VM stops after file transfer

  • Thread starter Thread starter TiagoRF
  • Start date Start date
T

TiagoRF

Guest
Hello!

Our fileserver stops (literally, goes to the stopped state) after migrating over 60gb of files.

Anyone experiencied this?

E1000 NIC, regular fileserver.

Cheers
 
udo said:
Hi,
is the e1000-nic on the host, or at the vm (or both)?

Udo
Click to expand...


I'm working with TiagoRF on this implementation, and the host is a Dell R610 (2 XEON Quadcore and 8GB RAM) with 4 Broadcom Gigabit Cards.

We have Proxmox 1.4 with 2.6.24 kernel and 2 VM (File Sharing and ISA), has reported here we opted to use e1000 in the 2 VM with intel drivers (from their website).

Each VM connect to a bond, like this

2 Physical broadcom - bond0 - Balance-rr - e1000 driver

2 Physical broadcom - bond1 - Balance-rr - e1000 driver

We have this results when high load happen

1 - the network card from the VM (File Sharing) dont respond...we need to restart the VM..even though the other bond and the other vm in the same host behave normally.

2 - the file sharing vm just STOPPED, status as stopped, the other vm behave normally as well in this situation.

I have go further and made some investigation on this the server in proxmox have 7.78 gb of ram and we were using

ISA - 4096 RAM and FS - 3072 RAM.... maybe this is too much because proxmox need for itself as well...

I have decided to decrease memory for the VM ISA - 3072 and FS - 3584.

And change bond method to active - backup....

Any ideas beside this ones....
 
Follow up to Sniffer's

Apparently our conflicts have been solved by adjusting the memory.

All the VM's seems stable now, and in production.

Dietmar, we haven't tried the latest kernel yet, but Sniffer states he's using it on Arch Linux and it's unstable regarding some bugs accepting some drivers etc.

However, we'll try it in a controled environment and we'll speak our minds.
 
by the way,

in PM VE 1.4 the branch is 2.6.24, and PM VE 1.5 (Default) came back to 2.6.18, that sure means something, correct?

Regards
 
tom said:
yes, that means in 1.4 we had 2.6.24 and in 1.5 we have 2.6.18 as default (and 2.6.24 and 2.6.32 as an option).

see http://pve.proxmox.com/wiki/Proxmox_VE_Kernel

but what is your question?
Click to expand...

Hi, first of all thanks for your support. Proxmox is great piece of work. Thanks for that to.

Regarding TiagoRF, he means that the change for 2.6.18 in 1.5 and not 2.6.24 like in 1.4 means stability, for instance cpu control works in 2.6.18 and in 2.6.24 dont work.

Like he said above 2.6.32 is a branch new kernel, i use it myself on Arch...but like other new stuff comes with bugs that need to be solved.

We have in production proxmox 1.4 2.6.24 and now we are very happy with it (we dont need cpu control), we commited the error to push to much memory in to the VMS and forgot that the host have needs as well.

We will wait until proxmox use 2.6.32 as the default kernel on installs, and maybe when it supports openvz if possible.

Regards
Sniff
 
You must log in or register to reply here.
Top Bottom