vzdump bogs down whole proxmox machine

[voodoo]

hi

I have a Proxmox machine with 16 gigs of ram an 3.2 ghz i7 and 2 terrabytes of diskspace
no raid or nfs, proxmox 1.9.
we have a very old program running which got virtualised with proxmox its a win2k3 server aprox. 250 gig big
Everytime i run vzdump it totally slows the whole server to a halt, its so bad it takes some times up to 5 minutes to get
the web interface to respond.

are there other ways to transfer our vm image to our new 2.1 server?

 

[mattnux]

hi

I have a Proxmox machine with 16 gigs of ram an 3.2 ghz i7 and 2 terrabytes of diskspace
no raid or nfs, proxmox 1.9.
we have a very old program running which got virtualised with proxmox its a win2k3 server aprox. 250 gig big
Everytime i run vzdump it totally slows the whole server to a halt, its so bad it takes some times up to 5 minutes to get
the web interface to respond.

are there other ways to transfer our vm image to our new 2.1 server?

yes of course you can move it to your new 2.1 server.

 

[gkovacs]

There are two things that can be done to mitigate the slowdown during vzdump backups:

1. Limit the IO bandwidth of vzdump
You can do this by adding the following line to /etc/vzdump.conf:

bwlimit: 20480

This will help server uptime by ensuring the network (and sequential file reading) IO will not exceed 20 MBytes / sec, so on a gigabit interface your services should run fine. This of course only helps if you backup over the network.

2. Change the Linux default IO scheduler to "deadline"
You can do this by adding "elevator=deadline" to /boot/grub/menu.lst, at the end of your running kernel line, so it looks something like this:

kernel          /vmlinuz-2.6.32-6-pve root=/dev/mapper/pve-root ro quiet elevator=deadline

More info here:
http://www.serverwatch.com/tutorials/article.php/3819446/Selecting-and-Tuning-a-Scheduler.htm

This will magically turn your server back to working normal during vzdump backups (and also much more responsive during any kind of heavy disk IO), regardless of using a one-disk setup or a 8-disk RAID array.

It's a shame Proxmox still uses the lame CFQ scheduler when the global Linux server community is starting to agree that CFQ is not suitable for server environments.

 

[tom]

I would not say that is that clear as you point it out here. AFAIK deadline does not support ionice (and some other features) so it really depends on the workload and a lot of other settings which one is "better".

 

[voodoo]

well now i ran into a new problem its not possible to copy this big imagefile to any other machine or device it everytime stalls exactly at 100 GB no matter if its scp ftp or cp to a usb disc
it looks like there is maybe a problem with the file system in itself. But the backup program inside my vm does a full backup everyday without a hitch so it must be releated to linux fs.
maybe its just the file size? is 300 GB too much too handle for ext3?

 

[dietmar]

is 300 GB too much too handle for ext3?

no, ext3 can handle up to 4TB.