OpenVZ Forum


Home » General » Support » NFS in VE == hang?
NFS in VE == hang? [message #33262] Thu, 02 October 2008 17:23 Go to next message
fatbrother
Messages: 12
Registered: September 2008
Location: Novosibirsk, Russia
Junior Member
I'm trying to use NFS in the VE.

I have 2.6.18-12-fza-amd64 #1 SMP Sun May 18 09:57:08 UTC 2008 x86_64 kernel on Debian Etch and vzctl version 3.0.22-1dso1.

Inside of the VE I have CentOS 5.2 environment made from contributed template http://download.openvz.org/template/precreated/contrib/cento s-5-x86_64-default.tar.gz with latest updates installed by yum update.

I used vzctl set VE --feature nfs:on --save and got NFS in the list of available filesystems in VE.

NFS server is Linux 2.4 with export options (rw,root_squash). It has lockd running. (May be old NFS server is part of the issue? But Solaris 10 and 11 clients work fine with this server under rather serious load, including clients in Solaris zones)

I tried to mount /home directory by running
VE# mount -t nfs server:/home /home
It mounts and I can do ls -l /home and change directories in /home.
Then I'm trying to log as NIS user with ~ directory in /home. It also seems to work, but as I log out, session hangs. I could see sshd and bash processes running in VE, I was able to kill them, but was unable to umount /home (it said device busy).

In HN /var/log/messages there was a line
[time] [HN hostname] kernel: nfs: server server not responding, still trying

nfsstat on HN shows 106 retrans, but number of retrans does not seem to grow.

First time I "solved" it by restarting VM.

Then I remounted share with
VE# mount -t nfs server:/home /home -o rw,vers=3,intr

It hang again, and this time I could not kill hung processes and could not stop VM. Now I cannot vzctl stop VE nor vzctl enter VE, and I probably will need to restart HN (currently there are no other VE on it).

Anybody has any idea?

[Updated on: Thu, 02 October 2008 17:36]

Report message to a moderator

Re: NFS in VE == hang? [message #33266 is a reply to message #33262] Thu, 02 October 2008 18:11 Go to previous messageGo to next message
maratrus is currently offline  maratrus
Messages: 1495
Registered: August 2007
Location: Moscow
Senior Member
First of all, please try the latest kernel.
As you're using Debian I could only suggest you to download the latest kernel from the git repository:

http://git.openvz.org/ (linux-2.6.18-openvz project).
Re: NFS in VE == hang? [message #33267 is a reply to message #33266] Thu, 02 October 2008 18:51 Go to previous messageGo to next message
piavlo is currently offline  piavlo
Messages: 159
Registered: January 2007
Senior Member
You need to run portmap in HN too Smile , Strange but its' a known old unfixed issue.
Re: NFS in VE == hang? [message #33268 is a reply to message #33267] Thu, 02 October 2008 19:19 Go to previous messageGo to next message
fatbrother
Messages: 12
Registered: September 2008
Location: Novosibirsk, Russia
Junior Member
Portmap was running on HN.
Probably resolved [message #33272 is a reply to message #33266] Fri, 03 October 2008 01:53 Go to previous message
fatbrother
Messages: 12
Registered: September 2008
Location: Novosibirsk, Russia
Junior Member
It was not exactly a Virtuozzo issue. It was NFS/network issue. Embarassed

Router between HN and NFS server cuts off fragmented packets, so I cannot send UDP packets > MTU.
I've set rsize and wsize mount options to 1024 and it seems to work fine.
Previous Topic: vps.mount not mounting
Next Topic: VE in separate subnet (other than HN)
Goto Forum:
  


Current Time: Fri Apr 26 16:17:36 GMT 2024

Total time taken to generate the page: 0.01737 seconds