Issue on 10.1B03

[anoobbhaskar]

Our progress version is 10.1B03 (RISC) in HP UX box(version 11.23). Our system usually experiencing a hanged broker and some times all brokers went to SENDING mode and persists for some time. This would even cause the performance issue and some time brokers were not available to process a new request(s). I have also seen that there are high usage of RAM and IO in our server. Our Serever has 24GB RAM and SAN is connected. The database size is 70GB and the maximum appeserver/webspeed brokers are 85 and minimum is 30.


I have seen the following comments from one of the forum and I would like to upgrade to new version?
I believe the below information and issues I am getting from the server is almost similar and this could be one of the reson for the performance issue . Can you please let me know your comments as well.

Unused AppServers
Performance is reduced when AppServers trim when they are no longer needed. Unused
AppServers hang and then are not available for future processes that need them.This issue is resolved
within the Open Edge 10.1B0307 (or higher) and Open Edge 10.0B0546 (or higher) service packs.
Note: Increasing the number of servers uses more RAM on the server. The application may run slower, but
users should be able to connect with less change of failure. Open Edge 10.1B03 hot fix 7 and higher (or
Open Edge 10.1B05 hotfix 46 or higher) resolves this issue. After either patch is installed, restore these
Pool Range values to their original values.

 

[Cringer]

I've no idea about your specific problem, but whenever I've come across things sitting at SENDING it's usually because code is connecting to them but not disconnecting afterwards. Have you checked your code?

 

[anoobbhaskar]

Yes. I am not sure this issue is related to the code because all available brokers became sending mode and persists for some time. It will be available after some time. Usaully one broker will be in a SENDING mode if a program has iss ue when the user tries to access the program correct?

Would be great if you can validate the statements described under "Unused Appservers" are true ? I have also seen some of the posting related to 10.1B in this forum discussing about the PURGE lock. I have also seen that status in the promon.

I've no idea about your specific problem, but whenever I've come across things sitting at SENDING it's usually because code is connecting to them but not disconnecting afterwards. Have you checked your code?

 

[cj_brandt]

Our policy is to not install hot fixes unless it is an emergency. If I thought the issue for the agents stuck in SENDING mode was related to a Progress bug, I would try to upgrade to the most recent version of OE - not just a hot fix on my current version.

If you are having a problem when unused app servers trim back, why not set your min and max servers to the same value to avoid the agents being trimmed ?