HACMP issue

[ogniemi]

2 nodes (one is a HOME NODE for RG1 and the other for RG2)
2 resource groups - in normal operation each ONLINE on its HOME NODE

(for each RG both nodes are configured to be participating in takoever)

in RG1 there is shared VG configured. VG is built on EMC storage. The VG is online on one node only during cluster normal operation.

NodeA: halt -q
NodeB: gets RG1 and varying on the shared VG

Problem/case:
NodeA started with RG1 online (has varied on the shared VG).
NodeB started with RG2 online.

Removing all FC cables from NodeA. It detects that hbeat network over hdisk (EMC) is down, BUT HACMP IS NOT INITIALIZING TAKEOVER!!!!

When run "lsvg -o" on NodeA with unplugged FCs it still shows that shared VG is online!!! No access to EMC storage.

why HACMP didn't do takeover RG1 to NodeB after FCs removed?


thx in advance,
r,m.

 

[RMGBELGIUM]

how long did you remove the fc cables? Because HACMP triggers after X missed heartbeats only.As long as the X isn't reached ,your node will remain online. ( perhaps shortening your interval of heartbeating will help )

rgds,


R.

 

[sbix]

HACMP normally doesn't perform any "POLLING" to see if disks are still available once they have been assigned to a node.
You use disks as heartbeat path, but this doesn't imply a takeover once this heartbeat path fails because there are other paths via normal network communication.
The hearbeat over disk is treated like a communication path and it is not used as a check on the disk resource.
If you need a such polling you can use the application monitor facility