topic Re: R80.20 - More ClusterXL State Information in General Topics

R80.20 - More ClusterXL State Information

HeikoAnkenbrand — Wed, 20 Mar 2019 20:07:56 GMT

With "cphaprob stat" you can now see more information about the cluster state.

Event Code
State change
Reason for state change
Event time

Now you can see the reason for the cluster state change:

Nice!

Regards,

Heiko

Re: R80.20 - More ClusterXL State Information

JozkoMrkvicka — Fri, 28 Sep 2018 11:58:12 GMT

Each important cluster event that affects the Cluster Members has a unique code that appears in the /var/log/messages file and dmesg.

Each cluster event message starts with the prefix CLUS-XXXXXX.

Event Codes are described here:

ClusterXL Error Messages

More info here:

Monitoring Cluster State

Heiko Ankenbrand‌ Can you try also "show cluster state" command ? The output should be exactly the same as "cphaprob stat".

Manual failover can be in R80.20 performed also via clish command "set cluster member admin up/down" according Initiating Manual Cluster Failover

Re: R80.20 - More ClusterXL State Information

HeikoAnkenbrand — Fri, 28 Sep 2018 13:26:15 GMT

Jozko thx, for the ClusterXL Error Messages link.

I also find the new status information very helpful:

A problem was detected, but the cluster member still forwards packets, because it is the only member in the cluster, or because there are no other Active members in the cluster. In any other situation, the state of the member is Down.

ACTIVE(!)
ACTIVE(!F) - Cluster member is in the freeze state.
ACTIVE(!P) - This is the Pivot cluster member in Load Sharing Unicast mode.
ACTIVE(!FP) - This is the Pivot cluster member in Load Sharing Unicast mode and it is in the freeze state.

and

LOST - The peer cluster member lost connectivity to this local cluster member (for example, while the peer cluster member is rebooted).

Regards,

Heiko

Re: R80.20 - More ClusterXL State Information

Alex_Sazonov — Wed, 03 Oct 2018 23:02:57 GMT

It's good to know that (see CXL admin guide😞

ClusterXL commands were integrated in clish
"Cluster failover count" is synced between cluster members and survive reboot (if only one node was restarted)
There is useful history view with ">show cluster failover or #cphaprob show_failover"
ClusterXL failover info is available in CPView and CPView history -> Advanced -> ClusterXL. This is good for troubleshooting of past events when both members were rebooted
As Jozko Mrkvicka‌ mentioned you can see CXL CLUS-XXXXX events in /var/log/messages but you can also improve functionality with ">set cluster member idmode name or #cphaconf mem_id_mode name" to show member's name instead of cluster ID. This configuration affects these local logs:
- /var/log/messages
- dmesg
- $FWDIR/log/fwd.elg
In /var/log/messages you can see both local and remote CXL member state changes or policy freeze events: CLUS-1XXXXX are local events and CLUS-2XXXXX are events from remote member.