Service group won't stay put on reboot

[Jeremy_Loukin1]

We have a pair of servers clustered. Solaris 8 VCS 4.0MP1.

Here is the situation. App owner wants the servers to be manual fail over only. The trick is each server gets rebooted monthly automatically via cron ( don't ask )

I have removed the -evacuate from /etc/rc0.d/K10vcs and the other start/stop scripts /etc/.

There are no agents on this server. Just storage and ip. It probes everything on the secondary node then starts the service group on the primary box.

2007/10/02 22:36:46 VCS NOTICE V-16-1-10438 Group clustmon has been probed on sy
stem nodeB
2007/10/02 22:36:46 VCS INFO V-16-1-50007 Initiating auto-start online of group
clustmon
2007/10/02 22:36:46 VCS INFO V-16-1-10493 Evaluating nodeA as potential ta
rget node for group clustmon
2007/10/02 22:36:46 VCS NOTICE V-16-1-10233 Clearing Restart attribute for group
 clustmon on all nodes
2007/10/02 22:36:46 VCS NOTICE V-16-1-10301 Initiating Online of Resource clustm
on_public_ip (Owner: unknown, Group: clustmon) on System nodeA

2007/10/02 22:36:46 VCS INFO V-16-1-10304 Resource node_public_ip (Owner:
unknown, Group: node_grp) is offline on nodeB (First probe)
2007/10/02 22:36:46 VCS NOTICE V-16-1-10438 Group node_grp has been probed
 on system mmfdbq0003b

then it fires everything up on nodeA.

[Hywel_Mallett]

Assuming it's the same as VCS on Windows, there is an attribute, AutoFailOver, which can be set for the service group. By default it's set to 1 (which is why it fails over). Add the attribute (if it doesn't already exist) and set it to 0.

[Roger_Zimmerman]

Hi, Jeremy,

I learned that you want to have a startup of your applications distributed via all available nodes in starting time of the cluster, but they all start only on one node every time you reboot your cluster.

Ok, then you have to modify the SystemList attribute of all your service groups. E.g. you have two nodes and two apps, then the SystemList attribute should read like this:

group SG1 (

SystemList = { node1 = 0, node2 = 1 }

...

)

group SG2 (

SystemList = { node2 = 0, node1 = 1 }

...

)

Then SG1 and SG2 start on different nodes everytimes your cluster comes up and all needed nodes are online.

[Gene_Henriksen]

First a warning: Modifying the scripts can lead to serious problems if you call support and do not let them know this. In one instance, a user modified the scripts for the agents. As a result there was an agreement between his employer and VERITAS that for them to have support, he could not administer systems with VERITAS products.

Second, if you want manual failover, turn off AutoFailover as mentioned before.

I don't understand "there are no agents on this server". If VCS ihas resources defined then agents are running. If the service group can failover from one server to another, then the agents are running on both and on the server where it is offline, then it  runs the monitor at the OfflineMonitorInterval (300 seconds) to make sure of the status.

To make service groups start on different systems, set the node name in AutoStartList to the node name you want the SG to start on.

So I suppose the App owner is there 24 hours per day in case of failure so they can order it to be started on the other node.

At least that is the way it looks to me after teaching VCS for over 7 years.

[Roger_Zimmerman]

Ok, then it should read like

group SG1 (

SystemList = { node1 = 0, node2 = 1 }

AutoStartList = ( node1, node2 )

...

)

group SG2 (

SystemList = { node2 = 0, node1 = 1 }

AutoStartList = ( node1, node2 )

...

)

Normaly there are all the nodes in the AutoStart list where the service group can start automaticly. Then the service group also comes up on the second node if the first node in SystemList is not online. But the primary attribute for the priority of starting for one service group keeps the sequence in SystemList :)

Ok, this is not that important in small clusters then in clusters with a bigger seeding. But who knows if this cluster is growing some day? :-)=)