Zombie connections and CPU overload

[nscheffey]

Hello all-

 

I'm running FMS 11.0.5 hosted on an Xserve running OS X 10.6.8. We have been experiencing an issue where users will initiate a simple find on an indexed field and get spinning beachball of death. Their statistics in the admin console will max out (elapsed time/call 15001916), and CPU usage on the server skyrockets (150-170%, screenshot attached). Attempts to disconnect the user via admin console will successfully kick them out of their client, but the user in the statistics panel doesn't disappear, and if they reconnect there will be two instances of their name listed.

 

Server CPU usage will continue to be high for 20-30 minutes, with all users experiencing FM slowness, and then finally the ghost user connection will disappear from the admin console, and CPU usage will return to normal levels.

 

Any ideas what is causing this?

 

Thanks,

Nate

 

[mr_vodka]

Have you tried reindexing the field?

[nscheffey]

No, although it doesn't seem to be related to one particular field. Definitely related to one particular file though. Maybe I should reindex all of them.

[Wim Decorte]

This seems like symptoms of the "thread lock issue".   The fact that the server recovers after a fashion is the fix that was made in v5.  Before that the server would not recover and you'd need to reboot.

 

Without looking at the solution or what the users are doing it is hard to propose a fix.  But try to limit the finds that users do on fields that would take a long time or be meaningless   Like searching for the character "a" in a name field across 1,000,000 records...

 

Giving the server more processing power will help too.

[nscheffey]

Thank you Wim. Is there a KB article on this issue or anywhere I could get more information? Is this fixed in 12?