no specific events trigger the hang… cant get thread dumps from console(i dont have the -Xrs switch in server.bat) when this occurs…
it occurs at random times and during random transactions… it could be SAP, could be DB i have nothing in logs to point me in either direction and ive had both at 10 since this started happening…
ill check for that patch, but we’re on 4.6 and that should be included in that release…
We JUST had a major hang. It was due to too many threads on our SQL Server Trading Networks database. We had to reboot the database server and that took care of the problem.
Do you have a firewall between SAP-gateway and IS? If this firewall has an automatic cut off after time out you will get many lost threads in the gateway and IS.
This did only happen with java.exe getting huger: Do you have the repository cleaned very frequently? If not, the repository begins to block IS.
We had similar symptons (w/ plenty hangups), but no SAP adapter use. After analysis found out some threads not being closed/killed by IBM JVM. Changed to Sun JVM 1.3 and the issue never happened
IS hang can be casused by many things but casue is usually same or related. In most cases it is due to exhausted thread pool and all requests are blocked! In Versions 4.6 or earlier you cannot tell exactly how may threads are active. Admin using JVM get active thread API but it does not return a true value. Only reflection that has been true at some point in time. Your IS may infact be out of threads and increasing them will only extend the life but will hang later. Also having threadpool too large like 500-1000 may imact server on some JVM with garbage collection. You need to investigate what porcess hanged threads! Another possibility is IS JVM process is ok but listener manager has crashed that will look like a crash, you cannot connect with anything to IS but process is ok. You can chek this by setting some test scheduled job to run. And chek if its still running after IS hanged.
However scheduler can also cause hang for same resource exhaustion reason. There are severl patches even for 6.0.1. to fix this porblem. I know 4.6 had same issue. If you run scheduler jobs alot make sure you have patch to start jobs on different sessions!
If you’re going to use 6.0.1, and you want to do the test with scheduled jobs mentioned above by Igor, check also the patches for scheduled jobs since there are bugs known that prevent the schedular to pick up new tasks.
I would suggest the following in order:
putting GC on scheduler.
Increase the number of threads for scheduled jobs.
ensure that you have the latest service packs installed on the 4.x versions.
check for any patches/fixes for 4.x versions from wm-support.
We were provided a patch by webMethods back when we were on 4.6 to split the scheduler thread pool and user schedule tasks so that they would have their own thread pools. This patch never really solved the problem of the server getting hung up for us.