Veritas Cluster Software
VCS resources are on-off, on-only and persistent. In case of failover, resource are not switched individually but the entire service group. Critical attribute of a resource defines whether service group fails over when the resource faults. When a resource faults, VCS takes action to clean up the resource.
VCS service groups are failover, parallel and hybrid. Clusterservice group is a special service group for resources required by VCS. Freezing a service group prevents VCS from taking action when a service group or system faults.
Agents are VCS processes that manage resources of pre-defined types according to commands received from VCS engine (HAD). Agents can be created in C++, Perl and shell scripts. Agents are bundles, enterprise or custom. hashadow monitors and restarts HAD, if required. HAD runs at high priority of 0 and nice level of real time. Agent has entrypoints and agent framework. HAD core dumps are written to /var/VRTSvcs/diag/had directory.
Cluster communications use LLT and GAB. LLT is a high performance and low latency replacement of IP stack. GAB runs on top of LLT and manage cluster membership and communications.
By default, VCS monitors resource every 60 seconds. Default no of threads per agent of a resource type is 10. MonitorInterval, OfflineMonitorInterval, MonitorTimoue, OnlineTimeout, ToleranceLimit and OfflineTimeout.
Cluster states at boot time are LOCAL_BUILD, RUNNING,
VCS service groups are failover, parallel and hybrid. Clusterservice group is a special service group for resources required by VCS. Freezing a service group prevents VCS from taking action when a service group or system faults.
Agents are VCS processes that manage resources of pre-defined types according to commands received from VCS engine (HAD). Agents can be created in C++, Perl and shell scripts. Agents are bundles, enterprise or custom. hashadow monitors and restarts HAD, if required. HAD runs at high priority of 0 and nice level of real time. Agent has entrypoints and agent framework. HAD core dumps are written to /var/VRTSvcs/diag/had directory.
Cluster communications use LLT and GAB. LLT is a high performance and low latency replacement of IP stack. GAB runs on top of LLT and manage cluster membership and communications.
By default, VCS monitors resource every 60 seconds. Default no of threads per agent of a resource type is 10. MonitorInterval, OfflineMonitorInterval, MonitorTimoue, OnlineTimeout, ToleranceLimit and OfflineTimeout.
Cluster states at boot time are LOCAL_BUILD, RUNNING,
posted by Usman Wahid at
5:01 PM
0 Comments:
Post a Comment
<< Home