would the package be restarted when the cluster is being reformed?

[Hanry Zhou]

I have the situation of the cluster being reformed when a node in 2-nodes SG had a moment of slow response.

Now, my question is, how come the package was not restarted? I thought whenever the cluster is reformed, the package based on the cluster will be also restarted. Isn't ?

[Kent Ostby]

Hanry --

If the cluster doesn't completely go down, then the package will continue to run where it was unless that node leaves the cluster.

If the package was already down, it will not restart simply because of a cluster reformation.

If you are looking at syslog.log information, you can post it here and I can take a look at the messages that we're seeing.

[Devender Khatana]

Hi,

Only those packages which were running on nodes which experienced problem and caused reformation will be restarted on adoptive nodes. Even on those nodes also in case of local lan switch over it will continue running without restart.

HTH,
Devender

[Geoff Wild]

and then, only if auto_run is enabled for the packages....


Rgds...Geoff

[Mel Burslan · Accepted solution]

Question is:

is the package going down when this cluster reformation happens ? If yes, your cluster configuration (networking wise) is not a good one and has to be reworked one way or another as this is not an acceptable behavior, especially you are running clustered environment for high availability not just for kicks.

Secondly, it sounds like you are experiencing this situation frequently. If so, your auto_run may be set to true but when the package fails over to the second node due to a cluster malfunction of any kind, any reason, it is set to false to prevent unnecessary jumping back and forth. So, when the package fails over, and you are sure the other node is fine, you need to enable the package for running on the now alternate node.

Example
CLUSTER STATUS
MyCluster up

NODE STATUS STATE
node1 up running
node2 up running

PACKAGE STATUS STATE AUTO_RUN NODE
pkg1 up running disabled node2
pkg2 up running disabled node2

now as you see both pkg1 and pkg2 are disable to fail over to the node1 if there is a failure. If you are sure node1 is in good shape to run them, you need to enable them using the following commands:

# cmmodpkg -e -n node1 pkg1
# cmmodpkg -e -n node1 pkg2

at this time the disabled should turn to enabled when you run a cmviewcl command.

[melvyn burnard]

The simple answer is no, if the cluster does a reform, provided the node a package was runing on remains in the cluster it will remain there.