Hello

I answer to myself partially, it seems that the problem occurs when the VMs have timeout to access to their hardware,in
my case the VM are on an NFS shared storage. I know that sometime the NFS server has timeout and after come back.

In the previous version of libvirt,qemu, system centos 6 the result was the VM fell with a file system in read only.

I tried to repeat  the error in switching on/off the NFS on the servers . I got some errors messages like

INFO: task kjournald:360 blocked for more than 120 seconds.
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message=..
....
[<fffffffff8000a08bd>] wake_bit_funtion +0w0/0x23
......
   [<fffffffff8800a0ead>] :jbd:journal_get_write_access+0x22/0x33
;;;;
   [<fffffffff800013ccd>] :ext3:ext3_dirty_inode+0x63/0x7b

If I try action on VM ( suspend shutdown , destroy )
the process becomes defunct and I can't do anything else

qemu      8142     1 17 10:35 ?        00:48:51 [qemu-system-x86] <defunct>
root      8151     2  0 10:35 ?        00:00:40 [vhost-8142]w
root      8153     2  0 10:35 ?        00:00:00 [kvm-pit/8142]

The major problem isn't that we see an defunct process with the ps command. It 's that we see
a domain with libvirt in an anormal state and we can't stop it, restart it =.....

So my question ;
Is there a way to hide, to don't list process in this state across the libvirt, to have the possibility
to reuse the VM without rebooting the hypervisor.
Can libvirt just list active process ?
Or an elegant to remove pending process ?


Is the choice for the disk performance can have an inpact to qemu ( none, writethrough, writeback ...)
Actually I 'm using Fedora 23 , libvirt 1.3.3 I will try with an older version in centos 7.2 to see what happen and
if I got the same errors.

Thanks




Michel Villeneuve <Michel.Villeneuve@univ-brest.fr> a écrit :

Hello

since I changed my hypervisor from centos 6.3 to Fedora-23
I had many problems with differents VMs.

Very often once by day ( I have about 150 VM ),
some VMs crash or  freeze indifferently and I got
messages like this on console.

   [<fffffffff8000a08bd>] wake_bit_funtion +0w0/0x23
.....
   [<fffffffff8800a0ead>] :jbd:journal_get_write_access+0x22/0x33
;;;;
   [<fffffffff800013ccd>] :ext3:ext3_dirty_inode+0x63/0x7b

And so the VM is crashed and can't be accessed but often the ping command can respond

Before my migration I never meet these problems, It 's strictly the same VMs between the  6.3 and
fedora23 release. I just changed the parameter :

  <type arch='x86_64' machine='rhel-6.0.0'>hvm</type>
to
  <type arch='x86_64' machine='pc-i440fx-2.4'>hvm</type>
and do a virsh define

I tried some other version of the parameters without success.

and I also added a lockd manager in fedora23.

Before and I used libvirt.0.9.5 or 1.0.2 on centos 6.03 without lockd

A major problem with these crashs is that  the VMs couldn't be destroyed
by the virsh command, the qemu process is notified as defunct by the ps command
If I try a virsh destroy
I get in log file

2016-04-20 20:32:47.318+0000: 5541: info : virEventPollRunOnce:641 : EVENT_POLL_RUN: nhandles=11 timeout=-1
2016-04-20 20:32:55.028+0000: 5567: debug : virProcessKillPainfully:368 : Timed out waiting after SIGTERM to process 8720, sending SIGKILL
2016-04-20 20:33:00.032+0000: 5567: error : virProcessKillPainfully:398 : Failed to terminate process 8720 with SIGKILL: Périphérique ou ressource occupé

or on console

Failed to terminate process xxx with SIGTERM: Device or resource busy
and the VM is still in the list in a "Stopping" state

Result of ps ps on the the qemu process attached to the VM

qemu      8720     1  0 avril20 ?      00:07:16 [qemu-system-x86] <defunct>
root      8733     2  0 avril20 ?      00:00:01 [vhost-8720]
root      8735     2  0 avril20 ?      00:00:00 [kvm-pit/8720]

libvirtd seems to be in an anormal state. If I restart the libvirtd
the virsh command just hang and never remove the VM from the list.

The only seems to reboot the hypervisor but all the VMs in production too=..

Is there a way to remove the process qemu in defunct state without reboot the hypervisor.
Perhaps the probleme come from the VM parameters which have been created on 6.3 Centos and libvirt <1.0 version. Do I need to convert some other parameters ?

I 'am trying to put a new hypervisor in aFailed to terminate process X with SIGKILL: Device or resource busy version level less than fedora23 perhaps a Centos 7.2 to
see what 's happen and if there is a problem like mine.

Thanks

PS:
I put the log_level to 1
----------------------information in logfile
[root@kvmserver6 ~]# ls -al /var/lib/libvirt/qemu/domain-1-TEST-VM-A
total 8
drwxr-x---   2 qemu qemu 4096 20 avril 22:01 .
drwxr-x--x. 18 qemu qemu 4096 20 avril 22:01 ..
srwxrwxr-x   1 qemu qemu    0 20 avril 22:01 monitor.sock
/var/lib/libvirt/qemu/channel/target/domain-1-TEST-VM-A/


[root@kvmserver6 ~]# cat /var/log/libvirt/qemu/TEST-VM-A.log
2016-04-20 20:01:36.714+0000: starting up libvirt version: 1.3.3, package: 1.fc23 (Unknown, 2016-04-06-15:17:39, thinkpad2), qemu version: 2.4.1 (qemu-2.4.1-8.fc23), hostname: kvmserver6.univ-brest.fr
LC_ALL=C PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin QEMU_AUDIO_DRV=none /usr/bin/qemu-kvm -name TEST-VM-A,debug-threads=on -S -machine pc-i440fx-2.4,accel=kvm,usb=off -m 1024 -realtime mlock=off -smp 1,sockets=1,cores=1,threads=1 -uuid 1e4c27e4-123e-719a-9fdf-f783d34cbb40 -no-user-config -nodefaults -chardev socket,id=charmonitor,path=/var/lib/libvirt/qemu/domain-1-TEST-VM-A/monitor.sock,server,nowait -mon chardev=charmonitor,id=monitor,mode=control -rtc base=utc -no-shutdown -boot strict=on -device piix3-usb-uhci,id=usb,bus=pci.0,addr=0x1.0x2 -drive file=/var/lib/libvirt/images/POOL_PROD4/TEST-VM-A.img,format=raw,if=none,id=drive-virtio-disk0 -device virtio-blk-pci,scsi=off,bus=pci.0,addr=0x5,drive=drive-virtio-disk0,id=virtio-disk0,bootindex=1 -drive if=none,id=drive-ide0-1-0,readonly=on -device ide-cd,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0 -netdev tap,fd=26,id=hostnet0,vhost=on,vhostfd=28 -device virtio-net-pci,netdev=hostnet0,id=net0,mac=52:54:00:77:11:11,bus=pci.0,addr=0x3 -chardev pty,id=charserial0 -device isa-serial,chardev=charserial0,id=serial0 -vnc 0.0.0.0:0,password -k fr -device cirrus-vga,id=video0,bus=pci.0,addr=0x2 -device ES1370,id=sound0,bus=pci.0,addr=0x4 -device virtio-balloon-pci,id=balloon0,bus=pci.0,addr=0x6 -msg timestamp=on
char device redirected to /dev/pts/1 (label charserial0)
qemu: terminating on signal 15 from pid 5541

Michel Villeneuve
Tel 02 98 01 71 61


Michel Villeneuve
Tel 02 98 01 71 61