Article ID: 127068, created on Oct 2, 2015, last review on Oct 2, 2015

  • Applies to:
  • Virtuozzo 6.0

Symptoms

prlsrvctl net operations get stuck

[root@vz6 ~]# prlsrvctl net info Bridged --timeout 100
Failed to retrieve the list of Virtual Networks: PrlJob_Wait: Operation timeout. The operation could not be completed due to a timeout.
Virtual Network Bridged does not exist.

Or

[root@vz6 ~]# prlsrvctl net list
^ะก

The process stack is:

[root@vz6 ~]# ps auxfww | grep prlsrvct
root      562372  0.0  0.0 509372 12368 pts/4    Sl   12:36   0:00 prlsrvctl net list

[root@vz6 ~]# cat /proc/562372/stack
[<ffffffff810c9cca>] futex_wait_queue_me+0xba/0xf0
[<ffffffff810cae60>] futex_wait+0x1c0/0x310
[<ffffffff810cc7c1>] do_futex+0x151/0xbb0
[<ffffffff810cd29b>] sys_futex+0x7b/0x170
[<ffffffff8100b122>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff

parallels.log is flooded with messages

10-02 09:06:08.383 F /IOCommunication:13672:23060/ IO server ctx [write thr] (sender 2): Error: write thread is not started!
10-02 09:06:08.383 F /IOCommunication:13672:23060/ IO server ctx [write thr] (sender 2): Error: write thread is not started!
10-02 09:06:08.383 F /IOCommunication:13672:23060/ IO server ctx [write thr] (sender 2): Error: write thread is not started!
10-02 09:06:08.383 F /IOCommunication:13672:23060/ IO server ctx [write thr] (sender 2): Error: write thread is not started!

The issue manifests itself on heavy loaded nodes with a lot of containers and virtual machines.

Cause

This behavior is recognized as product issue PSBM-19921

Workaround

Virtual network management is possible via vznetcfg command.

To release vznet lock, prlsrvctl process should be killed, for example:

[root@vz6 ~]# ps aux | grep prlsrvctl
root     1021766  0.0  0.0 510164 13812 pts/1    Tl   09:23   0:00          \_ prlsrvctl net list
[root@vz6 ~]# kill -9 1021766
[root@vz6 ~]# rm /tmp/vznetcfg.lck
rm: remove regular empty file `/tmp/vznetcfg.lck'? y
[root@vz6 ~]#

Since the issue is triggered by heavy load on the node, service parallels-server restart (results in restarting of all vms) should restore prlsrvctl functionality.

Search Words

IO server ctx

parallels.log

prlsrvctl net list

c62e8726973f80975db0531f1ed5c6a2 2897d76d56d2010f4e3a28f864d69223 0dd5b9380c7d4884d77587f3eb0fa8ef

Email subscription for changes to this article
Save as PDF