Operating System - HP-UX
1838275 Members
3557 Online
110125 Solutions
New Discussion

Re: SUBNET Monitoring in the PKG configuration

 
Q4you
Regular Advisor

SUBNET Monitoring in the PKG configuration

We recently had a complete n/w outage. The packages which were configured no subnet monitoring stayed up ( have seprate HUB lan for a cluster) but the package which has subnet monitoring enabled, were halted.

Question ->

What are the PROs and CONs for having SUBNET monitoring enabled ?

TIA,

-Q
2 REPLIES 2
melvyn burnard
Honored Contributor

Re: SUBNET Monitoring in the PKG configuration

Well this is what lets the package manager decide to move a package between nodes if a monitored subnet fails.
If you do not monitor a subnet in your package, it will not switch nodes when the subnet fails.
It is briefly described in the comments in the package configuration file.
My house is the bank's, my money the wife's, But my opinions belong to me, not HP!
Stephen Doud
Honored Contributor

Re: SUBNET Monitoring in the PKG configuration

Configuring a SUBNET by way of the package configuration file tells Serviceguard to halt the package on the affected node (and potentially fail the package to the adoptive server) when the SUBNET specified goes down. Te SUBNET is considered DOWN (by Serviceguard) when the primary NIC that supports the subnet and any standby NIC that would normally switch in for it - has been determined to be non-functional.

The PROs/CONs stem from the automation value of Serviceguard and whether a human attendant is available.
Some customers prefer not to have Serviceguard force a package down or failover when a network administrator is available to resolve the issue. The reasoning is the outage time to network repair is less than application shutdown/failover/restart.
Such customers wait for users to squeek about communication failure as a signal to investigate the network condition.

For systems which do not have staffing to manage network problems it may be advisable to have Serviceguard automatically fail the package to an adoptive node rather than wait for someone to complain - particularly where revenue is directly reliant on the package application.