1748150 Members
3520 Online
108758 Solutions
New Discussion

package disk fails

 
himacs
Super Advisor

package disk fails

Hi Admins,

 

Cluster : 11.20 

 

 I have active-passive 2 node cluster.My query is if any of the LUN is not accessible from active node, then what will be the behaviour of cluster and package.Will it try to fail over.

 

And if LUN is entirely failed and not accessible both the nodes, what will be the cluster status/pckage status.

 

 

I find below comment in HP docs.

Specifying the generic_resource_evaluation_type is optional. If not specified,

the default value is during_package_start.

 

i find during_package_start and generic_resource is commented in package file.

 

Regards

himacs

2 REPLIES 2
Matti_Kurkela
Honored Contributor

Re: package disk fails

If your package configuration includes monitoring for the accessibility of the LUN, or the packaged application is monitored by the package configuration and the LUN failure causes the application to fail, then the package will failover for sure.

 

If there is no explicit monitoring for LUN accessibility and your application will not die but will instead hang if the LUN is not accessible (or your package does not include a service that detects application failure), then the package may not failover automatically.

 

When the failover is triggered, the application shutdown scripts (if any) must complete successfully even if the application is hanging because of the LUN being inaccessible. If the application shutdown script fails, the package halt will fail and the failover cannot complete, and the package will remain down. (You can choose to avoid this issue by enabling the appropriate fail_fast parameter in the package configuration, at the price of having the entire node intentionally crashed by Serviceguard at the beginning of package failover. If you don't want that, you should write your application shutdown scripts with as robust error handling as you can.)

 

If the application shutdown script also hangs, the package will remain in the halting state until the halt script times out (if a timeout has been configured) or the administrator resolves the situation manually.


If the failover gets as far as trying to start the package on another node, and the LUN is entirely failed and not accessible on the failover node too, then the package start-up will fail in the VG activation phase and the package will remain down.

If the cluster nodes can still communicate with each other, the cluster will be up, but the package will be down.

MK
prasanth479
Regular Visitor

Re: package disk fails

generic resource is a monitoring mechanism implemented in Serviceguard 11.20

 

This is a replacement to legacy ems resource monitoring mechanism. Using generic resources we can write custom monitoring script for any resources eg. a LUN.

 

Evaluation type for a generic resource can be

 

1. during_package_start

 

2. before_package_start

 

As name implies, this indicates at what stage a particular resurce status is evaluated. 

 

during_package_start - This resource status is evaluated by the package during the pkg start and you can configure the resource monitor as a service inside a pkg

 

before_package_start- resource status is checked by the package manager before starting a pkg.