Runtime Object deletion failure for non-existing stream leads to node reboot

[Sangeeta.meena@…]

immutil_saImmOiRtObjectDelete returns SA_AIS_ERR_NOT_EXIST incase the runtime object doesn't exist with IMMND(after immnd restart case)and since immutilWrapperProfile.errorsAreFatal is not reset to 0,IMMSV does abort and the system reboots.

if (lgs_cb->ha_state == SA_AMF_HA_ACTIVE)

{

TRACE("Stream is closed, I am HA active so remove IMM object");
SaNameT objectName;
strcpy((char *) objectName.value, stream->name);
objectName.length = strlen((char *) objectName.value);
(void) immutil_saImmOiRtObjectDelete(lgs_cb->immOiHandle, &objectName);

}

log_stream_delete(s);

[hafe]

Sangeeta:
"It is on single node so the runtime object will not be synced."

Anders:
"If you are running single node and kill the IMMND, then this ammounts to a restart of the entire IMMSv. This is a special case of the more general case where the IMMD detects loss of IMMND coord and fails to elect a new coord. The really correct way to handle this is for the IMMD to request the AMF to restart the cluster. But the current AMF implementation apparently does not support this case.

Ironically, even thoug the IMMSv and AMF fail to escalate to cluster restart, the LOGSv fails
(correctly because the runtime object should still exist) and this causes then node to restart,
which in this case equals cluster restart :-))"

[anders]

This is not a defect of the LOGSv but a defect (or lack of enhancement) in the AVSv implementation.
What is missing is support for cluster restart, or SAF midleware restart.