Edit: Hopefully you have already put on these patches by now
Originally posted April 10, 2018 on AIXchange
I received this information from Chris Gibson a few weeks ago. If you use PowerHA, I recommend checking to see if you’re on affected levels of AIX:
High Impact / Highly Pervasive APAR
IJ02843 – PowerHA node halt during ip changes
“USERS AFFECTED:
- Systems running PowerHA System Mirror on the
- AIX 7100-05 Technology Level or
- AIX 7200-02 Technology Level with
- rsct.basic.rte at 3.2.3.0.
************************************************************** - PROBLEM DESCRIPTION:
- An improvement in obtaining adapter state information from
- AHAFS event responses introduced some errors in handling
- * internal tracking of monitored IP addresses.
* - This can result in a core dump of the hagsd process any time
- an IP change occurs at the OS layer. This means the failure
- cannot happen while a cluster is running stable with no
- changes occurring, but it is a risk during startup, shutdown,
- or a failover scenario, and cannot be predicted beyond that.
Problem summary
A flaw in handling of monitored IP changes during some
adapter state improvements in RSCT 3.2.3.0 has led to the
risk of a hagsd core dump in a couple code paths.
Problem conclusion
Transition of IP lists during a monitoring change has been
corrected.”
Here’s some additional information:
In a PowerHA cluster, if an IP address is changed on one of the AIX nodes, the node may reboot unexpectedly due to a core dump of the hagsd process. This can happen when a Service IP is configured during normal PowerHA startup/shutdown/failover, or other operations resulting in an IP change.
Affected AIX Levels and Recommended Fixes
Minimum Affected Level Maximum Affected Level Fixing Level Interim Fix
7100-05-00
rsct.basic.rte 3.2.3.0 7100-05-02
rsct.basic.rte 3.2.3.0 7100-05-03
IJ02843 iFix
7200-02-00
rsct.basic.rte 3.2.3.0 7200-02-02
rsct.basic.rte 3.2.3.0 7200-02-03
IJ02843 iFix
Note: Applying the ifix requires PowerHA to be stopped on the node prior to applying the fix.