I did indeed discover that the problem was in the app. When one
instance of my app hangs, JavaMonitor says that wotaskd is hung and
all kinds of bad things happen. Either the apache adaptor does not
redirect to a functioning instance or if it does, it does so after
a long pause.
How long? The default is 15 seconds. If you have multiple hung
instances, it could redirect to one of them.
The problem is that it never times out. Any user trying to go to that
instance just hangs forever.
Is there some reason that JavaMonitor/wotaskd/Apache Adptor can not
deal correctly with a hung instance? Is there anything I can do to
make it behave better?
Define better. It deals with it fairly well, given that it can only
assume what some symptoms might mean.
See above. It is workable for 15 seconds, but it seems to hang forever.
Also, I have to turn direct connect off to be able to do the refuse
new sessions, so is there another way to figure out which app is
hung so that it can be killed?
Look at the stack trace, like below. Also look for growing numbers
of worker threads. If you have more than 16, you probably have a
problem.
I will look into that.
Thanks,
Frank
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (Webobjects-dev@lists.apple.com)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/webobjects-dev/archive%40mail-archive.com
This email sent to arch...@mail-archive.com
