Allocation failed managing allocation failure

I have webmethods 6.5 running on WIN 2000.
Implemented Rosettanet PIP 3A4 Buyer. The flow takes a 3A4 xml converts it into a PIP 3A4 and submits to uses by https.
I started IS from command line with verbose and I see this message

<af[1]:>
<AF[1]: managing allocation failure, action=1 (4032/1339030520) (3145728/3145728
)>
<GC: Sun Jul 10 16:05:43 2005
<gc(1):>
<gc(1):>
<gc(1):>= 32), weak 4312, final 34751, phantom 0>
<af[1]:>

During this message, the flow service seems to hang for 2-3 min.
Does this message, indicates any error ? how to overcome this error ?

I assume you’re referring to Integration Server (webMethods is a company, not a product).

Search the 'net for “managing allocation failure” and you’ll see lots and lots of info about garbage collection, which is what your 2-3 minute “hang” is. Change your memory and/or garbage collection strategy settings on startup (more memory available to the JVM) or adjust your integration to make more efficient use of the available memory.

yes I am referring to IS 6.5.

The box has around 1G of RAM. JVM min/max 256/768.
GC in IS runs every 30mins

Apart from these, what else can I update/changed

UMG,
To be honest your box really doesn’t have enough memory. 1GB isn’t enough to accommodate both the IS server(unless you have no flow services and your IS is doing something basic like serving as a Reverse Invoke Server) and the Windows OS. 2GB would do it. With your current memory your JVM is going to have to swap to dsk which will make your performance less than optimal.

As far as your GC goes, every 30 minutes? Sounds like you have it scheduled? I would not do that to start with. Take a look at your flow service, make sure it is optimized first ie dropping pipeline variables when you should, eliminate unnecessary mappings etc. You didn’t mention details on document size. A large document? How large? Flow service doing anything else?

markg
http://darth.homelinux.net

The GC is not scheduler, the watt.server.cache.gcMins=30 min.
the document are small – 50-100K. Most of the service that are used are RosettaNet service. Custom service are every simple and direct mapping

watt.server.cache.gcMins does not control JVM garbage collection. It controls the removal of service cache data.

Depending on your document volume, your machine may indeed be undersized. What’s the document volume?

Try changing the JVM min/max from 256/768 to 768/768. If IS is the only thing that is running on this machine (if there are other things, this machine is really undersized), then might as well let allocate all the memory from the OS right up front.