RP/Tux 8.1 - restartsrv instances flood the process table (Doc ID 776901.1)

Goal

During a failure scenario in an MP configuration, in unusual circumstances it is possible for the BRIDGE to spawn
restartsrv processes in a tight loop, until an OS system limit is reached.  Subsequently the host becomes unusable,
the processes must be manually killed, and the application must be manually restored.

Here's an example scenario that leads to this situation.  The configuration has 2 LMIDs -- SITE1 (master) and SITE2
(secondary).
SITE1> tmboot -y
... everything ok
SITE1> some failure occurs, all IPC resources on SITE1 are removed.  Consequently SITE1's servers --- BBL, DBBL,
BRIDGE, etc. -- exit.
SITE2> servers are still running ok
SITE2> kill -9 <BBL> -- something kills the BBL.  Then, the BRIDGE notices that the BBL has died and tries to
restart it.

At this point, SITE2's BRIDGE starts restartsrv, which then tries to start the BBL.  However, since the DBBL can't be
contacted, the restart fails.  However, restartsrv keeps creating new processes until the user's OS limit has been
exceeded.

Solution

My Oracle Support provides customers with access to over a million knowledge articles and a vibrant support community of peers and Oracle experts.