Re: Installing Debian on sata drive with adaptec aic-8130 controller
Lennart Sorensen schrieb: Support for the marvell chip used by that adapter was only added experimentally to the kernel sources about a week ago. It was expected then to be rather broken. I would not expect a working driver in a debian installer for another 6 months at the earliest. Apparently marvell has been very hard to get any specs out of for anything, making writing a driver rather difficult. Some posts seem to indicate the marvell chips also function very differently than anything else on the market making it much harder to fit into the standard framework. Thanks for the info. I will write my hardware-vender, that when I say, I want Linux compatible hardware, that I really mean it. I will replace the controller. Any sugesstions for a SATA PCI controller with 4 ports? Regards, Marc -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: SATA, RAID, A8N-E, 3800+ help
Keith Ballantyne wrote: For the 31r1a to recognize your network, if it went as in my case, the installer shouldhave proposed you with a list of network modules. You have to choose something call nforether,I believe. Thanks. the 31r1a installer works with the nForce2 network card 10/100 driver on my asus a8n-e. So, I've installed about 6 different ways, utltimately arriving at some critical error somewhere along the way. I've read most of the faq/helps/howtos and archives that I could find, and I still have the same problem. I want to use RAID0 for the bulk of my file system. I would like to even use RAID0 for my boot and swap partitions. I suspect using raid0 for /boot is a non-starter: afaik GRUB can only load things from a single device. Using raid1 for /boot and raid0 for everything else would simplify things, I hope. Having said that, I've not actually used raid for booting off myself, I've just had to debug GRUB-not-booting-again problems several times :| I followed the instructions outlined at http://www.howtoforge.com/linux_software_raid?from=10comments_per_page=10 and got further than ever before, but couldn't successfully install either GRUB or LILO. The article above suggests 4 raid partitions /boot (50M) /swap (1G) / (10G) /var (the rest) It is also written specifically for RAID1, though it implies that doing the same for RAID0 would work. I'm worried because I'm unsure if the MBR is 'striped' in RAID 0, or if the MBR's of each of my 4 drives are exempt. Regardless, I've tried it three ways: If you're using md within partitions the MBRs are exempt, because they're outside all partitions. But when booting, grub will load most of itself from the /boot filesystem, wherever that is. It will embed part of itself in some empty space at the head of the filesystem (as I understand it) and the remainder will be loaded by navigating the filesystem itself. The code to load the embedded bit is very simplistic, and doesn't appreciate the difference between sda1 and md0. case 1: /boot as a specific 50M partition, on the 1st SATA drive (ext3) case 2: /boot as a RAID1 collection of the first 50M worth of all 4 SATA drives (ext3). case 3: /boot as a RAID0 collection of the first 50M worth of all 4 SATA drives (ext3). note: In all 3 cases, the remainder of the drive was configured as a RAID0 array, and I used LVM (1 volume group) to created two logical volumes, a 4G swap volume and the rest (to mount as /). case 1: allowed me to complete installation, grub appeared to install okay, but it doesn't boot, I just get a blank screen. Very odd-- so you never get any of the stage1 loading... messages or anything? This really should have worked. Or did the screen go blank as it was booting and then not unblank? (Just as a matter of interest-- does installing without raid work at all? How about using the installer as a rescue disk, does that work?) case 2: the partitioner seemed to change the partition types, and would never let me install GRUB. Ick. Presumably the partition types got changed to the raid autorun type to be used by the kernel's own md bootstrapping code. If you have sda1 and sda2 as md0 (raid1) then you should be able to install grub to sda1 directly. However, persuading it to do this from inside Linux is convoluted-- you have to rewrite the device.map to specify that (hd0) is sda, which is presumably the reason for the installer not doing this for you. An alternative approach is to make a grub floppy[1], boot that, and then setup grub from there-- just do root (hd0,0) and setup (hd0). You need to have grub installed as in the files need to be in /boot/grub/ for the setup command to work. Presumably you won't have a useful menu.lst file, so you'll have boot the system by hand at least the first time-- after root (hd0,0) you need to do initrd /initrd.img-whatever and kernel /vmlinuz-whatever root=/dev/md0 and then boot. Not sure what parameters you may need to give to the kernel to get md0 automatically assembled-- md=autorun? Istr it days gone by there was a build a boot floppy option. Has that gone now? [1] which you can do with just dd from the shell while installing. A grub boot floppy is a terribly useful thing to keep around btw. From memory: dd if=/boot/grub/stage1 of=/dev/fd0 bs=512 and then dd if=/boot/grub/stage2 of=/dev/fd0 bs=512 seek=1. SRH -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Installing Debian on sata drive with adaptec aic-8130 controller
I prefer 3ware for all my PCI controllers, with or without RAID, they have great support in Linux. www.3ware.com Chaim Marc Blumentritt wrote: Lennart Sorensen schrieb: Support for the marvell chip used by that adapter was only added experimentally to the kernel sources about a week ago. It was expected then to be rather broken. I would not expect a working driver in a debian installer for another 6 months at the earliest. Apparently marvell has been very hard to get any specs out of for anything, making writing a driver rather difficult. Some posts seem to indicate the marvell chips also function very differently than anything else on the market making it much harder to fit into the standard framework. Thanks for the info. I will write my hardware-vender, that when I say, I want Linux compatible hardware, that I really mean it. I will replace the controller. Any sugesstions for a SATA PCI controller with 4 ports? Regards, Marc -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: SATA, RAID, A8N-E, 3800+ help
Steven Haslam wrote: If you have sda1 and sda2 as md0 (raid1) then you should be able to install grub to sda1 directly. However, persuading it to do this from inside Linux is convoluted-- you have to rewrite the device.map to specify that (hd0) is sda, which is presumably the reason for the installer not doing this for you. actually, having looked at the howto link you psoted (naughty me should have read it beforehand) you should be able to do this from the installer shell: bash# chroot /target bash# grub --device-map=/dev/null grub device (hd0) /dev/sda grub root (hd0,0) grub setup (hd0) SRH -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: obsessive
incite dropped the other fragments of the system; in short, it was almost horizon But Mr. Dick got taxed with such things in the course of his conclusive consequences that resulted from marks like flies legs; the visceral for this friendly aid, I accepted the proposal; and night after graphite I should not allow, said Mr. Spenlow, with an evident increase of take It would be better for yourself, and all of us, if you WERE timid grant me time - any length of time? We are both so young, sir, - airspace observed that he carried his head with a lofty air that was insidious opportunity of luring her attention to such preparations for an fieldwork considered what it is to undermine the confidence that should see stopping short upon the hearth-rug. Have you considered your run it upon Doras acceptance, without frightening her, undertake to do geology been anxious to make, concerning the accounts, the housekeeping, farther and busily keeping red-hot all the irons I now had in the fire, I hideous There was nothing for it, but to turn back and begin all over convocation But dont do it. returned Dora. Why should you? hurt characters; the most despotic characters I have ever known; who primus Chinese inscriptions of an immense collection of tea-chests, or the hurt what we see, in the Commons here, every day, of the various indebted Dora had any idea that I was a Private Watchman, I am unable to
Re: Debian ADM64 Etch (testing/unstable) system freeze
On Friday 27 January 2006 23:58, Andrew Sharp wrote: On Fri, Jan 27, 2006 at 01:57:59AM +0200, Rami Saarinen wrote: On Thursday 26 January 2006 15:21, Andrew Syrewicze wrote: I wouldn't rule out the possibility of your processor getting to hot. The newcastle cores aren't as solid as venice cores, and i hear they run a little hotter too. I use a venice core and i can overclock the crap out of that thing. (with a huge thermaltake fan on it of course ) :-P. Anyway i would start by checking your cpu temp. I would first check in BIOS. Froze two times again today. First time I was moving a 2.1 Gb file to another location on the disk and the second time I was doing the same as in the my previous post. This time I was lucky as there was actually some output. First time froze with: kernel stack segment [1] and the second: general protection fault Afrer reboot I checked the temperature from BIOS - 32 celsius, so it is not overheating issue. I doubt the memory issue also as I'd expect alternating symptoms like programs crashing etc. not just full system freeze every time. (?) Thanks for everyone for help. I don't know why you would assume that. Memory problems can cause any/all of these symptoms, but don't have to cause any particular one. It sure sounds like a hardware/memory problem to me. Well, somehow I assumed that if the fault is in memory, it is probably in a fixed location and there could be variance of which program gets the faulty part. For example I might assume that typical memory error is that the value stored in the memory is changed when it is fetched and thus would cause various symptoms from rampant crashes to system freeze. But then again I am no memory expert. (Firefox does seem to be unstable at the moment). Anyway I am going to run memtest seriously this time and I am also trying to borrow some other memory to see if the problems persist. Thanks all for help. -- Rami Saarinen -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: SATA, RAID, A8N-E, 3800+ help
On Thu, Jan 26, 2006 at 11:52:44PM -0500, Keith Ballantyne wrote: I recently bought an Asus A8N-E mobo and AMD64 3800+ CPU. I then bought RAM, a case, and an ATI 300-based PCI Express 16 video card. I used netinstall sarge and managed to install 32-bit using the 2.6 kernel (as the 2.4 kernel wouldn't recognize my network card). ...a few days passed, and I realized that I wanted the 64-bit core. I recompiled the kernel, and then realized there was a 64-bit distribution, so I downloaded both the testing release (Etch) and the 31r1a netinstalls. 31r1a didn't recognize my onboard network, but Etch did. ...a few more days passed and I decided to buy 4 SATA II drives in an attempt to run RAID 0+1. I configured the BIOS and ran the installer, but the installer sees hda through hdd rather than a single RAID drive. In subsequent research (including this list archive) it appears that the BIOS RAID is considered 'inferior' to the software RAID support in Linux. So, my questions are: Unless you have hardware raid, just pretend the system has no raid at all (since really, it doesn't). 1) Why is the ASUS BIOS RAID inferior to software RAID on Linux? Because linux has better software raid than the bios. And it is not tied to specific hardware/bios versions, but can be moved between systems quite easily. It also is more cpu efficient, and gives you better ability to monitor the state of the raid. I suspect if windows had native software raid support, these fakeraid designs wouldn't even exist. 2) Is it possible to install 31r1a instead of the Etch release (I'm not overly keen on working with Etch, but will do it if it's a better 64-bit option...In the few days I played with 64-bit Etch it seemed to work well, but I didn't have most of the utilties I need available from the installer/package manager). I am surprised if the amd64 3.1 doesn't see the network byt i386 3.1 does. They should support the same hardware. There are some detection issues with nvidia ethernet in amd64 3.1, but usually doing 'modprobe forcedeth' on console 2 and going back to the installer fixes that. For some reason the network chip isn't in the pci device listing with 2.6.8 kernel on some nvidiabvoards. The driver loads fine though, when you ask for it manually. 3) Do I need to flash the BIOS for things to work? It sometimes helps. I don't see any problems here that look bios related though. 4) Is ATI 300 support better in Etch? Generally ATI is not what I would recommend in terms of good linux support. You will almost certainly need to use the packages for ati fglrx (a google search will find the unofficial debian packages for it along with instructions). Those may require the use of etch. Not sure. Len Sorensen -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: SATA, RAID, A8N-E, 3800+ help
On Sat, Jan 28, 2006 at 01:10:10PM -0500, Anthony DeRobertis wrote: The OP said any driver, and I pointed out that isn't true. Trying to have users tell the difference between software raid drivers and hardware raid drivers is, I think, harder than just having them tell software vs. hardware raid controllers apart. After all, the latter can be done with a high confidence by looking at the price tag. Linux software raid has a lot of advantages like being able to configure on a per-partition, not per-disk, basis. Many true hardware raids can't even do that. I think most of them can do volumes on a raid, which is essentially partitions. They show up as seperate drives then. Of course you can also partition the raid device since it is just a disk to the OS. However, I haven't actually seen benchmarks showing its faster. I'm not sure how you'd benchmark it, either... You'd have a lot of windows-vs-linux differences to filter out somehow. I was told by a 3ware dealer that many of their users run linux software raid on their 12 port cards, because they find with a modern cpu linux software raid beats the 3ware hardware raid chip. So they use it as a single card 12 driver sata controller instead. I certainly know in the past I got a drop in performance when going from software raid1 to an ibm serveraid 4m doing raid1. I was surprised and rather disappointed in that. The hardware raid was easier to manage though. Len Sorensen -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: high load average?
On Sun, Jan 29, 2006 at 11:21:03AM -0800, michael wrote: On Sun, 29 Jan 2006 18:30:59 +0100, Markus Neviadomski wrote have a look at top a while, when the server is working, for example delivering a website or something other. And, talk about the servers you are running: webserver, gameserver, databases? It is very difficult to help you without any information. It is a simple web, file server. After rebooting it, the load average has gone back to normal. Its just that we have other servers that do way more work and never seem to get over 1.00, while this one is in a small network and was showing some load averages over 1.00 at times. Don't know about your other servers, but if they are still running a 2.4 kernel, there certainly will be a difference. Linux 2.6 changed the way the load average is calculated in such a way that IO time spend by a process is also accounted (or accounted in a different way). Erik -- +-- Erik Mouw -- www.harddisk-recovery.com -- +31 70 370 12 90 -- | Lab address: Delftechpark 26, 2628 XH, Delft, The Netherlands -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: SATA, RAID, A8N-E, 3800+ help
On Sat, Jan 28, 2006 at 05:34:54PM -0500, Keith Ballantyne wrote: For the 31r1a to recognize your network, if it went as in my case, the installer shouldhave proposed you with a list of network modules. You have to choose something call nforether,I believe. Thanks. the 31r1a installer works with the nForce2 network card 10/100 driver on my asus a8n-e. So, I've installed about 6 different ways, utltimately arriving at some critical error somewhere along the way. I've read most of the faq/helps/howtos and archives that I could find, and I still have the same problem. I want to use RAID0 for the bulk of my file system. I would like to even use RAID0 for my boot and swap partitions. I followed the instructions outlined at http://www.howtoforge.com/linux_software_raid?from=10comments_per_page=10 and got further than ever before, but couldn't successfully install either GRUB or LILO. The article above suggests 4 raid partitions /boot (50M) /swap (1G) / (10G) /var (the rest) It is also written specifically for RAID1, though it implies that doing the same for RAID0 would work. I'm worried because I'm unsure if the MBR is 'striped' in RAID 0, or if the MBR's of each of my 4 drives are exempt. Regardless, I've tried it three ways: case 1: /boot as a specific 50M partition, on the 1st SATA drive (ext3) Makes sense. case 2: /boot as a RAID1 collection of the first 50M worth of all 4 SATA drives (ext3). Raid 1 only does two drives. At least in normal setups. You can stripe mirrors or mirror stripes of course to make bigger setups. case 3: /boot as a RAID0 collection of the first 50M worth of all 4 SATA drives (ext3). note: In all 3 cases, the remainder of the drive was configured as a RAID0 array, and I used LVM (1 volume group) to created two logical volumes, a 4G swap volume and the rest (to mount as /). case 1: allowed me to complete installation, grub appeared to install okay, but it doesn't boot, I just get a blank screen. case 2: the partitioner seemed to change the partition types, and would never let me install GRUB. Grub works fine if you do raid1 on two partitions. Trying 4 won't work. Too complex for the boot loader. case 3: would never allow GRUB (or LILO) to install on the RAID0 array. It kept indicating that status couldn't be reread from /dev/md/0. When I rebooted, I would reconstruct the partitions as outlined in case 3 above, and the same error would occur. The boot loaders don't know raid0 as far as I know. So, right now I have everything installed except an operational bootloader. Make /boot a raid1 using a 50M partition on the first two drives. Install grub to the MBR of both those drives. Or just make it a plain partition. If you use raid0 at all, reliability is obviously not an issue for the system overall, and performance for /boot is not an issue either so a plain single partition on the first drive is fine. Len Sorensen -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Installing Debian on sata drive with adaptec aic-8130 controller
On Mon, Jan 30, 2006 at 12:27:14PM +0100, Marc Blumentritt wrote: Thanks for the info. I will write my hardware-vender, that when I say, I want Linux compatible hardware, that I really mean it. I will replace the controller. Any sugesstions for a SATA PCI controller with 4 ports? Yeah, redhat/suse binary drivers != Linux support. Hmm, well for supported chipsets there are: Sil3112/3114 3ware cards (which do hardware raid.) Areca (although their drivers have not yet been merged into the kernel, so they take more work. Also hardware raid.) Some promise based cards are supported too, although I am not currently up to date on which ones. Len Sorensen -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Jumbo Frames?
I have recently moved my servers to an all gigabit ethernet network and I'd like to make use of jumbo frames. They're enabled on my switch (it's a Dell managed switch) - however, I am not sure if I have to make any changes to the servers themselves (i.e. changing the default MTU from 1500 to 9000?) - and I haven't been able to find anything by Googling that helps. I'm running Debian-amd64, so this isn't completely off-topic :) Thanks in advance for any replies, I'm sure plenty of people here use it or know how I can figure this out and confirm! - mike
Problems with apache-perl since perl upgrade
I'm writing this list as the issue seems to be architecture related. Since the recent perl security upgrade I have cgi scripts failing with errors like this: [error] Errno architecture (x86_64-linux-thread-multi-2.6.11.6-vs195-farm1.0) does not match executable architecture (x86_64-linux-thread-multi-2.6.10-9-amd64-k8) at /usr/share/perl/5.8/CGI/Carp.pm line 314.\nCompilation failed in require at /usr/lib/perl/5.8/IO/Socket.pm line 17.\nBEGIN failed--compilation aborted at /usr/lib/perl/5.8/IO/Socket.pm line 17.\nCompilation failed in require at (eval 366) line 5.\n at /usr/lib/www-perl-public/newpass_weather_query line 18\nBEGIN failed--compilation aborted at /usr/lib/www-perl-public/newpass_weather_query line 18.\n I'm running: apache-perl 1.3.33-6sarge1 perl5.8.4-8sarge3 GNU/Linux 2.6.8-11-amd64-k8-smp #1 SMP Sun Oct 2 23:21:12 CEST 2005 x86_64 I would appreciate any help dealing with this. Thanks. Don Hayward at pomobuli.net -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Problems with apache-perl since perl upgrade
* Don Hayward [EMAIL PROTECTED] [Jan 30. 2006 18:03]: I'm writing this list as the issue seems to be architecture related. Since the recent perl security upgrade I have cgi scripts failing with errors like this: [error] Errno architecture (x86_64-linux-thread-multi-2.6.11.6-vs195-farm1.0) does not match executable architecture (x86_64-linux-thread-multi-2.6.10-9-amd64-k8) at /usr/share/perl/5.8/CGI/Carp.pm line 314.\nCompilation failed in require at /usr/lib/perl/5.8/IO/Socket.pm line 17.\nBEGIN failed--compilation aborted at /usr/lib/perl/5.8/IO/Socket.pm line 17.\nCompilation failed in require at (eval 366) line 5.\n at /usr/lib/www-perl-public/newpass_weather_query line 18\nBEGIN failed--compilation aborted at /usr/lib/www-perl-public/newpass_weather_query line 18.\n I'm running: apache-perl 1.3.33-6sarge1 perl5.8.4-8sarge3 GNU/Linux 2.6.8-11-amd64-k8-smp #1 SMP Sun Oct 2 23:21:12 CEST 2005 x86_64 I would appreciate any help dealing with this. Thanks. Don Hayward at pomobuli.net -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED] Hi Don, I'm not sure if this is the same problem as I encountered on a i386 based box, but I did receive similar output: http://www.archivum.info/linux.debian.bugs.dist/2006-01/msg06209.html Did you restart apache after the upgrade? Cheers, Jonas -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Problems with apache-perl since perl upgrade
* Don Hayward [EMAIL PROTECTED] [Jan 30. 2006 18:03]: [error] Errno architecture (x86_64-linux-thread-multi-2.6.11.6-vs195-farm1.0) does not match executable architecture (x86_64-linux-thread-multi-2.6.10-9-amd64-k8) at /usr/share/perl/5.8/CGI/Carp.pm line 314.\nCompilation failed in require at /usr/lib/perl/5.8/IO/Socket.pm line 17.\nBEGIN failed--compilation aborted at /usr/lib/perl/5.8/IO/Socket.pm line 17.\nCompilation failed in require at (eval 366) line 5.\n at /usr/lib/www-perl-public/newpass_weather_query line 18\nBEGIN failed--compilation aborted at /usr/lib/www-perl-public/newpass_weather_query line 18.\n I'm running: apache-perl 1.3.33-6sarge1 perl5.8.4-8sarge3 GNU/Linux 2.6.8-11-amd64-k8-smp #1 SMP Sun Oct 2 23:21:12 CEST 2005 x86_64 I would appreciate any help dealing with this. Thanks. Don Hayward at pomobuli.net Hi Don, I'm not sure if this is the same problem as I encountered on a i386 based box, but I did receive similar output: http://www.archivum.info/linux.debian.bugs.dist/2006-01/msg06209.html Did you restart apache after the upgrade? Hi Jonas, Thanks, this seems to be a different problem -- Yes, I restarted apache-perl. And it is related to the IO modules. It occurs in scripts that use IO::File. Don Hayward at pomobuli.net On Mon, 30 Jan 2006, Jonas Bardino wrote: -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Jumbo Frames?
On Mon, Jan 30, 2006 at 07:26:56AM -0800, mike wrote: I have recently moved my servers to an all gigabit ethernet network and I'd like to make use of jumbo frames. They're enabled on my switch (it's a Dell managed switch) - however, I am not sure if I have to make any changes to the servers themselves (i.e. changing the default MTU from 1500 to 9000?) Yes. Set the MTU to 9000 on all machines. -- Lionel -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Jumbo Frames?
Thanks for the reply. I've tried that and now this is giving me confusing results: A has jumbo frames (mtu 9000) B has jumbo frames (mtu 9000) C has normal (mtu 1500) B - A = 3.4MB/sec C - A = around 50MB/sec I'm just scp'ing a 170 meg video file back and forth - it's over 10 times slower... does anything stand out to you instantly as to what the problem would be? Thanks in advance. - mike On 1/30/06, Lionel Elie Mamane [EMAIL PROTECTED] wrote: Yes. Set the MTU to 9000 on all machines.
Re: Jumbo Frames?
On 2006-01-30T13:58:56-0800, mike wrote: I'm just scp'ing a 170 meg video file back and forth - it's over 10 times slower... does anything stand out to you instantly as to what the problem would be? The difference does not make sense, but you do incur CPU overhead for encryption with scp so it might not be the best test. /Allan signature.asc Description: Digital signature
Re: Jumbo Frames?
Yeah - but I shouldn't be incurring that much on a dual-core opteron that's idle to slow it down by 10x on the same protocol :) Maybe FTP vs. SCP... but not a common denominator... On 1/30/06, Allan Wind [EMAIL PROTECTED] wrote: On 2006-01-30T13:58:56-0800, mike wrote: I'm just scp'ing a 170 meg video file back and forth - it's over 10 times slower... does anything stand out to you instantly as to what the problem would be? The difference does not make sense, but you do incur CPU overhead for encryption with scp so it might not be the best test. /Allan -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQFD3o45uDtNyOwreTYRArLGAJ9rUuIuMuPWFc51foiPigW9slUihwCfZqJM KP26gQqHcyqKFkNfaGKV0tE= =Bi6Q -END PGP SIGNATURE-
Re: Jumbo Frames?
On Mon, Jan 30, 2006 at 01:58:56PM -0800, mike wrote: Thanks for the reply. I've tried that and now this is giving me confusing results: A has jumbo frames (mtu 9000) B has jumbo frames (mtu 9000) C has normal (mtu 1500) B - A = 3.4MB/sec C - A = around 50MB/sec I'm just scp'ing a 170 meg video file back and forth - it's over 10 times slower... does anything stand out to you instantly as to what the problem would be? C - A will end up at 1500 since C can only do 1500, obviously. B - A, your switch is choking, is my *guess*. I suspect your switch it not working as well as you think it is, unfortunately. Is it firmware upgradeable? If so, check for updates. Also check the Dell forums. I have a unmanaged Dell switch, it exhibits the same behavior. Unfortunately, not upgradeable in my case ~:^( You should get better than 50MB/s however. What are the NICs? Using a hacked version of netcat, to eliminate disk I/O from the picture, I see steady 90ish MB/s. Between two 64/66 Intel NICs across my Dell switch. Between any other NIC hardware combos, I see 70-80 MB/s max. Actually, the only other hardware I have that isn't 32bit is a Broadcom 57XX, also 64/66. On a dual core, or even a single core, Opteron, the encryption overhead would be minimal. The heavy encryption calculations comes at key generation time, ie., when the connection is setup. Once it's going, I'm going to estimate that the overhead from encryption is less than 1% of CPU. The default encryption algorythms are pretty fast and efficient. Try netcat and just time it, for cleaner results. After the first time you run it, your caches should be warmed up, so you should get pretty clean results of just the network xfer speed. a -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Jumbo Frames?
On 2006-01-30, at 2:07 PM, Allan Wind wrote: The difference does not make sense, but you do incur CPU overhead for encryption with scp so it might not be the best test. This is a good point assuming that you're not using triple-DES this should be something like a factor of two (which is what I see with my aged 3GHz Xeon - G5|Opteron server) and using -c blowfish will significantly reduce that (45MB/s vs. 30MB/s). I think the bigger factor is that it requires the CPU to touch all of the data and thus prevents any sort of zero-copy optimization. Chris smime.p7s Description: S/MIME cryptographic signature
Re: Jumbo Frames?
On 1/30/06, Andrew Sharp [EMAIL PROTECTED] wrote: C - A will end up at 1500 since C can only do 1500, obviously. B - A, your switch is choking, is my *guess*. I suspect your switch it not working as well as you think it is, unfortunately. Is it firmware upgradeable? If so, check for updates. Also check the Dell forums. I have a unmanaged Dell switch, it exhibits the same behavior. Unfortunately, not upgradeable in my case ~:^( I sure hope not! It's a Dell PowerEdge 5324, and it touts all sorts of goodies. You should get better than 50MB/s however. What are the NICs? Using a hacked version of netcat, to eliminate disk I/O from the picture, I see steady 90ish MB/s. Between two 64/66 Intel NICs across my Dell switch. Between any other NIC hardware combos, I see 70-80 MB/s max. Actually, the only other hardware I have that isn't 32bit is a Broadcom 57XX, also 64/66. All the servers are identical config: * Dual-core Opteron 175 * 2x1 gig DDR3200 ECC * Seagate 250G SATA * Serverworks HT1000 chipset on a Supermicro H8SSL-i motherboard (specs: http://www.supermicro.com/Aplus/motherboard/Opteron/HT1000/H8SSL-i.cfm) * appears to be Dual-port Broadcom BCM5704C Gigabit LAN / Ethernet Controller Are there any sort of sysctls or echo foo /proc/somesettings that might help get more benefit out of jumbo frames? No magtter what I would expect it to at least be the same speed. Not 10x worse.
Re: SATA, RAID, A8N-E, 3800+ help
Lennart Sorensen wrote: I think most of them can do volumes on a raid, which is essentially partitions. They show up as seperate drives then. Of course you can also partition the raid device since it is just a disk to the OS. No. That's different from what Linux software raid can do. Linux software raid can say, ok, use sda1, sdb1, sdc1, sdd1 as raid1; sda2, sdb2, sdc2 as RAID-5 with sdd2 as a hot spare; and finally, sda3, sdb3, sdc3, and sdd3 as raid0. Most hardware raid can't do that. I was told by a 3ware dealer that many of their users run linux software raid on their 12 port cards, because they find with a modern cpu linux software raid beats the 3ware hardware raid chip. So they use it as a single card 12 driver sata controller instead. Interesting. I have several 3ware 9000s in the field, never tried that... The hardware raid was easier to manage though. Definitely. Especially with libata's hot-swap deficiencies. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Please help - kernel crashes often
mike wrote: Sure enough, one of the two others I tested failed within only 10-15 mins. Does this look appropriate for a dual-core, single chip, Opteron 175 in a 1u chassis? The max any CPU gets is 62C... That's pretty damn hot. Something closer to 40C is more normal... -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Please help - kernel crashes often
Harald Dunkel wrote: Hi Mike, A machine check exception indicates a hardware problem, i.e. a broken CPU. (I am not sure whether it could indicate bad ECC memory, too. Did you run memtest68?) ECC failures will generate MCE's. The MCE message *should* provide some hint as to what is wrong. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Jumbo Frames?
mike wrote: Thanks for the reply. I've tried that and now this is giving me confusing results: A has jumbo frames (mtu 9000) B has jumbo frames (mtu 9000) C has normal (mtu 1500) B - A = 3.4MB/sec C - A = around 50MB/sec Check to make sure that B is operating in 1000mbps, full duplex, using ethtool, and check to make sure the switch agrees. Duplex mismatches will kill any ethernet link. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Debian ADM64 Etch (testing/unstable) system freeze
Rami Saarinen wrote: Well, somehow I assumed that if the fault is in memory, it is probably in a fixed location Depends on the type of memory problem. Memory problems can cover everything from this one certain bit is stuck at 0 (what you're thinking of) to the memory timings/voltage/whatever are off, memory functions as a hardware random number generater as a result. Oh, and memory allocation is not random. The kernel is going to wind up in a certain spot every time. So will, e.g., init. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
RE: SATA, RAID, A8N-E, 3800+ help
I originally attempted to send this 29 Jan at 3:30PM, but it was rejected by my mail relay server, so I'm attempting to resend to provide a follow up on the thread. Does your installation work, are you able to boot it by other means? For example, try using a Grub-floppy, and see if that lets you boot. Thanks. It turns out I followed a similar path, but from a different source. Grub was hanging on install (I let it sit in a 50% complete state for about 8 hours to make sure...it was definately not doing what it was supposed to do). So I went to: http://xtronics.com/Reference/SATA-RAID-Debian.htm The instructions there (as everywhere else) don't necessarily say that RAID0 cannot be used for the boot partition, but I was guessing that was the problem. Since RAID0 is a stripe-set across multiple drives, I reasoned that it couldn't load a bootloader from them. To rectify this, I made my /boot mount be RAID1 rather than RAID0. When I did that, GRUB installed fine, but would not boot. So I followed the steps outlined in the web page above, notably the section where you start Grub (I did it from a floppy after invoking grub-floppy from an alternate console during one of my install attempts) and execute: root (hd0,0) setup (hd0) root (hd1,0) setup (hd1) ... In my case, I have 4 drives, so I continued through (hd3). After that grub successfully booted the system. So, now I'm on to making a chroot ia32 so I can get OpenOffice running. Ah, the joy :) - Keith -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
strages NIC lockups
After installing amd64 Debian on a SunFire x2100 I shelled into it to continue and the ethernet link failed. Mystified, I reloaded the forcedeth driver (eth0) without a problem but the interface still would not work. There were no other sign of failure other than it just did not work. The other interface (eth1) was fine. Rebooting did not help but power cycling did the trick. This happens at random times and power cycling corrects the problem. kernel: 2.6.12.6-xen It has failed with and without domU's running. Regards, David Koski [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Debian ADM64 Etch (testing/unstable) system freeze
Anthony DeRobertis wrote: Rami Saarinen wrote: Well, somehow I assumed that if the fault is in memory, it is probably in a fixed location Depends on the type of memory problem. Memory problems can cover everything from this one certain bit is stuck at 0 (what you're thinking of) to the memory timings/voltage/whatever are off, memory functions as a hardware random number generater as a result. Yes, very true. Oh, and memory allocation is not random. The kernel is going to wind up in a certain spot every time. So will, e.g., init. Yes. Somehow I ended up thinking that if the fault is in the memory area the kernel uses, the faulty behaviour would be more devastating and would occur more ofter. After all I have ran the system for hours without a problem. Anyway, I am glad to inform that yes it really was the memory that was causing the trouble. I let the machine run the memtest86+ last night and after 10 hours it had found four memory errors. Apparently I was too hasty at the first time. I have one more stupid question: as it may take couple of days for me to get the new memory. Is there any way to block / reserve the faulty memory area so that it would not be available for use? Thanks again for help! -- Rami Saarinen -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
