[Linux-cluster] CS5 / something weird during tests

[Alain Moulle]

Hi

On my two-nodes cluster with qdiskd :
when testing CS5 via a ifdown eth0 on node2(where is the heart-beat)
I have a strange behavior : the node2 is rebooted and service
is failovered by node1, fine. But after the reboot of node2 and
re-launch of CS 4 daemons, I can't see via clustat any information
from rgmanager, and cant't stop it if I try service rgmanager stop ...
It seems to be stalled in a strange state ...
An idea ?
Thanks
Regards
Alain Moullé

On node2 after reboot :
# clustat
msg_open: No such file or directory
Member Status: Quorate

  Member Name                        ID   Status
  ------ ----                        ---- ------
  xena140                               1 Online
  xena141                               2 Online, Local
  /dev/sdc                              0 Online, Quorum Disk

[root at xena141 log]# service rgmanager status
clurgmgrd (pid 7805 7804) is running...

whereas on the other node I still have service information :

[root at xena140 log]# clustat
Member Status: Quorate

  Member Name                        ID   Status
  ------ ----                        ---- ------
  xena140                               1 Online, Local, rgmanager
  xena141                               2 Offline
  /dev/sdb                              0 Online, Quorum Disk

  Service Name         Owner (Last)                   State
  ------- ----         ----- ------                   -----
  service:test_xena140 xena140                        started
  service:test_xena141 xena140                        started