09-06-2016 07:11 AM
I have a HA cluster with 3 applications (A1, A2, A3) and 4 nodes (N1-4).
Applications A1 and A2 I need to start in failover on nodes N1 or N2, they have some common components (files, ip and mount point). Application A3 I need to start in failover on nodes N3 or N4.
One issue is an order of the start. I need to start A1, then A3, then A2. Otherwise applications won't start correctly.
After initial start A3 can failover/restart at another server without any affect on A1 and A2.
Same for A1 and A2 - they can failover from N1 to N2 without any affect on A3. But A1 and A2 need to work on the same system.
I created service groups SG1 and SG2 are on nodes N1 and N2 for applications A1 and A2, SG3 is on the N3 and N4 for A3. I've set dependencies for service groups, so applications can start in correct order:
SG1 is child for SG2, online local firm dependency
SG1 is child for SG3, online remote soft dependency
SG3 is child for SG2, online remote soft dependency
And now they can start in correct order. But I have some problems when I'm trying to switch groups SG1 or SG2 to another node or make failover to another node. I'm receiving an error:
hagrp -switch SG2 -to rsb-ascmos0smm1 -localclus
VCS WARNING V-16-1-10206 Group dependencies are not met for group SG2 on system rsb-ascmos0smm1
Or if I kill an application and check logs during failover:
2016/09/06 12:19:41 VCS INFO V-16-1-10307 Resource SG2 (Owner: Unspecified, Group: SG2 ) is offline on rsb-ascmos0smm2 (Not initiated by VCS)
2016/09/06 12:19:41 VCS ERROR V-16-1-10205 Group SG2 is faulted on system rsb-ascmos0smm2
2016/09/06 12:19:41 VCS NOTICE V-16-1-10446 Group SG2 is offline on system rsb-ascmos0smm2
2016/09/06 12:19:41 VCS INFO V-16-1-10493 Evaluating rsb-ascmos0smm1 as potential target node for group SG2
2016/09/06 12:19:41 VCS INFO V-16-1-10163 Group dependency is not met if group SG2 goes online on system rsb-ascmos0smm1
2016/09/06 12:19:41 VCS INFO V-16-1-10493 Evaluating rsb-ascmos0smm2 as potential target node for group SG2
2016/09/06 12:19:41 VCS INFO V-16-1-50010 Group SG2 is online or faulted on system rsb-ascmos0smm2
2016/09/06 12:19:41 VCS INFO V-16-6-15015 (rsb-ascmos0smm2) hatrigger:/opt/VRTSvcs/bin/triggers/resfault is not a trigger scripts directory or can not be executed
How can I configure service groups or applications dependency in correct way?
OS: RHEL 6.8, 2.6.32-642.3.1.el6.x86_64
09-19-2016 10:26 PM
Super late reply but maybe you still need help. What you're seeing is the default behaviour but i think its addressed in Infoscale. If you have a 2 group dependency you could never just right click and switch to the other node, you had to offline the parent, then switch the child, and then online the parent. I might be wrong but it could be resolved in infoscale (I recall doing a switch of a SQL SG and its associated RVG SG recently in one step).
If its not, you can use Operations Manager's VBS feature to control your service group failover and fault dependency behaviour.