OpenVZ Forum

Home » General » Support » ps stuck (failed to read /proc/pid/cmdline for some process)
ps stuck [message #47099] Fri, 06 July 2012 09:03 Go to next message
Messages: 1
Registered: July 2012
Junior Member

I have critical problem. It is the second time it happens ( in a month )

The problem symptom is that load increases to very high numbers.

What I can see is that ps -ef command is stuck from the openvz server on specific process.

strace ps -ef shows that it is stuck on trying to read /proc/pid/cmdline of some process

Manual "ls" on /proc/pid stuck too ( for the specific process )

I can identify the ve on which the stuck process is but it's not possible to kill it.

My previous experience show only reboot of the openvz server helps.

The openvz server is installed on CentOS 6.2 , openvz version 2.6.32-042stab044.17

Any idea what is the problem ?

Thanks a lot,
Re: ps stuck [message #47328 is a reply to message #47099] Tue, 31 July 2012 06:50 Go to previous message
tjheeta is currently offline  tjheeta
Messages: 3
Registered: July 2012
Junior Member
This is a pretty ugly problem.

What is happening I think that processes are stuck in uninterruptable sleep. In the kernel log, there's probably something of the sort:
INFO: task java:806296 blocked for more than 120 seconds.
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.

We're finding this happening on NFS backed VM's. Adjusting the mount options to "soft,intr" doesn't really help. In fact, nothing really helps.

The VM cannot be shutdown causing the entire hardware node needing to be rebooted.

We're having this problem on all kernel versions. Not sure if this is related to the network card or something else.
Previous Topic: OpenVZ RHEL6 is much more stable now
Next Topic: nfsd kernel options
Goto Forum:

Current Time: Tue Dec 05 04:25:28 GMT 2023

Total time taken to generate the page: 0.02683 seconds