Tuning the Sun Cluster HA for Solaris Containers Stop_timeout property

The Sun Cluster HA for Solaris Containers components consist all of the resource type SUNW.gds(5). As described in Stop_command Property in Sun Cluster Data Services Developer’s Guide for Solaris OS the value for the Stop_timeout should be chosen so that the Stop_command can successfully return within 80% of its value.

Choosing the Stop_timeout value for the Zone Boot Component

The stop method for the zone boot component spends 60% of the value for the Stop_timeout performing a complete "shutdown -y -g0 -i0" within the zone. If that failed, the next 20% of the value for the Stop_timeout will be spent halting the zone performing a "zoneadm —z zonename" halt and perform some additional cleanup steps in order to force the zone into the state installed. Therefore the Stop_timeout value for the zone boot component should be computed so that 60% is enough to successfully shutdown the zone.

Choosing the Stop_timeout value for the Zone Script Component

The stop method for the zone script component calls the command or script configured for the ServiceStopCommand keyword. Therefore the Stop_timeout value for the zone script component should be computed so that 80% is enough for the configured ServiceStopCommand to succeed.

Choosing the Stop_timeout value for the Zone SMF Component

The stop method for the zone SMF component spends 60% of the value for the Stop_timeout using svcadm to disable the configured SMF service in the zone. If that failed, the next 20% of the value for the Stop_timeout will be spent to first send SIGTERM then SIGKILL to the processes associated with this SMF service. Therefore the Stop_timeout value for the zone SMF component should be computed so that 60% is enough to successfully disable the configured SMF service in the zone.