Re: Runaway Process

In article <CqD2tt.6qB_at_csn.org> losburn_at_teal.csn.org (Lynn Osburn) writes:
>ENV: 7.0.13 running Financials 9.4.2 on HPUX 9.04
>
>We see one, two, or occasionally three FNDLIBR processes running wild,
>consuming all CPU. Generally we find them in the morning when we come
>in, stop them, check to be sure that any users' long-running jobs haven't
>stopped, them kill them.
>
>Since they are Concurrent Manager processes, I suspect that they are
>resilient survivors of the nightly CONCSUB CONCURRENT FND ABORT. They
>seem to miss the kill, then become detached from the database (no
>corresponding oracle backend process), get reaped by init (new PPID=1)
>and then run wild.
>
>Anyone else seen this? Anyone care to speculate on a clean, sanitary
>solution? Anyone care to set up a pool to wager on the response time
>from Oracle Support? <grin>
>
>
We've seen the same problem when the database is shut down before _all_ FNDLIBR processes have terminated.

Sounds like you're doing a nightly shutdown for an export or backup. You might want to check the database shutdown time in the "alert" log against the concurrent manager shutdown time in the "mgrlog".

Until we arrive at a more elegant solution, we've scripted a hard-coded delay of 3 minutes between the CONCSUB abort and the database shutdown. This seems to do the trick in giving the FNDLIBR processes time to go away.

• Tim

  ---

Tim Heath                                   E-Mail: heath_at_network.com
Network Systems Corporation                 Phone:  (612) 424-1784
------------------------------------------------------------------------