rgmanager and restart host from pve gui

[RobFantini]

Hello
If a host is running rgmanager and a shutdown or restart is made from pve gui, then 'service rgmanager stop' should run before init6 or init 0 . That way the vm's could online migrate to another node. Instead the vm's shutdown then start up on another node.

 

[dietmar]

AFAIK, a simple 'service rgmanager stop' does not trigger any online migration?

 

[bread-baker]

Dietmar,
Sorry for the delay replying, I could not test until tonight.

Stopping rgmanager was mentioned on the wiki.

It does force migration of ha kvms. They stop on the node and then start on the other node:

fbc243 s009 ~ # qm list
      VMID NAME                 STATUS     MEM(MB)    BOOTDISK(GB) PID       
       500 Win7                 running    3072               0.00 1000476   
       510 winxp                running    2048               0.00 11996     
       520 etch-kvm             running    512                0.00 5267      
       530 sid-cam165           running    4200               0.00 667711    
       550 ubuntu-cam5          running    8100               6.00 667756    
      1003 ltsp-term-KVM        stopped    512                0.00 0         
      9998 ubuntu-template      stopped    1000               0.00 0         
      9999 wheezy-template      stopped    1024               2.00 0         
fbc243 s009 ~ # date
Tue Aug 21 20:54:39 EDT 2012
fbc243 s009 ~ # service rgmanager stop
Stopping Cluster Service Manager: [  OK  ]

It would be better if an online migration were done... On testing I thought that occurred when rgmanager was stopped from cli. But I guess I was mistaken.