PVE 4 Another migration problem

[alitvak69]

Some times after running migrate command I get this

[FONT=verdana][COLOR=#000000]Executing HA migrate for VM 100 to node virt2n3-la[/COLOR][/FONT]
[SIZE=2][FONT=verdana][COLOR=#000000]unable to open file '/etc/pve/ha/crm_commands.tmp.19096' - No such file or directory[/COLOR][/FONT][/SIZE]
[COLOR=#000000][SIZE=2][FONT=tahoma]TASK ERROR: command 'ha-manager migrate vm:100 virt2n3-la' failed: exit code 2
[/FONT][/SIZE][FONT=tahoma]
[/FONT][/COLOR]

In syslog

[/FONT][/COLOR]Oct 10 01:12:05 virt2n4-la pvedaemon[22397]: <root@pam> starting task UPID:virt2n4-la:00004A97:009DFD7E:5618C855:hamigrate:100:root@pam:
Oct 10 01:12:06 virt2n4-la pvedaemon[19095]: command 'ha-manager migrate vm:100 virt2n3-la' failed: exit code 2

Then I run from command line

[/FONT][/COLOR]ha-manager migrate vm:100 virt2n3-la
unable to open file '/etc/pve/ha/crm_commands.tmp.19823' - No such file or directory

touch /etc/pve/ha/crm_commands.tmp.1
touch: cannot touch ‘/etc/pve/ha/crm_commands.tmp.1’: No such file or directory


touch /etc/pve/ha/crm_commands.1
rm /etc/pve/ha/crm_commands.1

Then migration starts working

I noticed this line in syslog right before the migration

Oct 10 01:17:19 virt2n4-la pve-ha-crm[1206]: status change wait_for_quorum => slave

 

[t.lamprecht]

Whats the output of:
# ha-manager status
# ls -l /etc/pve/ha
# cat /etc/pve/resource.cfg
when that is happening?

It looks like you try to migrate an service which is not HA managed, but then the daemon should not choose an HA migration...
# ha-manager add vm:100
or adding VM 100 via the web interface was done?

 

[alitvak69]

It is hard to say what happen but as of now:

ha-manager status
quorum OK
master virt2n1-la (active, Sat Oct 10 17:26:49 2015)
lrm virt2n1-la (active, Sat Oct 10 17:26:47 2015)
lrm virt2n2-la (active, Sat Oct 10 17:26:50 2015)
lrm virt2n3-la (active, Sat Oct 10 17:26:46 2015)
lrm virt2n4-la (active, Sat Oct 10 17:26:53 2015)
service vm:101 (virt2n2-la, started)

root@virt2n2-la:~# ls -l /etc/pve/ha/
total 2
-rw-r----- 1 root www-data   0 Oct 10 17:26 crm_commands
-rw-r----- 1 root www-data  71 Oct 10 12:58 groups.cfg
-rw-r----- 1 root www-data 307 Oct 10 17:26 manager_status
-rw-r----- 1 root www-data  24 Oct 10 13:21 resources.cfg

root@virt2n2-la:~# cat /etc/pve/ha/resources.cfg 
vm: 101
    state enabled

Whats the output of:
# ha-manager status
# ls -l /etc/pve/ha
# cat /etc/pve/resource.cfg
when that is happening?

It looks like you try to migrate an service which is not HA managed, but then the daemon should not choose an HA migration...
# ha-manager add vm:100
or adding VM 100 via the web interface was done?

 

[t.lamprecht]

ha-manager migrate vm:100 virt2n3-la

You tried to migrate vm:100 while only vm:101 is under HA control, use:

qm migrate 100 <target>

for this VM when you want to migrate it.
In fact you should maybe use "qm migrate" for every machine, if it's an HA managed one the API redirects the call to the ha-manager

I saw the output you got only on an non configured HA (e.g.: no resource configured), with:

quorum OK
master virt2n1-la (active, Sat Oct 10 17:26:49 2015)
lrm virt2n1-la (active, Sat Oct 10 17:26:47 2015)
lrm virt2n2-la (active, Sat Oct 10 17:26:50 2015)
[...]

You should be good.

 

[alitvak69]

You tried to migrate vm:100 while only vm:101 is under HA control, use:

qm migrate 100 <target>

for this VM when you want to migrate it.
In fact you should maybe use "qm migrate" for every machine, if it's an HA managed one the API redirects the call to the ha-manager

I saw the output you got only on an non configured HA (e.g.: no resource configured), with:

quorum OK
master virt2n1-la (active, Sat Oct 10 17:26:49 2015)
lrm virt2n1-la (active, Sat Oct 10 17:26:47 2015)
lrm virt2n2-la (active, Sat Oct 10 17:26:50 2015)
[...]

You should be good.

Sorry but I am not good. I am having trouble with live migration under ha period.

The reason you see 101 now is because I cloned it (complete copy) from 100. Vm 100 is gone. The only one vm there is 101 and it is in ha mode.

This problem shown here is in addition to my earlier post about issues with live migration when ha is on.

Qemu migrates vm but migration leaves vm suspended on the target node. If you look at my other post it contains all details and logs.

At this point I believe there is either a bug in pve 4 ha causing it or I have a very weird problem that trigers race condition.

I read your development list thread labeled something like " ... Feedback " and I believe Gilles have a similar or related set of problems.

Can you please help? I can open a ticket if it speed things up.

Sent from my SM-G900V using Tapatalk

 

[alitvak69]

Sorry but I am not good. I am having trouble with live migration under ha period.

The reason you see 101 now is because I cloned it (complete copy) from 100. Vm 100 is gone. The only one vm there is 101 and it is in ha mode.

This problem shown here is in addition to my earlier post about issues with live migration when ha is on.

Qemu migrates vm but migration leaves vm suspended on the target node. If you look at my other post it contains all details and logs.

At this point I believe there is either a bug in pve 4 ha causing it or I have a very weird problem that trigers race condition.

I read your development list thread labeled something like " ... Feedback " and I believe Gilles have a similar or related set of problems.

Can you please help? I can open a ticket if it speed things up.

Sent from my SM-G900V using Tapatalk

The post I am referring to is

PVE 4 KVM Live migration problem.

Sent from my SM-G900V using Tapatalk