Date: Thu, 8 Jun 2000 11:38:24 -0700 (PDT)

This seems to be a recent regression, although I haven't had
the chance to trace it back to figure out when it set in.

[Thu Jun 08 08:55:09 PDT 2000] Unexpected error in MainServer thread.
(java.lang.O
utOfMemoryError)
[Thu Jun 08 08:55:10 PDT 2000] MainServer thread is stopping.

I stumbled across this when I was making the changes for removing
the _lwp_{create,exit} interposition code to make libjvm.so's
appcert clean.

Basically, if you run Volano Test on _b18, the server (i.e. Volano
server) runs out of memory as above.

I hadn't see this problem ever before, even in very long, over 24 hr
runs with Volano. I suspect it came in with the recent _b18 promotion
(although i am yet to confirm that conjecture).

Anyone have any reason to believe that any of the changes that went in
with the _b18 promotion could have caused this problem?

I ran the server with -XX:+PrintGC, -XX:+PrintScavenge and the end of the log showed the following:

[Full GC 59901K->39481K(64896K), 1.5336192 secs]
[Full GC 64892K->37464K(64896K), 1.6226856 secs]
[Full GC 58048K->37676K(64896K), 1.5371137 secs]
[Full GC 64888K->40414K(64896K), 1.6117614 secs]
[Full GC 60990K->40536K(64896K), 1.5979006 secs]
[Full GC 64894K->38512K(64896K), 1.7134310 secs]
[Full GC 59064K->38691K(64896K), 1.6058922 secs]
[Full GC 64895K->41391K(64896K), 1.7000513 secs]
[Full GC 61943K->41529K(64896K), 1.6477583 secs]
[Full GC 64895K->39495K(64896K), 1.7117184 secs]
[Full GC 60071K->39705K(64896K), 1.6421202 secs]
[Full GC 64893K->42578K(64896K), 1.7579038 secs]
[Full GC 63130K->42969K(64896K), 1.6970144 secs]
[Full GC 64886K->40930K(64896K), 2.0260502 secs]
[Full GC 61498K->41140K(64896K), 1.7126025 secs]
[Full GC 64892K->43576K(64896K), 1.7153169 secs]
[Full GC 64152K->43992K(64896K), 1.7414857 secs]
[Full GC 64889K->41966K(64896K), 1.7337730 secs]
[Full GC 62542K->42169K(64896K), 1.7227489 secs]
[Full GC 64893K->43861K(64896K), 1.7273669 secs]


The error that June reported of a problem in uninstalling
alternate stacks is orthogonal to this problem, and I have fixed
it in my workspace. (the above behavior of running out of memory
still remains).

No further information is available at this point.

I was running run.vtest from June's scripts.
The test was being run on phalanx and failed once after 4 hours
and a second time after about 30 minutes. The same test running
on hsba4 does not run into problems, and is still running
for more than 12+ hours.

In both cases, I was running with -server (i.e. compiler2),
but that is probably orthogonal to the problem (although
i haven't confirmed this either).

The log files on phalanx can be found under:

/tmp/VolanoTestrun.12434

Note that there is no core, because the server exited "cleanly"
(see the error log in that directory from which the above was excerpted).