Re: 8.04 server + ubuntu-xen-server dies under disk load on AMD Opteron

[David Kempe]

James Dinkel wrote:
>
> Have you thought about using KVM instead?  I know that doesn't solve
> your problem.  But KVM is built into the standard linux kernel.  Xen
> is a fork of the linux kernel.  So KVM should work just as reliably as
> your typical linux install.
>
>   
Also along the same line of recommendation - have you considered OpenVZ? 
If you have all linux guests, the OpenVZ container approach is far more 
performant. We have converted our whole hosting environment from Xen to 
OpenVZ because of poor disk and network performance with Xen. Haven't 
looked back - its been awesome.

Sorry to not actually fix your problem though :) (I think someone else 
has in this thread)

thanks

dave

-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

Re: XFS In Dapper [previously posted to ubuntu-users]

[David Kempe]

Nick Webb wrote:
> I've got a couple projects coming up that will have a file systems >= 
> 2TB and I'm thinking of using XFS for it.  Main feature of XFS I need is 
> the lack of fsck at startup (fsck for ext2/3 will take many hours with a 
> 2TB partition).  The file system will also likely have many large files, 
> so XFS seems to be a good choice for this as well.
>
>
>   
XFS is good, we use it on dapper all the time. My largest XFS filesystem 
is 5.5TB formatted.
I have to say 64-bit is the only way to go for this - the xfs repair 
tools can't handle larger filesystems in 32bit mode.
btw, one thing I found was that xfs_repair can chew massive amounts of 
ram to run a repair on a filesystem. I had a 2TB fs take nearly 8gb of 
ram (and swap) to repair it. It did a good job of repairing, and took 
ages. So XFS is not free of the fsck problem, just xfs_check is faster 
and perhaps less thorough (dunno).

Importantly, you can have data-loss on XFS if you lose power suddenly, 
perhaps more so than ext3. When files get corrupted on XFS, I have 
noticed they go to zero size, whereas in messy situations with ext3 I 
have noticed you are more likely to loose metadata than data. I still 
would stick with XFS anyday though, even just because the sheer increase 
in format time.

I have had good results on many different types of block devices as well.

thanks

dave


-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

Re: Forum Poll Questions

[David Kempe]

Michael Hipp wrote:
> Should we include some of the "big iron" like IBMs?
>
>
>   
does ubuntu run on them? (without building the entire OS from scratch?)

dave

-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

Re: File Server Tasks

[David Kempe]

Neal McBurnett wrote:
> I don't know if this is a "wild" objection :-) , but I'd suggest taking the
> NFS stuff out and using Jamie's name:
>
>  Samba (Windows file sharing) -- task
>
> or something like it.
>   


I would agree with that. I can't think of someone asking about getting 
NFS to work on our LUG lists recently or in my day to day work. I think 
we should just stick to samba, even for linux->linux sharing

dave

-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server
More info: https://wiki.ubuntu.com/ServerTeam

Re: 3ware 9650se

[David Kempe]

Ante Karamatić wrote:
> I could only say that 9650 is the best SATA controller I've worked with. 
> As for 5TB... If you create >2TB partition, make sure you have BIOS that 
> supports EFI boot loaders and install EFI loader after install, but 
> before reboot. If that's not the case, create small disk (not 
> partition!) for /. If you are doing software raid, then you will not 
> have these problems.

thanks. thats great info and probably saved me lots of time.
I have already planned to not boot off the large array - done that 
before and had troubles with the whole efi thingo. (I have some other 
machines that boot off usb drives because of this very problem!)

kernel-team - +1 for new driver updates in the kernel/install CD. This 
is an important part of an LTS release surely, and I am sick of buying 
servers has not having the nics/sata/ide/southbridge chips supported :( 
I have had like 5 firewalls in the last month or so with this problem. 
Had to take alot more effort in specifying hardware with my supplier 
than normal because of it :(

dave

-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

3ware 9650se

[David Kempe]

Gday,
wondering if anyone is using the above card with dapper?
Particularly, are you using software raid and jbod, or native raid6? 
kernel work ok?

I am looking to get one of these and a 3ware back plane for a file 
server I am looking to build.
Any comments or recommendations? I need 5TB of storage. I am tempted to 
go with software raid6, as I like having linux as the only raid 
controller I have to rely on :)

thanks

dave

-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: Loginprompt shows up before booting is finished in Ubuntu Feisty server

[David Kempe]

Anders Häggström wrote:
> Anyone know why this happens?

upstart is event driven, so things happen after other things. The login 
prompt comes up, but is overwritten by other things happening in the 
background. This will be different for different hardware etc, as the 
event driven init works that way.

the upside in decreased boot time, the downside is that things may 
happen in strange orders sometimes...

dave

-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: Ubuntu Server on Soekris

[David Kempe]

I have used a couple of board that are similarly specced and its fine.
We also use ubuntu on very low end Via EPIA board for a variety of 
applications (monitoring, firewalls, time-lapse photography). You need 
to install the -386 kernel on these chips, but aside from that its all good.

dave

Ingo Lantschner wrote:
> Hello,
> does someone have experiences with Ubuntu Server on Soekris Hardware?  
> Any links/hints would be fine!
> -> http://www.soekris.com/
> TIA Ingo.
>   


-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: Changing permissions of postfixs "local"

[David Kempe]

Hi Ingo,
You don't need to do that.
Just make a seperate file for the nagios alias and make it owned by the 
correct user and group. Postfix local agent should then deliver mail 
with the perms of the alias file.

From man 8 local:

*DELIVERY RIGHTS*
  Deliveries to external files  and  external  commands  are
  made with the rights of the receiving user on whose behalf
  the delivery is made.  In the absence of a  user  context,
  the *local*(8)  daemon uses the 
owner rights of the *:include:*
  file or alias database.  When those files are owned by the
  superuser, delivery is made with the rights specified with
  the *default_privs 
* configuration parameter.


I have done this before for this exact problem and it works fine

dave


Ingo Lantschner wrote:

Hello,
I am using Ubuntu Server 6.06 as base os for a Nagios systemmonitor. 
Yesterday I was looking for a way to feed alerts send as emails from 
RAID-controllers like 3ware. First it was quite simple: Adding a line to 
/etc/aliases

nagios: "|/usr/local/nagios/libexec/eventhandlers/handle-RAID-mail"

Writing the script was not the problem, but the permissions were. Nagios 
uses a pipe which is for obvious reasons not writable by nobody:nogroup.


In order to get around this problem I changed the user runing "local":
$ sudo postconf -e default_privs=nagios

Now the emails go straight into Nagios.

BUT: What are the security implications of tampering with the 
permissions of postfix? Any input is welcome - tia Ingo.


  
-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: motd - changes are lost during reboot

[David Kempe]

Ingo Lantschner wrote:
> Hello,
> a small but anoying problem, which I could not solve yet: Changes in / 
> etc/motd are lost during reboot. Why? How can I customize motd?
>
> TIA for any tipps! Ingo.
>   
check /etc/default/rcS

dave

-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: Sudo or NOT to sudo

[David Kempe]

Yeah, or just put the host in dns if you have that sort of control over 
your DNS setup :)

dave

Ante Karamatic wrote:
> No, this is not the only way to fix it. Actually, reinstall isn't a fix
> at all. You can fix it by booting in single user mode and fix
> your /etc/hosts and/or /etc/hostname.

-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: LAMP Production Server - Dell 2850

[David Kempe]

One 2850 should be able to handle 300 users on a fairly lightweight
php/mysql app. Your app will probably become CPU bound quickly,
obviously depending on its nature. You need to scale the machines in the 
direction of the problem. If you app is db intensive (lots of small 
transactions or accesses), you may even want a php/mysql pair on each 
server, and scale that way. I know a large add provider that does things 
that way, and use mysql replication.

Scaling totally depends on the application at hand, and odds on, you 
need to simply start with one machine, figure out what the bottleneck is 
after some exhaustive testing, and then design a solution that works. 
Even if that means tweaking the app to scale better. You probably need 1 
dev server, with monitoring and other services on it as well (cacti and 
nagios etc) and a testing server that is exactly the same as production 
servers. This is where Xen comes into its own :)

LVM is great, and I use it alot, but there are times when it doesn't 
make alot of sense. If you think you are going to add storage to the 
machines, then LVM make sense. When we build Xen servers, we absolutely 
use LVM to chop up the disks into resizable partitions. What you end up 
doing is figuring out what is best for the application. Servers serve 
the application :)

dave

Bill Asher wrote:
> David, In your recommendation, would there be any need for LVM then?
> I'm doing a Ubuntu LAMP install, after install how would you handle
> all the smylinks you are talking about?  I agree with your thoughts
> but am unsure how this effects your other partitions needed.
> 
> This is what I have now using LVM for the RAID5 drives: $ df -h 
> FilesystemSize  Used Avail Use% Mounted on /dev/sda1
> 62G  227M   59G   1% / varrun   1014M   48K 1014M   1%
> /var/run varlock  1014M  4.0K 1014M   1% /var/lock udev
> 1014M   96K 1014M   1% /dev devshm   1014M 0 1014M
> 0% /dev/shm /dev/mapper/data-home 34G  129M   32G   1% /home 
> /dev/mapper/data-usr   31G  364M   29G   2% /usr /dev/mapper/data-var
> 138G  227M  131G   1% /var
> 
> 
> 
> The application we are producing is internal only via web browser.
> Using Apache2, PHP5 and MySQL5.  My thoughts were to build 2 servers
> with Apache2 & any web services and 2 servers with MySQL for database
> services. One of each being production and development servers. This
> app will be used for about 300 users.  In your opinions is this over
> kill?  Or just do 4 servers each having Apache2 and MySQL on them, 2
> production and 2 development?

-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: LAMP Production Server - Dell 2850

[David Kempe]

Hi,
I am over having separate partitions for the sake of it. You just end up 
running out of space when you least need it.
I would suggest you just use the raid1 drives for the system, and 
allocate the whole raid5 to /home and redirect things like /var/www to 
/home/www.
I imagine the users are going to be users of the web application not 
system users.
oh and btw, grsecurity doesn't work with dapper's libc very nicely.
We have actually repatched it and are testing a very of dapper's libc 
that does work with grsec, but I doubt it will be widely available very 
soon.

Partitioning is a very application specific choice, but I would warn 
against too fine grained partitioning.

dave


Alejandro Sanchez Marín wrote:
> Hi Bill, i have a similar configuration and best practice is:
> 
> Logical drive 0: Put here / and swap partition if you need it.
> 
> Logical drive 1: Use LVM to split RAID5 hardware into 3 partitions and
> put here /var, /usr and /home partitions. About partition distribution

-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: LAMP Production Server - Dell 2850

[David Kempe]

Hi Bill,
Yes I have installed on a Dell 2850. Dapper should work fine.
Ubuntu rocks for servers, just like Debian did previously.

go for it :)

dave

Bill Asher wrote:
> We are planning on rolling out an internal web application to LAMP.  I’m 
> at the point of choosing a Linux distro that will be stable and growing 
> community.  I am fimilar with Linux and mostly FreeBSD in a production 
> environment.
> 
>  
> 
> I’m looking for encouragement to go with Ubuntu Server in our production 
> app because I really like the distro and its outlook.  Server admin, how 
> has your experience been, should we go with Ubuntu and has anyone 
> installed on Dell PowerEdge 2850s
> 
>  
> 
> Thanks!!
> 
>  
> 
> Bill
> 
>  
> 
> **B. Asher**
> 
> Network Administrator
> 
> National Auto Warranty Services
> 
> W: 636.625.2552 x.: 7715
> M: 314.749.6936
> 
> [EMAIL PROTECTED] 
> 
>  
> 

-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: Anyone using EVMS?

[David Kempe]

I am considering using it for an unfinished task for a client. They have 
1RU servers with 4 750Gb drives in them for file storage. I am having 
trouble with all the available tools, as the onboard hardware raid makes 
that into a 2.3Tb volume and the bios can't boot that. So we are booting 
off usb keys, but now fdisk and cfdisk and lots of other tools have this 
2TB limit. EVMS doesn't seem to suffer the same problems.
I haven't had time to fix it up yet, so I will report back once I do.

thanks

dave

David Abrahams wrote:
> Ubuntu installs and activates EVMS by default, but gives us no way to
> set up EVMS volumes in any installer, and AFAICT all of EVMS'
> functionality is provided by mdraid and LVM.  I've been unable to get
> a convincing answer out of the EVMS people about why anyone would
> convert a filesystem to EVMS
> (http://thread.gmane.org/gmane.linux.evms.devel/6446/focus=6464) when
> you can get all the touted UI benefits of the system by using the EVMS
> UI tools on LVM/mdraid volumes, which begs the question about why EVMS
> conversion writes any new metadata to the disk and what EVMS uses that
> metadata for.
> 
> So, if you're using it, what benefits are you getting?
> Are you just using the UI, or do you have a true EVMS filesystem?
> 
> And if nobody's using it, shouldn't the installer just leave it out?
> 

-- 
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: Ubuntu a poor choice for servers

[David Kempe]

whats there to look into?
we do it all the time. I even got it to work with hoary once you updated 
 mdadm to start at the right time.


dave

Carl Karsten wrote:

David Kempe wrote:


/boot on software raid is fine. did it today even


On the grub todo list:http://grub.enbug.org/TodoList
"Add support for software RAID. Add support for  LVM"

So normal grub does not support Raid, (or they haven't taken it off the 
list.)


But, Ubuntu's grub is patched.  But I have never bothered to see what 
the patch does.  Can someone look into that?


Carl K


--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: Ubuntu a poor choice for servers

[David Kempe]

David Abrahams wrote:

Could you give more details?  If I have to complain to the person who
sold me the hardware (knowing I'd be running linux on it), I want to
be able to justify myself.


the forcedeth driver is reverse engineering, much like the nv driver 
versus the nvidia binary driver. I think you could get the nvidia binary 
drivers to work with the nvidia nforce chipset/nic, but am certain thats 
not going to play nice with xen (at least not until xen becomes a subarch).
we have had crazy issues with forcedeth cards - like lockups where the 
card stops tx/rxing data completely, and only way to fix it is to pull 
the power cord, a reboot doesn't fix it. its related to acpi stuff 
mostly. also the driver is slow to get link, which causes problems for 
init scripts and it generally crap, not supporting features I like to 
use like ethtool or mii-registers.
I don't doubt for a second its going to get better, reverse engineering 
is always a work in progress, so maybe its just a matter of time. But 
frankly, if nvidia just damn opensourced decent drivers it would be a 
whole lot better. Its a freaking network card for crying out loud 
(although I know is all integrated in the chipset - probably to reduce 
the cost).
the better tyan mobos have broadcom nics in them, and AMD chipsets, all 
very well supported.


dave

--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: Ubuntu a poor choice for servers

[David Kempe]

/boot can't be on LVM - grub doesn't seem to like it. Though that might 
be fixed in dapper

/boot on software raid is fine. did it today even

dave

David Abrahams wrote:

1. Can the /boot partition be on a software RAID device?
2. Can the /boot partition be on an LVM logical volume?



--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server

Re: Ubuntu a poor choice for servers

[David Kempe]

David Abrahams wrote:

One concern I have about softraid is that there's (AFAICT) no RAID
possible for the /boot partition.


We use dapper and breezy on nearly all of our servers where we get the 
decision (thats 10s a week). Our base Xen dom0 server is software raid 1 
/boot and software raid 1 LVM, with 2gb /root VG for the dom0. The rest 
of the raid 1 is for lvm to allocate to domU machines. It works fine. 
The installer is fiddly to get to that situation, and in breezy there is 
a few strangenesses, but dapper seems fine with it all. We even adapted 
the dapper server cd to do all that automatically from a boot cd, and it 
works great. netinstall from 22mb iso with auto LVM and software RAID.

So saying it doesn't work is just plain crazy.
Anyway - that motherboard you have has got some issue. The onboard 
nvidia chipset is crap, mostly because the reverse engineered forcedeth 
drivers are seriously problematic. I have a few of those boards in 
servers, and have ended up with different network cards in them.



I wouldn't combine LVM with softRAID either, but well, that's just
my opinion.


Why not?



lvm and software raid works fine here.

dave

--
ubuntu-server mailing list
ubuntu-server@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-server