I am facing problems with cache in cluster environment. it is getting corrupted because of which , we arae facing some availability of system is not reliable and to resove this we offten restart the system.
In current environment, we have 2 Slave nodes and 1 master and in each slave node there are 3 MS configured and on master 2 BGMS are configured.
Is there any way to avoid cache corruption problem or improve the synchronisation process accross nodes.
The number one cause of issues with Windchill's server-side caches is misconfiguration.
Did you follow the cluster configuration steps? [Note these are much simpler in 10.2 than in previous releases -- and provide automatic election and failover of the master server manager as well.]
Are there by any chance log messages noting cache communictaion issues that you're not mentioning?
presently we have 10.2 M010 and it is clustered with 2 slave nodes, custering is done by following previous procedure as it used to be done in 9 and 10.0, i.e. manualy ..
another information is that, When we had 2 MS configured on each slave node? things were working fine, we did not had this issue but due to some circumstances we have incresed to 3 MS on each node.
cluster configuraton seems to be fine. but it is as per 9 version ..
could you please kindly provide clustering procedure to be followed in for 10.2 version.
Clustering configuration changed in important ways in 10.2. It is significantly simpler, but also significantly different. If you simply configured it as you would have in previous versions, then I suspect your configuration is wrong.
See the 10.2 documentation. Notably, there's no special configuration regarding a master any more, nor any notion of a pre-elected master (though you can limit the nodes that can play the role of master if you really want to).
One other note:
Offhand, the only way I can see moving from 2 to 3 MS per node causing the issue is if you over-commited server memory in the process such that you are swapping Java process memory from physical to virtual memory. Java processes behave quite badly in this situation. If something like this is done, then you should see warnings about very high percentages of time spent in garbage collection (GC) and bad performance overall. Cache communication issues may then result due to horribly long response delays, causing timeouts.