[SOLVED] Some VM migrations fail, but only between certain nodes

[fknorn]

Hi there,

As intro: I've recently changed a bunch of node names / hostnames.

Things mostly work, except for one particular oddity.

In my 6-node cluster, using Ceph as shared storage, with HA configured, fencing working, etc. I can live-migrate my VMs from anywhere to anywhere else, EXCEPT for one particular node.

Let's say I have nodes A, B, C, and X. Migrations A<=>B, A<=>C, A=>X, B<=>C, B=>X and C=>X work just fine BUT X=>B and X=>C fail, only X=>A works.

In other words, everyone can "send" VMs to X, but once on X, they can only be migrated to, A, but none of the other nodes. All other pairwise migrations (not involving X) work.

Any idea what might cause this? Any suggestions how to better debug this?

Here's the syslog:

[COLOR=#000000][FONT=tahoma]Nov  7 17:07:14 ve44-b1 rgmanager[3467]: Migrating pvevm:10071 to ve23-b1[/FONT][/COLOR]
[COLOR=#000000][FONT=tahoma]Nov  7 17:07:15 ve44-b1 pvevm: <root@pam> starting task UPID:ve44-b1:0000B0A8:021FC6B2:545CEE33:qmigrate:10071:root@pam:[/FONT][/COLOR]
[COLOR=#000000][FONT=tahoma]Nov  7 17:07:15 ve44-b1 task UPID:ve44-b1:0000B0A8:021FC6B2:545CEE33:qmigrate:10071:root@pam:: migration aborted[/FONT][/COLOR]
[COLOR=#000000][FONT=tahoma]Nov  7 17:07:15 ve44-b1 pvevm: <root@pam> end task UPID:ve44-b1:0000B0A8:021FC6B2:545CEE33:qmigrate:10071:root@pam: migration aborted[/FONT][/COLOR]
[COLOR=#000000][FONT=tahoma]Nov  7 17:07:15 ve44-b1 rgmanager[3467]: migrate on pvevm "10071" returned 150 (unspecified)[/FONT][/COLOR]
[COLOR=#000000][FONT=tahoma]Nov  7 17:07:15 ve44-b1 rgmanager[3467]: Migration of pvevm:10071 to ve23-b1 failed; return code 150[/FONT][/COLOR]
[COLOR=#000000][FONT=tahoma]Nov  7 17:07:15 ve44-b1 corosync[3107]:   [CMAN  ] daemon: read 0 bytes from fd 35[/FONT][/COLOR]
[COLOR=#000000][FONT=tahoma]Nov  7 17:07:15 ve44-b1 corosync[3107]:   [CMAN  ] daemon: Freed 0 queued messages[/FONT][/COLOR]
[COLOR=#000000][FONT=tahoma]Nov  7 17:07:15 ve44-b1 pvedaemon[45220]: command 'clusvcadm -M pvevm:10071 -m ve23-b1' failed: exit code 239[/FONT][/COLOR]
[COLOR=#000000][FONT=tahoma]Nov  7 17:07:15 ve44-b1 pvedaemon[3751]: <root@pam> end task UPID:ve44-b1:0000B0A4:021FC679:545CEE32:hamigrate:10071:root@pam: command 'clusvcadm -M pvevm:10071 -m ve23-b1' failed: exit code 239[/FONT][/COLOR]

Thanks for your help !!

 

[fknorn]

Anyone ... ?

 

[mir]

What cpu emulation does the the concerned vm's use and are the hardware on the nodes identical?

 

[fknorn]

Yes, they are.

But I solved the issue. It had to do with SSH authentication I believe.

I manually tried to SSH from every node to every other node, and I noticed that in some cases I got the usual first-time-SSH-question (The authenticity of host ......... can't be established). Upon answering all those with "yes", all migrations now work.