Regarding lockLUN

[Minoru Asano]

Hello,

I would like to know regarding lockLUN

[ENVIRONMENT]
RHEL AS 4 up4
Serviceguard 11.16.10

[QUESTION]
There are two cluster member as Serviceguard for Linux.
When a package failover from host-A to host-B, is there possibility to appear like following
by using cmviewcl command ?

Host-A -----> "unknown"
Host-B -----> "UP"

Thank you for advice.
Best Regards.
/Minoru.Asano

[Matti_Kurkela · Accepted solution]

Yes, it's possible.

It means Host-A does not respond at all, so Host-B does not know if Host-A is crashed or just lost all the ways to communicate with Host-B.

You'll have to inspect Host-A to see what's going on.

If it's running, check "uptime": it should have rebooted itself right after losing network connections to Host-B. Obviously Host-B managed to access the lockLUN first, and Host-A rebooted to avoid the split-brain situation. As cmviewcl (on host-B I think) is reporting host-A as unknown, there may be a network problem preventing host-B from getting information about host-A.

After the network problem is fixed, cmviewcl on host-B should be able to say with confidence that host-A is "down". To make it re-join the cluster, run "cmrunnode host-A"
or reboot host-A.

If host-A is not running, find out why and fix it (Overheating? Loss of power? Hardware failure?). If there are no network problems and /usr/local/cmcluster/cmcluster.rc has the default value, it should automatically re-join the cluster when it reboots. If not, "cmrunnode host-A" is the command to make it re-join the cluster manually.

MK

[Steven E. Protter]

Shalom,

If you wish, explain the subject of the post versus the question.

Host A is not accessible. Common cause is inconsistent /etc/hosts networking. If you are not using that (see /etc/nsswitch.conf) then it could be DNS.

Start with ping and make sure the nodes recognize each other.

SEP