this 3rd system to function stably. Disks dropped from the array except the disks are fine which I proved by deleting the disk group and starting again. This behaviour is consistent with Western Digital's discussion about TLER (time-limited error recovery) and their admonition to use enterprise disks instead of desktop disks. I was wondering if there are ways of getting around this problem without having to buy another set of disks. Googling suggests some options to try but I thought I'd used the list as a sounding board first before embarking on options. Thanks!
Re: has xdvi been superseded?
I know it's in the repo. Since it was not installed when installing TeX, I was wondering if xdvi has become an optional package. That, perhaps, it's been superseded by another previewer. I have not kept up with TeX developments and was hoping someone who has may know more. On 07/04/2012 05:25 AM, Dr Andrew C Aitchison wrote: On Tue, 3 Jul 2012, Ken Teh wrote: xdvipdfmx is not installed as part of a TeX install on SL6. Has xdvi has been superseded by another previewer? Hm. I see xdvipdfmx in the SL6.2 repo, eg ftp://ftp.scientificlinux.org/linux/scientific/6.2/x86_64/os/Packages/xdvipdfmx-0.4-5.1.el6.x86_64.rpm There is an xdvik package for SL6 in the epel repo with an xdvi command.
LTO tape drives
Hi all, I need some help and advice with LTO tape drives. I haven't kept up with tape technology for several years now and from a quick google, it appears that the only games in town are LTO and DDS. We do have a single DDS deployment and we are looking at replacing it with LTO because of capacity limitations. Some questions: (1) It appears that LTO-5 is the current technology. But most single drives use SAS as an interface. Again, I am not familiar with SAS. All I know is what the acronym stands for. Is it a simple matter of buying a SAS card? Will it run out of the box on a SL6.x box? Any issues with drivers? (2) I see that LTO-4 is still available. And a quick google suggests that Ultra160 or 320 LVD SCSI is the more common interface for these drives. I have several of the Adaptec 29xxx cards, made available when we transitioned to SATA. Is this viable or would you recommend against using LTO-4? It appears to be about 5 years old. I think it would be preferable to use technology that has at least some life to it but I am not inclined to work very hard to make LTO-5 work if it is not well supported on Linux. We are not looking to do anything fancy. A single internal drive and a box of tapes. Backups the old fashioned way. Specific recommendations welcomed. As in, buy this card, buy this drive! Thanks!
Re: network manager questions
TUV needs an option in kickstart to turn off NM for designated cards. Btw, NM_CONTROLLED=no in ifcfg-eth0 is not sufficient. When you do this, I lose DNS as well since apparently, NetworkManager usurps dhcp-client's role in this. When I chkconfig NetworkManager off, everything works. So I now have that in my kickstart script for my desktops. Is Enterprise Linux mostly installed on laptops? I would have thought that desktops still make a large fraction of its deployment. In fact I would almost bet on it since Linux is still not trouble-free when it comes to installing on laptops. In which case, it seems like a really bad idea to foist the NetworkManager on people. On 04/06/2012 01:28 PM, Konstantin Olchanski wrote: On Thu, Apr 05, 2012 at 02:46:50PM -0500, Ken Teh wrote: Is it true that the network manager service turns off the network when there is no activity? Think of it this way. The network manager was invented to handle Wifi on laptops. (And it works well enough for that). When used for any other purpose, well, what do you expect, it was not invented for that. This is the new-think engineering through the but it works on my laptop! paradigm. In practice. I use the network manager on most server-type computers (as it comes pre-installed, pre-enabled with SL6). As long as you remember to open the network connection editor and enable the right available to all users and enable on boot buttons, it seems to work well enough, as long as nothing goes wrong. When things go wrong in server-type-specific ways, well the network manager does not know what to do, even for the simplest cases, like the network link going down for an hour (maintenance of UPS power to the network switch). I have seen it drop it's IP address and never ask for another one (should have issued a DHCP request when network link came back up, then maybe should have tried again every hour). I have also seen the network manager drop it's IP address (and never ask for another one) after an eth device hang (eth chip vs driver compatibility) when a simple ifconfig down/up would have recovered the system. I tend to think that these days one should go back to static IP addresses for server-type machines, after all, all DHCP, network manager co do is assign the same IP address to the same machine over and over and over again with the only variation when they fail to do the boring thing and you have a machine down, staying down until somebody physically walks to it to reboot it.
network manager questions
Is it true that the network manager service turns off the network when there is no activity? I just discovered that my desktops lost connection to the authentication server. So, screen locks, gdm logins, remote ssh just stopped working. Only when I logged in as root on the console did I notice that the network was disabled and when I clicked on the the panel icon, it reconnected. Is there some special config for the network manager to stop it from doing this? Is there a kickstart option for this as well? I have, for the moment, disabled the network manager service and edited the ifcfg-ethX files so they are no longer controlled by NM. I did this for my headless servers, but I was surprised (annoyed is more accurate) that this also affects standard client desktops.
adobe yum repo
I just found out there is an adobe yum repo where you can keep the flash-plugin up-todate. Is any one using it?
httpd-2.2.3-63.el5-8.1 and cve 2012-0053
Is this update to httpd available? Supposedly addresses cve 2012-0053.
Re: xrandr is driving me batty
I can add modelines to whatever is running because xrandr reports them. When I try to set them with xrandr --output default --mode modename I always get 'configure crtc 0 failed' There's also some esoteric message when I first load the modeline, could not set crtc 262 So, still no go on xrandr. And it's skimpy on what it reports. Avoid Viewsonics? Stiffer drink? Would appreciate any other insights. On 01/27/2012 06:03 PM, Orion Poplawski wrote: On 01/27/2012 04:37 PM, Ken Teh wrote: The new method of configuring displays doesnt seem to be an improvement over the venerable system-config-display. I have several monitors that it doesn't seem to know. Viewsonics in particular and I end up with 800x600 or if it's generous 1024x768 resolutions. When the monitor can do 1920x1200. What are folks doing to get around this problem besides having a stiff drink? Perhaps: xrandr --addmode VGA1 1920x1200 replace VGA1 with the appropriate output name.
Re: xrandr is driving me batty
That was also what I was doing. Removing xorg.conf and *hope* the autoconfig works. I don't have any problems using proprietary drivers. Problem is that they stop working every time there is a kernel upgrade. At least this is my experience and I have to rebuild them again. For a few machines, this is not a problem. But when you have to do this for 50 machines... On 01/27/2012 07:58 PM, Konstantin Olchanski wrote: On Fri, Jan 27, 2012 at 05:37:01PM -0600, Ken Teh wrote: The new method of configuring displays doesnt seem to be an improvement over the venerable system-config-display. I did not know there is a new method for this. After system-config-display was removed, I though the only way is either autoconfigure (rm xorg.conf), or vi xorg.conf. But no matter, I specify NVIDIA graphics for all purchases, install the NVIDIA proprietary driver (from NVIDIA or from ELREPO) and use nvidia-settings to configure the display (including multiple display configurations). So which tool are you using?
xrandr is driving me batty
The new method of configuring displays doesnt seem to be an improvement over the venerable system-config-display. I have several monitors that it doesn't seem to know. Viewsonics in particular and I end up with 800x600 or if it's generous 1024x768 resolutions. When the monitor can do 1920x1200. What are folks doing to get around this problem besides having a stiff drink?
epel repo enabled ok?
Quick question? I needed some packages from epel so I added the epel rpms from SL6x. Is it okay to leave the yum.repos.d/epel.repo enabled? Are there packages on epel that will clobber the SL6x packages during the nightly updates? Thanks!
SL6x: sbin in path??
Is there some reason why regular users get /usr/local/sbin, /sbin, /usr/sbin in their PATH? SL5x didnt do it and I believe that was also the case for SL4x. Can I remove it safely?
kernel version syntax
I want to confirm the following assumptions I have about the kernel version numbers obtained with 'uname -r' That, with each release, the version is 2.6.x-blah-blah-blah where x is fixed for a given release. For instance SL5.x is always 2.6.18 while SL6.x is always 2.6.32. The series of numbers that follow the hyphen is some other beastie that is part of the distro and not the original kernel version that Linus releases.
Fwd: Re: Farewell from Troy
Will the add-on rpms still be signed by Spiky Hair? :-) Many thanks for all your efforts. You guys are the best. Ken On 08/24/2011 01:40 PM, Troy Dawson wrote: Hi, I have loved all the years that I have been a developer and architect for Scientific Linux, but it is time for me to move on. I have accepted a job offer from Red Hat to work on their new openshift project. ( https://www.redhat.com/openshift/ ) My last day working for Fermilab, and on the Scientific Linux project will be September 2, 2011. Thank you to everyone who has encouraged, thanked, and helped me over the past 8 years that I have worked on Scientific Linux. I have said it before, and I'll say it now, The Scientific Linux community is one of the best communities there is. Troy
sd?
How does the kernel name the sata drives? Seems to change around on reboot. I have a machine with 2 sata drives that are mirrored. And a large disk array on a megaraid card which makes up the third drive on the machine. I reinstalled the machine by first pulling out the megaraid card. Installed SL6.1 on my RAID1 mirrors. I saved the partition info from sda and sdb, grubbed both disks, then shut the machine down to reinstall the megaraid card. Reboot the machine and now the megaraid drive is sda, and my system disks are sdb and sdc. I realize that the system mounts the disks with uuid's so it really doesn't matter. It just means I have to make a note of it in the RAID1 info I saved so several years from now when I really have to replace a mirror member, I will remember that sdb became sdc or did sdb stay as sdb and sda became sdc.
Re: SL6.1 kickstart persistent networkmanager prevents unattended install
Can you tell us what you type (the kernel parameters) in order to specify that you are installing from a kickstart? On 08/10/2011 02:18 PM, Johnson, Kent A (GE Healthcare) wrote: Greetings, Sorry to bother but I've tried to do my homework first, to no avail... I'm kickstarting the standard SL6.1 install DVD from a USB key to provide an unattended install. I've tried everything I can find to try to prevent the networkmanager popup to configure the ethernet interface during the install (occurs immediately after disk partitioning). I've tried many different forms of the kickstart network command to fully configure eth0 or leaving the network line out completely but the nm popup persists. The only way that I've been able to stop the nm popup during the kickstart install is to disable the ethernet hardware in the BIOS! I've read all the RH/SL docs and searched the web far and wide... So, it seems that anaconda starts networkmanager any time that it finds an ethernet device (active or not) whether the kickstart references network or not. I'm not adding any external repos or anything else that would require network. What am I missing? thanks in advance for any guidance, Kent
Re: SL6.1 kickstart persistent networkmanager prevents unattended install
I've never done one of these before so I really can't help. I usually serve my kickstart via nfs in which case I want the network manager to set up the nic with dhcp. If the machine is to be a server, I will override the network config in the kickstart. My guess is that you have to mount the boot.iso image that's distributed with SL6x, read-write, on a loopback, and copy your ks.cfg file to it. That is, the boot.iso image needs to be modified. It's not enough to to copy write the iso and the ks.cfg to the usb key. Or you need to modify the initrd image that's on the boot.iso image. It's not clear from the documentation which root (/) it refers to - the syslinux image or the initrd image. My guess is it's the initrd image. Is this what you did? On 08/10/2011 02:44 PM, Johnson, Kent A (GE Healthcare) wrote: I'm booting from the ISO DVD and entering TAB at the install splash screen and then adding ks=hd:/dev/sdb1:/ks.cfg to read the kickstart from a USB key. The kickstart is below (the latest attempt)... #platform=x86, AMD64, or Intel EM64T #version=DEVEL # Firewall configuration firewall --enabled --ssh # Install OS instead of upgrade install # Use CDROM installation media cdrom # Root password rootpw --iscrypted $1$Qp/kIHFC$d/jWHGmcg5jeiwIAt/82P1 # System authorization information auth --useshadow --passalgo=sha512 # Use text mode install text firstboot --disable # System keyboard keyboard us # System language lang en_US # SELinux configuration selinux --disabled # Installation logging level logging --level=info # Reboot after installation reboot # System timezone timezone --isUtc America/Chicago # System bootloader configuration bootloader --location=mbr # Clear the Master Boot Record zerombr # Partition clearing information clearpart --drives=sda --all # Disk partitioning information part /boot --fstype=ext4 --ondisk=sda --size=700 part / --fstype=ext4 --ondisk=sda --size=2 part /export/home1 --fstype=ext4 --ondisk=sda --size=5 part swap --fstype=swap --ondisk=sda --size=4096 %packages @additional-devel @base @basic-desktop @client-mgmt-tools @compat-libraries @debugging @desktop-debugging @desktop-platform @desktop-platform-devel @development @directory-client @eclipse @fedora-packager @fonts @general-desktop @graphical-admin-tools @ha @ha-management @hardware-monitoring @ice-desktop @infiniband @input-methods @internet-applications @internet-browser @java-platform @large-systems @legacy-unix @legacy-x @load-balancer @misc-sl @network-file-system-client @network-tools @nfs-file-server @performance @perl-runtime @print-client @scalable-file-systems @scientific @security-tools @server-platform @server-platform-devel @spins @storage-client-fcoe @storage-client-iscsi @storage-client-multipath @system-admin-tools @system-management @system-management-messaging-client @tex @virtualization @virtualization-client @virtualization-platform @virtualization-tools @x11 %end thanks regards, kent -Original Message- From: Ken Teh [mailto:t...@anl.gov] Sent: Wednesday, August 10, 2011 2:38 PM To: Johnson, Kent A (GE Healthcare) Cc: scientific-linux-us...@fnal.gov Subject: Re: SL6.1 kickstart persistent networkmanager prevents unattended install Can you tell us what you type (the kernel parameters) in order to specify that you are installing from a kickstart? On 08/10/2011 02:18 PM, Johnson, Kent A (GE Healthcare) wrote: Greetings, Sorry to bother but I've tried to do my homework first, to no avail... I'm kickstarting the standard SL6.1 install DVD from a USB key to provide an unattended install. I've tried everything I can find to try to prevent the networkmanager popup to configure the ethernet interface during the install (occurs immediately after disk partitioning). I've tried many different forms of the kickstart network command to fully configure eth0 or leaving the network line out completely but the nm popup persists. The only way that I've been able to stop the nm popup during the kickstart install is to disable the ethernet hardware in the BIOS! I've read all the RH/SL docs and searched the web far and wide... So, it seems that anaconda starts networkmanager any time that it finds an ethernet device (active or not) whether the kickstart references network or not. I'm not adding any external repos or anything else that would require network. What am I missing? thanks in advance for any guidance, Kent
6.1 kickstart is characterless
The installation screens during a 6.1 kickstart install have no words on them. It's like it was not able to load a font table so the screens, dialogs, buttons, are all blank. The install works. I'm not particularly bothered by it but thought you might want to know.
Re: 6.1 kickstart is characterless
Yes, it seems to be video hardware. Everything is fine if I choose the basic video install instead of the standard install. On 08/09/2011 12:14 PM, Christopher Tooley wrote: On 2011-08-09, at 9:17 AM, Jos Vos wrote: On Tue, Aug 09, 2011 at 10:17:31AM -0500, Ken Teh wrote: The installation screens during a 6.1 kickstart install have no words on them. It's like it was not able to load a font table so the screens, dialogs, buttons, are all blank. The install works. I'm not particularly bothered by it but thought you might want to know. I've done several SL 6.1 kickstart installs in the last week, but I never saw anything wrong, all info was displayed as usual. Having said that, when installing with kickstart I don't have dialogs, but buttons and other text I see during the install all look fine. -- --Jos Vosj...@xos.nl --X/OS Experts in Open Systems BV | Phone: +31 20 6938364 --Amsterdam, The Netherlands| Fax: +31 20 6948204 This may be a long shot, but there may be something wrong with your video card. I've had garbled text on a terminal and it was due to my video card dying for some reason. Christopher Tooley ctoo...@uvic.ca Systems, HEP/Astronomy UVic
where is autofs.schema?
The redhat/autofs.schema is not in /etc/openldap/schema. It was there in 6.0 but it's not there in 6.1. Where did it go? Ken
Re: where is autofs.schema?
Agreed, my apologies. /usr/share/doc/autofs-5.0.5 I found it with a 'locate autofs.schema'. Should have thought of it earlier. On 08/02/2011 12:56 PM, Chris Tooley wrote: Perhaps as a service for those who are looking for it in the future you could post where you found it? -Chris On 11-08-02 10:29 AM, Ken Teh wrote: Never mind. I found it. See what I mean about these bits of stuff darting all over the place. I can perhaps understand it with a major version update. But a release update? On 08/02/2011 12:03 PM, Ken Teh wrote: The redhat/autofs.schema is not in /etc/openldap/schema. It was there in 6.0 but it's not there in 6.1. Where did it go? Ken
Re: The Clone Wars – CentOS vs. Scientifi c Linux
On 07/27/2011 04:43 AM, Tom H wrote: On Wed, Jul 27, 2011 at 3:25 AM, Andreas Petzold andreas.petz...@kit.edu wrote: Can we please stop emails like this? It's getting really annoying. This is a technical support mailing list. +1 Second, thirdPlease stop. Use some other forum for discussions like this.
kickstart: lvm on an md device
Is it possible to kickstart an installation and have / constructed on an lvm volume that is itself constructed on top of an md RAID 1 device?
Re: kickstart: lvm on an md device
Never mind..I found the example in the TUV's documentation. Sorry. On 07/18/2011 05:35 PM, Ken Teh wrote: Is it possible to kickstart an installation and have / constructed on an lvm volume that is itself constructed on top of an md RAID 1 device?
Re: NM_CONTROLLED?
I suggest turning off the NetworkManager service if it is appropriate. I set this parameter to no thinking it would be sufficient but then I started getting all sorts of weird behaviors like the network interface would stop working. To be honest, I did not investigate but then I was setting the machine up as a server and wanted the network interface on and to stay on after the machine boots. On 07/05/2011 12:43 PM, Tom H wrote: On Tue, Jul 5, 2011 at 1:19 PM, Todd And Margo Chester toddandma...@gmail.com wrote: In SL 6.0, 64 bit, I have the following parameter in my ifcfg-ethxxx called: NM_CONTROLLED=yes What is this? NM_CONTROLLED == controlled by Network Manager It's only effective if set to no in order not to have the NIC under NM's control.
non-stop printer notification
I have a user who's print job was cancelled and he now gets dialog box popping up every minute telling him so and suggesting he find out why. And little else. Sound familiar? Is there a simple way of shutting up the machine short of shutting it down?
Re: non-stop printer notification
I found 500 of these in the queue on the print server, cancelled all of them. They are no longer listed but it still pops up on the client machine. Ken On 06/14/2011 02:10 PM, Chris Tooley wrote: On 11-06-14 11:59 AM, Ken Teh wrote: I have a user who's print job was cancelled and he now gets dialog box popping up every minute telling him so and suggesting he find out why. And little else. Sound familiar? Is there a simple way of shutting up the machine short of shutting it down? You can try clearing the printer queue... Perhaps it's continually trying to print and failing? Or, try restarting CUPS: /etc/init.d/cups restart -Chris
Re: kickstart resolv.conf problem with SL6
You can use DNS names in the script sections of the kickstart. It's documented in the TUV's 6.0 manuals. Don't forget to use -o nolock if you are mounting NFS. On 04/05/2011 04:28 AM, Ahmed El Zein wrote: - Original Message - From: Ahmed El Zein ahmed.elz...@anu.edu.au Date: Tuesday, April 5, 2011 6:23 pm Subject: kickstart resolv.conf problem with SL6 To: scientific-linux-us...@fnal.gov Dear SL users,I am testing SL6 and my kickstart scripts are failing. Specifically wgets are unable to resolve the hostname of the spacewalk server. I took a look and both /etc/resolv.conf and /mnt/sysimage/etc/resolv.conf are missing the nameserver line. They both just have a single line: search domain -- These kickstart scipts work just find on my Centos 5.5 systems, so I am not sure if this is a SL6 issue or some change in the kickstart options that I am not aware of. I have looked but I can't find any relevant difference between the RHEL5 and RHEL6 kickstart doco. And help/info in this regard would be much appreciated. A quick follow up; After some more testing I noticed that when anaconda first runs the resolv.conf file is fine. By the time the pre script runs the resolv.conf is just the one line as I described it. Thanks, Ahmed
Re: gdm and PreSession/Default
I solved this problem. For those of you who are required to put up the DOE warning banner or any other sort of disclaimer that aborts a gui login, the solution is to put the attached file in /etc/X11/xinit/xinitrc.d. The scripts in xinitrc.d are run from xinitrc-common. The distro's first script is 00-start-message-bus.sh. Choose a name so that the DOE warning script is run before it. No point starting up dbus if the user chooses to decline the disclaimer and abort the login. On 03/08/2011 12:41 PM, Ken Teh wrote: SL6's gdm does not honor the exit code from /etc/gdm/PreSession/Default. In SL5x, it used to abort the login if the script returns a non-zero exit code. I used this feature to put up a zenity dialog which the user had to click yes in order to continue logging in. Now, the user logs regardless of the script. In fact, the distro's Default script is empty. It used to contain the xsetroot and sessreg bits. Any workarounds? === WARNING: This e-mail has been altered by MIMEDefang. Following this paragraph are indications of the actual changes made. For more information about your site's MIMEDefang policy, contact MIMEDefang Administrator postmaster@localhost. For more information about MIMEDefang, see: http://www.roaringpenguin.com/mimedefang/enduser.php3 An attachment named '000-doewarning.sh' was converted to 'defang-1.binary'. To recover the file, right-click on the attachment and Save As '000-doewarning.sh' defang-1.binary Description: defang-1.binary
nfs kickstart scripts
The install and boot iso's no longer issue a boot prompt. I used to make use of this feature in SL4 and SL5 in order to download a kickstart from an nfs server. The kickstarts would typically specify our local mirror as a install source. It also allowed to customize kickstarts for particular machine configs. Is this still possible in SL6? Is it done differently?
Re: nfs kickstart scripts
A little more detail: The iso images for SL4 and SL5 used to stop with the boot: prompt after loading the kernel. I would then specify boot: linux=nfs:server:/path/to/kickstart On 03/09/2011 05:32 PM, Ken Teh wrote: The install and boot iso's no longer issue a boot prompt. I used to make use of this feature in SL4 and SL5 in order to download a kickstart from an nfs server. The kickstarts would typically specify our local mirror as a install source. It also allowed to customize kickstarts for particular machine configs. Is this still possible in SL6? Is it done differently?
Re: nfs kickstart scripts
Yes, I've tried it. Now with all the iso images. None of them gives a boot prompt. I should correct my previous message. The boot prompt used to show up after syslinux loads, not the kernel. According to TUV docs, the prompt should show, but it doesn't. On 03/09/2011 06:02 PM, Bluejay Adametz wrote: A little more detail: The iso images for SL4 and SL5 used to stop with the boot: prompt after loading the kernel. I would then specify boot: linux=nfs:server:/path/to/kickstart Have you tried hittingtab when the boot screen comes up? That seems to allow modifying the boot options and may do what you want. There is a note to this effect on the boot screen, but it's not real obvious. I missed it for a while. - Bluejay Adametz, CFII, AP, AA-5B N45210 Be careful what you teach You might have to learn it one day. -Tunnell's Terse Transmogrification of Fido Fisher's Fortuitious Formulary NOTICE: This message, including any attachments, is only for the use of the intended recipient(s) and may contain confidential and privileged information, or information otherwise protected from disclosure by law. If the reader of this message is not the intended recipient, you are hereby notified that any use, disclosure, copying, dissemination or distribution of this message or any of its attachments is strictly prohibited. If you received this message in error, please contact the sender immediately by reply email and destroy this message, including all attachments, and any copies thereof.
Re: nfs kickstart scripts
Ok, I got it. You were right. The tab option did the trick. The ks spec ks=nfs:server:/path/to/kickstart needs to be added to the kernel boot options. It says so in the TUV docs, but it also says to type this in at the boot prompt. Except, of course, there is no boot prompt unlike SL4x and SL5x. On 03/09/2011 06:02 PM, Bluejay Adametz wrote: A little more detail: The iso images for SL4 and SL5 used to stop with the boot: prompt after loading the kernel. I would then specify boot: linux=nfs:server:/path/to/kickstart Have you tried hittingtab when the boot screen comes up? That seems to allow modifying the boot options and may do what you want. There is a note to this effect on the boot screen, but it's not real obvious. I missed it for a while. - Bluejay Adametz, CFII, AP, AA-5B N45210 Be careful what you teach You might have to learn it one day. -Tunnell's Terse Transmogrification of Fido Fisher's Fortuitious Formulary NOTICE: This message, including any attachments, is only for the use of the intended recipient(s) and may contain confidential and privileged information, or information otherwise protected from disclosure by law. If the reader of this message is not the intended recipient, you are hereby notified that any use, disclosure, copying, dissemination or distribution of this message or any of its attachments is strictly prohibited. If you received this message in error, please contact the sender immediately by reply email and destroy this message, including all attachments, and any copies thereof.
gdm and PreSession/Default
SL6's gdm does not honor the exit code from /etc/gdm/PreSession/Default. In SL5x, it used to abort the login if the script returns a non-zero exit code. I used this feature to put up a zenity dialog which the user had to click yes in order to continue logging in. Now, the user logs regardless of the script. In fact, the distro's Default script is empty. It used to contain the xsetroot and sessreg bits. Any workarounds?
6.0 installation problems
I'm trying to install 6.0 x86_64 on a machine using the enterprise storage option. The machine hosts a 16TB disk array. The OS is to be installed on a software raid set of smaller disks separate from the 16TB array. A couple of questions: (1) Is the enterprise storage option the right option? I'm not using any of the FCoE, iSCSI, etc., options. Just plain hardware raid 5. (2) I'm having trouble with the graphical install. The network dialog is too large for my screen and I cannot see the buttons at the bottom of the dialog box. Is there a text-based install? Appreciate any help you may have. Thanks Ken
Re: 6.0 installation problems
The video problem was solved using the second option on the install list - ie installing with the basic video driver. On 03/03/2011 03:13 PM, Troy Dawson wrote: On 03/03/2011 02:55 PM, Ken Teh wrote: I'm trying to install 6.0 x86_64 on a machine using the enterprise storage option. The machine hosts a 16TB disk array. The OS is to be installed on a software raid set of smaller disks separate from the 16TB array. A couple of questions: (1) Is the enterprise storage option the right option? I'm not using any of the FCoE, iSCSI, etc., options. Just plain hardware raid 5. I don't know about any enterprise storage option so I'm not sure what to say. Scalable Filesystem is really just XFS support (2) I'm having trouble with the graphical install. The network dialog is too large for my screen and I cannot see the buttons at the bottom of the dialog box. Is there a text-based install? Installation instructions are here http://www.scientificlinux.org/distributions/6x/installing/ I suggest that you try to find some option to change the video resolution. You really don't want to try text install. Text install instructions are here http://www.scientificlinux.org/distributions/6x/installing/install-guide-text-full.html NOTES about text install * You are unable to do customized partitions in text mode. * You are unable to review and edit partitions in text mode. * You are unable to put your boot loader anyplace special in text mode. * You are unable to select packages and/or groups in text mode. You only get a minimal install. This is the way the installer comes from The Upstream Vendor. Troy
apc or tripplite or something else
Is APC (or some other make) preferred over tripplite as a linux friendly ups? Is everyone using nut or are the vendor supplied software good enough? I have only used tripplite and nut and it has been less than satisfactory. It's always a struggle to put one together. Thanks!
compiling a newer kernel for SL5x
I need some pointers on compiling a new kernel for SL5x. I haven't done one in a long time and am not familiar with the directory layouts. I used to unpack the tarball into /usr/src/linux-2.x.yy and create a link /usr/src/linux to point to this directory. Do all the compilation and installing from /usr/src/linux. Is the preferred method to unpack the tarball into /usr/src/kernels and simply work in /usr/src/kernels/linux-2.x.yy? I have a vague notion that compiling add-on modules for the running kernel rely on files in /usr/src/kernels/linux-2.x.yy? Is this correct? Btw, I posted a question about SL5.5's latest kernel not detecting my DVD drive which was connected to the MB's IDE port. It turns out the Fedora 12's kernel sees it just fine. Without any BIOS tinkering. Hence my interest in updating the kernel for SL5x. I'd also appreciate any caveats (except for auto kernel updates via yum) you may have of running a later kernel with SL5x. Thanks! Ken
Re: SL 4.8 Live CD questions
Hi Larry, I would also be interested in the CERN Howto on building a Live CD. I have my own script for building my own SL Live CD. The live CD is not as full fledged as Urs'; I use the live image only for embedded purposes, but one can always learn new tricks. Please post the URL when you get a chance. I share Urs' opinion that an up-to-date Linux box is secure enough for online banking. Ken Urs Beyerle wrote: Hi Larry, some comments from my side ... First, good to hear that you like the SL LiveCD. The LiveCD is just build after SL releases a new version. Normally the LiveCD is not updated afterwards. You have to wait for the next SL release to get an updated LiveCD. The LiveCD comes with a write/read file system. All changes are written to RAM. Just run yum update and the software on the LiveCD will be updated. This can take some time depending on your internet connection and age of your LiveCD and maybe fill up your memory or maybe not. Or you can just run yum update firefox to get the latest firefox. Because all changes are stored in memory, after a reboot they are lost again. However, you can save the changes on an USB stick, see http://www.livecd.ethz.ch/save.html This should lead to an uptodate LiveCD. One thing you cannot do: You cannot update the kernel of a LiveCD. Editing the LiveCD iso image would be theoretically possible, but I would not do it, because the data is stored in a special way (compressed, etc.) Hope this helps, Urs PS. I would be interested in the CERN excellent Howto of building an own live CD. PS. I use for internet banking an uptodate Linux installation. In my option this is enough secure. P. Larry Nelson wrote: Hi all, the following article has convinced me to go the Live CD route when doing online banking. I had been using a squeaky-clean and bare-bones Windows XP installation (and the latest Firefox) as a guest OS in VMware, but even that method, I've read, has potential security issues. http://voices.washingtonpost.com/securityfix/2009/10/avoid_windows_malware_bank_on.html To that end, I've downloaded and burned my first LiveCD (SL 4.8) and have been playing around with it - quite slick, I must say! However, it has raised some questions, particularly in regards to what I'm using it for. First (probably a Troy/Connie question), how often do security updates get incorporated into the ISO image? For example, there was just a security fix for xpdf/gpdf and the firefox in the SL4.8 LiveCD certainly does not have the latest bug fixes. And if the ISO image doesn't get updated then what's the best course for maintaining a patched LiveCD? I know that one can build one's own LiveCD, and the CERN site has an excellent Howto, so conceivably I could build one and keep it up to date with the latest bug/security fixes, but I'm also aware that there seems to be software out there that will let one edit ISO images to extract or add files (ISO Master for Linux is one I found, but not tried). Any thoughts on this and/or does anyone have experience editing an ISO image? Thanks! - Larry
Re: SL53 disk I/O is slow
I solved the problem and thought I'd pass it along. The BIOS also has a SATA AHCI setting which was disabled. After I enabled it, the disks are now recognized as /dev/sd? and I'm getting more reasonable disk speeds. I also do not get the ata_piix message about no available legacy ports. Ken Ken Teh wrote: I just installed SL5.3 on a Supermicro PDSBE motherboard and its disk i/o is painfully slow; about 3MB/s. The system has SATA drives but the kernel sees them as /dev/hdx devices. There is also a ata_piix message at the beginning of bootup that says no available legacy port. I'm guessing the failure to recognize the SATA drives as /dev/sdx and the slow disk i/o are related to this cryptic message. I ran a Fedora 11 live CD on the system and it can do disk i/o easily 20-30 times faster which is closer to what I expect. 100MB/s or more. I'm pretty sure the problem is kernel related. I tried switching the SATA mode in the BIOS to compability instead of enhanced. It didn't make any difference. I wasn't expecting any. The compatibility vs native stuff, I thought, was something that was done when SATA support for spotty. The SATA controller is an Intel ICH8 which I figure should be well supported. I also looked at the dumps of hdparm both under SL53 and Fedora 11. The features enabled are the same for both. There are additional features listed under Fedora 11, all SCT (SMART Command Transport) related. There is also a whole bunch of dma modes displayed in both hdparm -t dumps. On SL53 there is a * next to udma5 while it's next to udma6 under Fedora. I'm not sure what this means. Any ideas on how to proceed? Build a custom kernel? Which I am reluctant to do since I rely on SL for updates. Ken
Re: Programs for backing up data?
You need to be clear about what you mean by backup. If you are keeping a mirror of your data, rsync is fine. But sometimes you want a version of a file from 3 months ago. rsync is not going to do that for you. The appropriate terms are redundancy and archival and the question you need to ask yourself is which is it you want. Maybe both? Good luck Avetisyan, Aram wrote: Hello, We have a few desktops running CERN's version of Scientific Linux (SLC 4.8) and would like to automatically backup the data on these via the network all of them are on. Can anyone recommend a good way to do this? I know of rsync and I've found some rather long lists of software online, but I'm not sure which method is best. Thanks. -- Aram
SL53 disk I/O is slow
I just installed SL5.3 on a Supermicro PDSBE motherboard and its disk i/o is painfully slow; about 3MB/s. The system has SATA drives but the kernel sees them as /dev/hdx devices. There is also a ata_piix message at the beginning of bootup that says no available legacy port. I'm guessing the failure to recognize the SATA drives as /dev/sdx and the slow disk i/o are related to this cryptic message. I ran a Fedora 11 live CD on the system and it can do disk i/o easily 20-30 times faster which is closer to what I expect. 100MB/s or more. I'm pretty sure the problem is kernel related. I tried switching the SATA mode in the BIOS to compability instead of enhanced. It didn't make any difference. I wasn't expecting any. The compatibility vs native stuff, I thought, was something that was done when SATA support for spotty. The SATA controller is an Intel ICH8 which I figure should be well supported. I also looked at the dumps of hdparm both under SL53 and Fedora 11. The features enabled are the same for both. There are additional features listed under Fedora 11, all SCT (SMART Command Transport) related. There is also a whole bunch of dma modes displayed in both hdparm -t dumps. On SL53 there is a * next to udma5 while it's next to udma6 under Fedora. I'm not sure what this means. Any ideas on how to proceed? Build a custom kernel? Which I am reluctant to do since I rely on SL for updates. Ken
trouble installing 4.8, etc.
I'm having trouble installing SL4.8 X86_64 from the DVD. It keeps coming up saying it can't find certain packages, speex for one. I've checked the media and the installer says the media is fine. Any ideas? On a somewhat related note, I've not been able to install the later 4.x releases via kickstart. For some reason, 4.x machines whether at install time or boot time take forever to accept their IP assignments via DHCP. This is not a big problem after it's installed; it just takes forever to boot. But, during kickstart, it times out after which you have to do it interactively. The 5x machines do not show this behavior at all. I also didn't have this problem with earlier versions of 4x, say around 4.2-4.4. This is not a pressing issue for me since I've pretty much moved on to 5x. But I'm just wondering if anyone else has encountered this situation. Thanks! Ken
dual head with ATI Radeon 43xx
Has anyone successfully run a single desktop across 2 monitors with an ATI Radeon 4300 series graphics card with SL5.3?
Re: Updates and long-running user processs
I was at a linux conference and met someone who sells a tool to patch linux kernels live without rebooting. http://www.kspice.com Dr Andrew C Aitchison wrote: What do other groups do about updating applications and machines with long running processes ? My users run two sorts of long running processes, with different problems when it comes to updates. First, I have users who never log off. Thus applications like firefox and pdf viewers will be running when they are updated. Some time later these applications may try to load and run plugins which have been removed/updated. Second, I have users with long running calculations (often weeks or more) which would be interrupted if the machine were rebooted into an updated kernel. User-writing code often check-points, so the actual calculation time lost is not significant, but calculations in commercial packages such as Mathematica and Maple are often less good about check-pointing. How do people balance the disruption of killing user processes against the need to update to the latest versions of software ? Thanks,
Re: Updates and long-running user processs
Sorry, typo. http://www.ksplice.com/ Ken Teh wrote: I was at a linux conference and met someone who sells a tool to patch linux kernels live without rebooting. http://www.kspice.com Dr Andrew C Aitchison wrote: What do other groups do about updating applications and machines with long running processes ? My users run two sorts of long running processes, with different problems when it comes to updates. First, I have users who never log off. Thus applications like firefox and pdf viewers will be running when they are updated. Some time later these applications may try to load and run plugins which have been removed/updated. Second, I have users with long running calculations (often weeks or more) which would be interrupted if the machine were rebooted into an updated kernel. User-writing code often check-points, so the actual calculation time lost is not significant, but calculations in commercial packages such as Mathematica and Maple are often less good about check-pointing. How do people balance the disruption of killing user processes against the need to update to the latest versions of software ? Thanks,
Re: Strategies for setting up scientific packages?
I have one called ssetup which I wrote. Originally, it was 'setup' (idea stolen from VMS) but Redhat commandeered the name. My packages are stored, one per directory, on a packages server which is NFS mounted on clients. All the command does is run a .setup file in the appropriate directory which actually does the proper environment setup. Graham Allan wrote: We have been using UPS/UPD from fnal for this (mainly for root and packages like that): http://www.fnal.gov/docs/products/ups/ Being somewhat disconnected from fnal here, I am not really sure how well supported ups/upd is these days, it is hard to get much information on it. I'd be interested if anyone from fnal might comment. I guess the most popular package to do this kind of thing is modules, http://modules.sourceforge.net/ Graham On Wed, May 27, 2009 at 06:09:57PM -0700, Matt Harrington wrote: This isn't specifically a Scientific Linux question, but I suspect many of the list's readers are in the same boat as me. We have about 30 scientific packages, of which about 20 are command-line only and about 10 are GUI applications. Rather than have massive, slow, and unmaintainable .cshrc/.bashrc files, people use an application called prepare to set up each app as necessary. prepare originally came from Johan Postma at EMBL Heidelberg and unfortunately its website seems to have disappeared. It's a clever csh script which detects the architecture in use and then sources an appropriate csh file to set up environment variables and aliases. Originally it worked with IRIX and OSF/1, and when Linux came on the scene I made the necessary modifications. The idea is that prepare ccp4 will set up the CCP4 package for whatever type of computer a user is currently using: SGI, Tru64 Alpha, Linux Alpha, Linux x86, or Linux AMD64. Simply typing prepare gives a list of applications currently configured for the computer in use. This has worked well, but I haven't revisited this issue in 15 years and am wondering how the rest of the scientific world solves this problem. All comments welcomed. Matt UCSF
Intel I7
Does SL5.2 run on the Intel I7 processors? Any magic involved? Thanks!
kernel lockd does not honor requested lockd ports
I've fixed the various nfs ports in my firewall config and have propagated these ports to /etc/sysconfig/nfs. All the ports are honored except for the lockd ports. I've even tried setting the ports in sysctl.conf and appending them to the kernel boot in grub.conf. rpcinfo -p shows that the kernel (2.6.18-128.1.6.el5) basically ignores me. NFS clients are mounting via NFSv3. Ignoring the lockd numbers creates apparently creates problems for some applications, presumably because the application is requesting file locks. For example, firefox won't run when launched in a user's home directory that is mounted remotely from the server. Has anyone seen this problem? What's the fix? Ken
Re: determining the appropriate sata driver
Hi Troy, I'm looking for something more basic. I'm trying to craft an init script for an initrd image to load appropriate drivers. Nothing fancy. Just the basics. Like what sort of hard drives are on the system. SATA or IDE? The SATA issue confuses me since there is a plethora of SATA drivers. Does the init script have to try each one to see if it finds a match or is there something in /proc or /sys that provides some identification? The only tool I have in my initrd image is busybox and it doesnt have an lspci equivalent. The /proc system contains a subdirectory that lists all devices on PCI but as bus.device files. I tried cat'ing the files but they are not ascii. I'm guessing it's possible to walk these files to extract the information I need. I can dig deeper but to save time, I'd ask if someone already knows the answer. Ken Troy Dawson wrote: Mark Stodola wrote: I believe most of anaconda's magic comes from probing the PCI and USB bus for vendor and device IDs. If you dig into a driver (for example, e1000e), you will find a pcitable listing all of the IDs the driver supports. If you look in /lib/modules/kernel/ you will find a modules.pcimap and modules.usbmap among other bus types. Hope that helps. Cheers, Mark Ken Teh wrote: How does a system determine the appropriate sata driver? Specifically, how does anaconda figure out to write alias scsi_hostadapter ata_piix in my modprobe.conf and to bind it into my initrd image? I took apart boot.iso but didnt find anything readable that indicated how this was determined. Thanks! Ken Hi Ken, On SL 5, alot of that data comes from hwdata, which is in the /usr/share/hwdata directory. The main ones people and programs look at is pci.ids and usb.ids But there are other files in there. Troy
help decoding yum message about aufs
I'm playing with aufs and have built and installed aufs for the 5.2 distro from Urs Beyerle's src rpm. N.B. the src rpm builds the aufs and the kernel-module-aufs rpms. I have my yum configured to exclude kernel* and the only enabled repos I use are sl and sl-security. I'm a little confused by the yum cron message below. The aufs src rpm I got from Urs is 20070210.cvs. The message seems to suggest that there is an aufs package in my yum enabled repos. But I have looked at the 5.2 install tree and I don't find any such rpm. Can someone clarify? Customizations are tedious and having a known environment to make customizations helps a great deal. If the aufs and kernel-module-aufs are part of the standard SL yum repos, then I wont bother building them and would prefer to rely on both aufs and the kernel being kept in sync. Thanks! Ken YUM - security Error: Missing Dependency: kernel-module-aufs = 0.20090202.cvs-6.sl5 is needed by package aufs
rebuilding kernel rpms
A simple question: I'd like to rebuild the current kernel rpm with minor mods to the config. The procedure I'm following is what's documented for RHEL 4, that is, install the src.rpm do an rpmbuild -bp cd to /usr/src/redhat/BUILD/kernel-vers/linux-vers/ make menuconfig to create the custom .config. At this point, how do I fold the .config that I generate back in so I can simply say 'rpmbuild -ba' and have it generate the kernel rpms with my mods? Thanks!
Re: rebuilding kernel rpms
Excellent! Thank you very much Akemi. I was browsing through the spec file and encountered all those sections that you wrote up in the wiki. I did not wish to dig through all those details to figure out how to spec file works. Your recipe is exactly what I need! Thanks! A short followup: You have a warning at the beginning of the wiki that it is almost not necessary to build a custom kernel. My problem is that I need a kernel that's able to mount a squashfs fs on a loop device during boot-up. The standard kernel has both squashfs and loop as .ko. I believe I need them built-in. If I'm wrong, please advise. Thanks again! Ken Akemi Yagi wrote: On Thu, Feb 12, 2009 at 2:27 PM, Ken Teh t...@anl.gov wrote: A simple question: I'd like to rebuild the current kernel rpm with minor mods to the config. The procedure I'm following is what's documented for RHEL 4, that is, install the src.rpm do an rpmbuild -bp cd to /usr/src/redhat/BUILD/kernel-vers/linux-vers/ make menuconfig to create the custom .config. At this point, how do I fold the .config that I generate back in so I can simply say 'rpmbuild -ba' and have it generate the kernel rpms with my mods? There is a well-written CentOS wiki on how to build custom kernels. The instructions are based on CentOS-5 but there are notes whenever the procedures differ for CentOS-4. One important thing is that you do not want to do the building as root. This is a dangerous practice even if you are an experienced user (mistakes can happen). Akemi
yum --installroot anomaly?
It's not a show stopper but I don't think it's supposed to happen this way. I'm working in /root/livecd and installing the core package group into a subdirectory called 'live' with the either commands. Both produce the same outcome. yum --installroot=/root/livecd/live install ... yum -c live/etc/yum.conf --installroot=/root/livecd/live install ... The directory is first prep'ed following Urs Beyerle's instructions on making live cds. The install works fine but it creates /root/livecd/live/root/livecd/live/var/lib/yum and writes a file called transaction-all. there. Note the directory path which repeats the current working directory. I would have thought the installroot option would reorient yum so whatever it does is relative to the install root, ie, the transaction file is written to /root/livecd/live/var/lib/yum instead of remembering where it was working from.
Re: comps xml advice
Troy Dawson wrote: Mark Stodola wrote: Ken Teh wrote: I would like to install a minimal SL system plus/minus certain packages. Since it's minimal, I figure the easiest way is to install core (maybe base) and manually add and subtract packages in the ks file. Another possibility is to modify the comps.xml file to turn type='default' to type='optional' for those packages I don't want. Of course, that means I have to host the installation tree on a server inside of fetching them via http from a mirror. Both methods are about equivalent. What I don't want to do is sit in front the machine and manually check and uncheck packages. The above approaches work only because it's a minimal install. Anything more would get pretty tedious. Are there cleverer ways of doing this? Manually checking/unchecking is probably about the fastest way of doing it, certainly much faster than hand typing each package name or editing a bunch of 8+ character strings in an xml file. I'd either do the clicking at install time or use the system-config-kickstart tool on an existing setup to generate a kickstart before you get going. If you're not the mouse type, once you get in to the package lists in the installer, you can use the up/down keys and spacebar to toggle packages extremely fast. What I've done is sorta a combination of what you both are saying. I did an install with only core, and then added the packages I wanted using yum. I then took the kickstart file made during install (/root/anaconda-ks.cfg) and added the packages that are listed in the yum log (/var/log/yum.log) I've edited the comps.xml file, and I've edited kickstart files, and I believe editing the kickstart file is easier. Troy ok, thanks! I thought perhaps there were other ways I was not aware of. I'm a kickstart file person myself. Manual is maybe good for a onesie but gets old and unreliable for N 2 systems. Also, I like kickstarts because it's both a record of what you did. And it can be customized per machine.
Re: sci-linux as a pseudo-embedded os, have any of you installed it in this fassion?
I recommend using a live CD distro. This may or may not apply to you but typically an embedded system is something you want to be able to switch off and on with a button. If your embedded SL mounts a hard disk partition as /, especially if it is mounted read-write, you really should use 'shutdown' to shut the system off. Otherwise you run the risk of disk integrity problems. It's happened to me before and when it happens you will need to build the system again. A live CD distro runs entirely out of RAM so it's CD image remains pristine through reboots. Salvador Aguinaga wrote: Hello! I've wondered if any of you have used scientific linux on a x86 hardware to run a single application ( like an embedded OS ) and disable updates and remove unneeded packages? Or if you there is a better alternative to accomplish the same thing. // Sal Aguinaga // Northwester University // Evanston, IL
when is an sda an hda?
I'm getting confused with the sda/hda naming conventions. I thought all SATA disks were sd devices. They were a while back but apparently, not anymore. And, I can't seem to make any sense of when an sda is an hda. I'm currently installing a system with a SATA system disk that has a IDE CDROM. A systemrescuecd (Gentoo based kernel) identifies the disk as an sda. But the 5.2 installer says it's an hda. There's a single IDE connector on the MB on which hangs a CDROM drive. Apparently, it's not an hda. What is it? An sda? What gives?
Re: NFS woes
Dr Andrew C Aitchison wrote: On Wed, 8 Oct 2008, Ken Teh wrote: reverted back to NFSv3 and have had no problems with it. Are folks satisfied with NFSv4? Or, am I the stupid one for trusting the default on the distro? I'd appreciate any general remarks you might have on how to move forward. I've been put off NFSv4 becuase I cannot understand how to export several directory trees in differenet partitions. NFSv4 seems to want you to merge them into a virtual partion and export that, but the kernel cannot cope with the different bits of the virtual partition having different permission modes. So I can't export a read only partition and a read-write partition without the kernel becoming dangerously confused :-( That I was able to do. The following /etc/exports is an example: /x @myclients(fsid=0,rw,) /x/packages @myclients(ro,...) /x/home @myclients(rw,...) I went back to NFSv3. I completed my transition to NFS3 today. We'll see if it performs better. I expect it will from past experiences.
Re: NFS woes
It's worse today. I have a lot of stale stateid messages both in dmesg and in /var/log/messages. Here's a sample Oct 8 00:00:41 prelude kernel: NFSD: stale stateid (48bebbc4/0004ddd3/0004ef27/ 0006)! Oct 8 00:01:12 prelude last message repeated 44507 times Oct 8 00:02:13 prelude last message repeated 90519 times Oct 8 00:02:48 prelude last message repeated 52779 times It's going bonkers. Since midnight, it's put out 90 thousand such messages every minute. I think I made a bad choice using NFSv4. It's been problematic ever since I stood up the server and it's only gotten worse. The last server I stood up I reverted back to NFSv3 and have had no problems with it. Are folks satisfied with NFSv4? Or, am I the stupid one for trusting the default on the distro? I'd appreciate any general remarks you might have on how to move forward. Thanks! Donald Tripp wrote: Anything in /var/log/messages that would indicate anything? On Oct 7, 2008, at 11:03 AM, Ken Teh wrote: One of my NFS servers is very noisy, spewing out NFSD: preprocess_seqid__op: magic stateid! Does anyone know what this means? NFSv4 seems to be problematic and difficult to diagnose. These are dmesg messages with no timestamps which makes it difficult to correlate it with anything else. I'm going to throw out a couple of possibilities. (1) It started after adding a SL5.2 client, (2), we had our networks reconfigured to go behind a divisional firewall, we were previously behind a lab-wide perimeter firewall. Without better system diagnostics, I'm grasping at straws. Any remarks or advice would be appreciated. Thanks! Ken
wine emulator
Are they any problems running the wine emulator on SL5x? I recall wine used to be part of RedHat's distributions. But, I dont see it in SL5x. Is or was there a reason why it was left out? It appears to be part of the EPEL repo which I am not familiar with. One of my users wants it on a server. I'm reluctant to install it if it will hang up the server. I've never used it myself. I've played around with VMWare and while it works, there were the occasional hiccups. Which is ok if it's your own machine. Not so good on a multi-user server. I'd appreciate any advice. Thanks!
install via http for SL4.7 does not work
I'm unable to install SL4.7 via the url method. Originally, I thought it was something to do with our local mirror but I've tried it against http://ftp.scientificlinux.org and I get exactly the same error. The Alt-F4 (?) screen shows the following (I'm copying it from a Ctrl-S screen, so typos galore) : URL_STAGE_MAIN- url is http://ftp.scientificlinux.org/linux/scientific/47/i386 url address ftp.scientificlinux.org url prefix /linux/scientific/47/i386 mntloop loop0 on /mnt/runtime as /mnt/source/SL/base/stage2.img fd is 22 mntloop loop7 on /tmp/update-disk as /mnt/source/SL/base/updates.img fd is 22 umounting loopback /tmp/update-disk loop7 transferring http://ftp.scientificlinux.org//linux/scientific/47/i386/SL/base/product.img to a fd transferring http://ftp.scientificlinux.org//linux/scientific/47/i386/disc1/SL/base/product.img to a fd umounting loopback /mnt/runtime loop0 umount loopback /tmp/product-disk loop7 LOOP_CLR_FD failed for /tmp/product-disk loop7 (No such device or address) and it just keeps repeating itself over and over again. I noticed that it first mounted loop7 on update-disk, then tried to umount it as product-disk. Is this a typo in anaconda? I've posted this before maybe a year ago with SL4.6 which IIRC also does not work. I believed it worked with SL4.5. And, I know for a fact it works with SL5.x
tripplite support in gnome power manager?
I was startled to see the power manager icon on a 5.2 system that I had just installed and plugged into a tripplite ups. I did connect the computer to the ups via the usb cable provided. The system seems to know all about the UPS system; it's make and model and has some information on the battery charge. I need some advice on how to proceed here. I was going to use the manufacturer's software but first, I'd like to find out more about how the gnome power manager works vis-a-vis the ups, whether it's preferable to use the tripplite software, etc. I would appreciate any help or guidance you might have. Thanks! Ken
yum groupinstall
I suppose I could test this if I had a sacrificial machine, but I'm hoping someone can simply confirm that it does work. I have a 4.2 server machine which was not installed with the gnome desktop. I need to do so now and was wondering if 'yum groupinstall gnome' would do the trick. Thanks!
Re: yum groupinstall
Thanks, that's what I wanted to hear. Whether 'groupinstall gnome' was sufficient or whether I had to manually handle the dependencies as well. And, tips for a more complete solution that would satisfy my expectations. Troy Dawson wrote: Ken Teh wrote: I suppose I could test this if I had a sacrificial machine, but I'm hoping someone can simply confirm that it does work. I have a 4.2 server machine which was not installed with the gnome desktop. I need to do so now and was wondering if 'yum groupinstall gnome' would do the trick. Thanks! Hi, This all depends on if you already have all the X stuff installed and just need the gnome desktop, or if you installed something very light, like a server. If you already have X installed and working, like if you installed KDE and you now want gnome installed, then you can almost do what you said, but not quite. First - get the list of groups yum grouplist Then - do a groupinstall with the group in quotes, so what you would want to do is yum groupinstall GNOME Desktop Environment and that will pull in all the gnome stuff. But, the odds are that you are asking because you have a server installed, and want a full graphical desktop, including the graphical login screen. If you do just the gnome desktop, it is going to get you just that, the gnome desktop and enough X stuff to satisfy dependancies. But it won't get you the gdm, or any of the X configuration stuff. So you probrubly want yum groupinstall X Window System yum groupinstall GNOME Desktop Environment as well as yum groupinstall Graphical Internet yum groupinstall Printing Support yum groupinstall System Tools yum groupinstall Text-based Internet Troy
debugging a http based install
I just started using the http based kickstart install. Done via the kickstart url option. I have installed a couple of SL5.1 systems successfully and tried today to install a SL4.6 machine. I boot the system with the disc1 iso image and provide the kickstart file via NFS. Like so: boot: linux ks=nfs:installserver.my.net:/SL/46/ks/myks In my kickstart file I specify our local SL mirror as the install server. Like so: url --url http://mirror.anl.gov/pub/scientific-linux/46/i386 Works for 5.1 but not for 4.6. I've attached the Alt-F2 screen. Can someone provide some pointer on how to debug this? How do I go about isolating the problem? I can revert back to using my old method which is an NFS based install. But http based installs are so convenient especially since we have high-speed access to a local SL mirror. Thanks! Here's the fragment of the Alt-F2 dump. It's hand copied so they are likely typos. I'm also not sure where it begins. But it appears to be looping. * trying to mount CD device hdc * mntloop loop0 on /mnt/runtime as /mnt/source/SL/base/stage2.img fd is 21 * mntloop loop7 on /tmp/update-disk as /mnt/source/SL/base/updates.img fd is 21 * unmounting loopback /tmp/update-disk loop7 * transferring http://mirror.anl.gov//pub/scientific-linux/46/i386/SL/base/product.img to a fd * transferring http://mirror.anl.gov//pub/scientific-linux/46/disc1/SL/base/product.img to a fd * unmounting loopback /mnt/runtime loop0 * unmount loopback /tmp/product-disk loop7 * LOOP_CLR_FD failed for /tmp/product-disk loop7 (No such device or address) * URL_STAGE_MAIN - url is http://mirror.anl.gov/pub/scientific-linux/46/i386 * url address is mirror.anl.gov * url prefix is /pub/scientific-linux/46/i386
Re: %post NFS access
I've used mount's in the kickstart %post section like forever and I never needed 'nolock' until 5.x. All I did was 'mount -rvt nfs server:/remote /local'. I also did not start the nfslock service. Jon Peatfield wrote: On Fri, 20 Jun 2008, Steve Traylen wrote: snip It has been ages since I 've done this but certainly in the past the mount needed both a nolock and nfsvers=2 mount -o nolock,nfsvers=2 I'm pretty sure that we don't need to add nfsvers any more - I know we don't use it. I'm not sure if the nolock is really needed any more but we still do that - though we also start up the nfslock stuff too as well :-) We now do nfs mounts in %pre as well since we found that nfs access to our extra local repos (via file:// locations) is faster (and uses less memory) than either ftp: or http: forms of access from anaconda.
Re: SL 5.1 file system corrupted easily!!!
After installing a RAID 1 system, you should not shutdown the system until the system has a chance to complete the sync of both disks. Maybe this is the reason for your problem. I'm a little confused why you want to deactivate the second channel and reactivate it once a week to resync. I have systems that have run with both disks for 4 years or more without any problems. I think this is the way it is supposed to work. The way you are doing it you are better off not configuring the system as RAID and simply doing a manual sync once a week. Timmy wrote: Hi Roelof and others: I uninstalled the SiI software raid card and installed SL 5.1 again! SL 5.1 seemed to be able to set up a RAID pair by itself even without the add-on raid card! Everything was very smooth. My config is: 2 x 160GB SATA harddisks installed on 2 separate SATA channels. My idea is to keep one hardisk as working system disk 24 hours (science calculation, business, BT, everything etc and therefore it will be worn out to death easily ;) and the other harddisk as Spare disk. It will be rebuilt and cloned per week against the working system disk. Another file backup software will take care the incremental backup everyday! Sound very perfect! ;) Ok! After the above mentioned SL 5.1 system was built and became stable, I shut down the computer. Then, I turned on the computer and deactivated the 2nd SATA harddisk (named sdb) from within BIOS. SL 5.1 booted with the single hardisk(sda) successfully (It claimed Degraded RAID mirror when SL 5.1 started up). I ran the system for a few days. Then, I activated sdb again and hoped that SL 5.1 could rebuild it and cloned sda to it. However, SL 5.1 claimed that it could not find many files and stopped at the command prompt. I rebooted the computer, deactivated sdb again in the BIOS and booted to SL 5.1 successfully with only sda again! So, I face a problem: How to REBUILD / CLONE sda to sdb with Linux command mdadm?? :(( Timmy P.S. After I had unplugged the SiI software SATA RAID card, SL 5.1 never stops at RedHat Nash 5.19.6 . any more. I guess this model of card is not compatible with the current SL 5.1! Avoid it! :(( I never understood why redhat included the alpha quality dm-raid drivers. I call this alpha, since basic functionality like rebuilding a raidset (!) still is missing. DM-raid is a kludge to be able to do dual-boot with a windows system using these raid drivers. Never, ever use dm-raid on a server. Use mdadm instead. It performs just as well (dm-raid is also softwaid after all) and it is stable. If you only need mirroring, you can also use LVM to create a mirror. (I never used that myself, YMMV) Another advantage of mdadm is that it is hardware independent. Your raidsets will keep on working even when transferred to a different brand of sata controller. By the way: if you try to install an mdadm based raid system on a drive that previously has been part of an intel/promise/highpoint raid set, it will probably fail after the first reboot. This is because the dm-raid system is not active during the anaconda run, but is activated before mdadm in the regular startup. DM-raid will detect the old raid signature and lock the disk, preventing mdadm from using is. Apart from a complete disk wipe, you can set the 'nodmraid' kernel parameter in grub to prevent this from happening. Destroying the old raid-set using the bios tool before installing mdadm often will not work in my experience. Roelof Timmy wrote: I have installed SL 5.1 many times! The file system can be corrupted easily! I don't know whether it is RedHat Linux's weakness. Here are my experiences: Case 1. The system might stop here at boot time: RedHat Nash 5.1.19... After rebooting, it might boot normally or the system went corrupted and no more boot again! Case 2. I installed 2 SATA harddisks as mirror pair on a SiI SATA software RAID card. Installation is successful but after several days' operation, the system could not read the RAID pair. Formatting was needed. Case 3. After several days' downloading of a movie file through Ktorrent, the file system went corrupted again! Formatting harddisk required. I heard that downloading BT file hurts hard disk very much. However, Case 1 and Case 2 do not have any relationship with BT task. Any technique or trick to avoid this?
Re: SL 5.1 file system corrupted easily!!!
You might find this little writeup useful. It's a little dated but I can vouch that it works. http://www.dell.com/downloads/global/power/1q04-hul.pdf Samuel Halicke wrote: On Jun 19, 2008, at 11:50 AM, Timmy wrote: [snip] So, I face a problem: How to REBUILD / CLONE sda to sdb with Linux command mdadm?? :(( Timmy The simplest route is to just use dd(1): dd if=/dev/sda of=/dev/sdb bs=1M conv=notrunc,noerror,sync 21 | tee /tmp/dd-errors.log This will create an exact, bootable clone, including partition tables and UUID. It may take a long time, and there is no progress indicator with dd, but it works. If there are errors, you can check the file tee writes. See the dd manpage for more details. Other alternatives include partimage, available here: http://www.partimage.org/Main_Page -- and also, the mdadm command itself -- the short version: set up RAID 1, partitions to Linux raid autodetect, put the first drive (your everyday drive) as the first in the set and the second as the mirror. The md driver should automatically rebuild the mirror for you. Just make sure you partition the disks the same way. There's plenty of information on the web about working with md/mdadm. -- Samuel Halicke[EMAIL PROTECTED] FNAL/CD/SF/FEFFCC/252/U (T): 630.840.4308(P): 630.266.0042
Re: Missing dependencies.
Hah! I was just about to post the same thing. Have you got a rope around you? I'm sure all of us will be more than willing to pull you up when you've fixed it. *wink* Troy Dawson wrote: *Troy starts going down the steep step of dependancies and sighs* I had checked pretty good on SL4, and then forgot that this is a major evolution update for SL5, so I didn't check as much. This is going to take a little bit for me to figure out ... and I should have done it yesterday, I apologize. Expect the dependancies to be fixed in a little bit. I'll let you know when they are pushed out. Troy Al Russell wrote: I tried to do yum update and got this response. [EMAIL PROTECTED] ~]# yum update Loading kernel-module plugin Setting up Update Process Setting up repositories Reading repository metadata in from local files Excluding Packages in global exclude list Finished Resolving Dependencies -- Populating transaction set with selected packages. Please wait. --- Package evolution.i386 0:2.12.3-8.el5_2.2 set to be updated -- Running transaction check -- Processing Dependency: libegroupwise-1.2.so.13 for package: evolution -- Processing Dependency: libnssutil3.so for package: evolution -- Processing Dependency: libexchange-storage-1.2.so.3 for package: evolution -- Processing Dependency: libgtkhtml-3.14.so.19 for package: evolution -- Finished Dependency Resolution Beginning Kernel Module Plugin Finished Kernel Module Plugin Error: Missing Dependency: libegroupwise-1.2.so.13 is needed by package evolution Error: Missing Dependency: libnssutil3.so is needed by package evolution Error: Missing Dependency: libexchange-storage-1.2.so.3 is needed by package evolution Error: Missing Dependency: libgtkhtml-3.14.so.19 is needed by package evolution [EMAIL PROTECTED] ~]# I am using the vanilla yum configuration obtained from clean SL 5.1 install. What am I missing? Thanks, Al
Re: Missing dependencies.
Follow-up question: I discovered this missing dependencies as part of an network install in the %post section. My procedure is to NFS mount a remote directory where I keep several post install scripts that configure the machine exactly the way I want it. The NFS mount does not seem work anymore. I have the following lines in my %post section %post ... [ -d /mnt/SL ] || mkdir -p /mnt/SL mount -rvt nfs server.anl.gov:/SL /mnt/SL if [ $? = 0 ]; then echo server.anl.gov:/SL mounted else echo failed to mount server.anl.gov:/SL exit 0 fi cd / /mnt/SL/51/post/postcommon ... I can mount the directory fine after the installation reboot and can run the scripts manually. Any ideas what the difference might be between a post-configure after a reboot and a post-configure during the anaconda %post section? Obviously, the kernel is different. But, I load my kickstart script via NFS from the same server where I want to fetch the post-configure scripts in the %post section. Ken Teh wrote: Hah! I was just about to post the same thing. Have you got a rope around you? I'm sure all of us will be more than willing to pull you up when you've fixed it. *wink* Troy Dawson wrote: *Troy starts going down the steep step of dependancies and sighs* I had checked pretty good on SL4, and then forgot that this is a major evolution update for SL5, so I didn't check as much. This is going to take a little bit for me to figure out ... and I should have done it yesterday, I apologize. Expect the dependancies to be fixed in a little bit. I'll let you know when they are pushed out. Troy Al Russell wrote: I tried to do yum update and got this response. [EMAIL PROTECTED] ~]# yum update Loading kernel-module plugin Setting up Update Process Setting up repositories Reading repository metadata in from local files Excluding Packages in global exclude list Finished Resolving Dependencies -- Populating transaction set with selected packages. Please wait. --- Package evolution.i386 0:2.12.3-8.el5_2.2 set to be updated -- Running transaction check -- Processing Dependency: libegroupwise-1.2.so.13 for package: evolution -- Processing Dependency: libnssutil3.so for package: evolution -- Processing Dependency: libexchange-storage-1.2.so.3 for package: evolution -- Processing Dependency: libgtkhtml-3.14.so.19 for package: evolution -- Finished Dependency Resolution Beginning Kernel Module Plugin Finished Kernel Module Plugin Error: Missing Dependency: libegroupwise-1.2.so.13 is needed by package evolution Error: Missing Dependency: libnssutil3.so is needed by package evolution Error: Missing Dependency: libexchange-storage-1.2.so.3 is needed by package evolution Error: Missing Dependency: libgtkhtml-3.14.so.19 is needed by package evolution [EMAIL PROTECTED] ~]# I am using the vanilla yum configuration obtained from clean SL 5.1 install. What am I missing? Thanks, Al
SL 5.1 unable to mount kickstart file via NFS
I'm trying to kickstart via nfs. At the syslinux boot prompt, I type linux ks=nfs:slserver.privnet:/SL/51/ks/standalone The slserver logs an authenticated mount request, so it's accepted the request and offered the share. But, the anaconda installer fails to register the offer and suggest I continue with an interactive install. On screen 3, I see the following logged: INFO : doing kickstart... setting it up INFO : url is slserver.privnet:/SL/51/ks/standalone INFO : file location: nfs://slserver.privnet:/SL/51/ks/standalone ERROR : failed to open /tmp/mnt/standalone: Permission denied ERROR : failed to copy file to /tmp/ks.cfg I've done NFS installs since RedHat 6. Most times when it failed it was usually an NFS problem. And I suspect this is the case here. Perhaps some NFS4 weirdness? I'd appreciate suggestions on how to proceed.
Re: SL 5.1 unable to mount kickstart file via NFS
More on this problem: I tried to start the system via the rescue mode to see if I could manually issue the NFS mount command. Unfortunately, the rescue mode does not work. So far the only thing that works is manually feeding the CDs. Which is what I'm trying to get away from. :\ Ken Teh wrote: I'm trying to kickstart via nfs. At the syslinux boot prompt, I type linux ks=nfs:slserver.privnet:/SL/51/ks/standalone The slserver logs an authenticated mount request, so it's accepted the request and offered the share. But, the anaconda installer fails to register the offer and suggest I continue with an interactive install. On screen 3, I see the following logged: INFO : doing kickstart... setting it up INFO : url is slserver.privnet:/SL/51/ks/standalone INFO : file location: nfs://slserver.privnet:/SL/51/ks/standalone ERROR : failed to open /tmp/mnt/standalone: Permission denied ERROR : failed to copy file to /tmp/ks.cfg I've done NFS installs since RedHat 6. Most times when it failed it was usually an NFS problem. And I suspect this is the case here. Perhaps some NFS4 weirdness? I'd appreciate suggestions on how to proceed.
Re: SL 5.1 unable to mount kickstart file via NFS
Thanks for your response. The fact that it works for you is encouraging. If I do my nfs install, ie, linux ks=nfs, I don't get the console-2 bash prompt. I don't think anaconda is far enough along. That's why I tried to load linux rescue. I should have written early that anaconda actually crashes in the rescue mode. I get some sort of exception and I see on console-1 an abnormal exit message following a dump. I tried a plain old linux text but I can't nfs mount because the network is not prepped yet. Any more ideas? Jon Peatfield wrote: On Tue, 3 Jun 2008, Ken Teh wrote: More on this problem: I tried to start the system via the rescue mode to see if I could manually issue the NFS mount command. Unfortunately, the rescue mode does not work. So far the only thing that works is manually feeding the CDs. Which is what I'm trying to get away from. :\ Just to confirm we do sl51 installs with nfs kickstart files all the time and it works fine for us - in our case we are PXE booting with pxelinux but that shouldn't make much difference anyway. If you boot from the CDs and get to the console-2 bash prompt can you NFS mount the directory from that server by hand? Can you run tcpdump on the nfs server and see if that sheds any light on what the problem might be? -- Jon
Re: SL 5.1 unable to mount kickstart file via NFS
Jon Peatfield wrote: On Tue, 3 Jun 2008, Ken Teh wrote: Thanks for your response. The fact that it works for you is encouraging. If I do my nfs install, ie, linux ks=nfs, I don't get the console-2 bash prompt. I don't think anaconda is far enough along. Yup the bash on console-2 isn't started until stage2 has been loaded. That's why I tried to load linux rescue. I should have written early that anaconda actually crashes in the rescue mode. I get some sort of exception and I see on console-1 an abnormal exit message following a dump. Is that booting from CD? Are there other related messages on the VCs? I've never seen it crash like that, is the hardware known to be working? I tried a plain old linux text but I can't nfs mount because the network is not prepped yet. Any more ideas? Well you could go far enough through the manual install (booted from CD) to set up networking or manually bring up networking in the console-2 bash... (you don't have much to work with in that environment but I think it should include at least one of ifconfig or ip). Did running tcpdump (or wireshark or whatever) on the server show anything? -- Jon No, I haven't tried tcpdump yet. syslog shows that the server heard the nfs mount request. I've also verified that I can mount the volume from another SL5.0 machine.
Re: SL 5.1 unable to mount kickstart file via NFS
OK, I found the problem. Should have paid attention to the console-3 error.log. I had made a manual installation and had taken the anaconda-ks.cfg that was left in /root as the basis for creating custom kickstart files. That file was mode 600. That was it. Duh!! Anyway, the rescue mode does crash. The hardware is a Dell Precision and I was able to install it manually. But I cannot bring the machine up with the rescue CD. I can try to jot down what the screen says but I see only a fraction of the dump when the rescue CD exits abnormally. If Connie or Troy think this is useful, I will try to do it. Thanks for everyone's help and patience. Ken
Re: gpdf problems
November 2007 was the reported install date. John Summerfield wrote: Ken Teh wrote: I have a 4x system that where gpdf has just developed a problem rendering a pdf file where it did not before. The same said file is rendered without problems with ghostscript. We reloaded the file from its original URL and they are binary identical. Anyone seen a problem like this before? Ias gpdf been updated recently? rpm -qi gpdf
installing on asus p5b-vm se
I tried to install 5.1 x86_64 on a p5b-vm se. Apparently, the installation kernel does not see the SATA disk. I booted the system with Ubuntu's latest offering, 8.04, and looking at dmesg I see the disk listed. Its kernel is 2.6.24. I'm guessing the 5.1 install disk's kernel is an earlier version. What is its version and can someone suggest how I might solve this problem? Thanks! Ken
gpdf problems
I have a 4x system that where gpdf has just developed a problem rendering a pdf file where it did not before. The same said file is rendered without problems with ghostscript. We reloaded the file from its original URL and they are binary identical. Anyone seen a problem like this before?