Support Forum

by **stampy** » Thu Oct 08, 2009 8:46 pm

our TS server is periodically experiencing "high" cpu states that slows down responsiveness for all users and this is bad. Sometimes these "slow" events last for 30 minutes. My educated guess is that users may be executing complex filters or filters that search on large task branches in TS. I say this because i noticed a while ago that executing a report on a "large" filter produced a similar affect on our server. Report running is currently disabled.

we have a total of about 30k tasks. One task branch contains about 16k tasks and the remaining tasks are split between several other branches with no more than 5000 in a single branch. Also, the branch with 16k tasks have very few users who can access that branch, so its safe to say that that branch is not being filtered on.

my question is: is there a way to log or monitor what filters are getting executed and maybe what users are executing them?

by **waldorml** » Thu Oct 08, 2009 9:22 pm

I'm curious, what version of TrackStudio are you using and with what DBMS? I ask this as my group is working on updating our installation of TrackStudio and want to try to establish some point of reference.

While I don't know about checking what filters are being used and by whom (I'm fairly certain the default filter view for a task is stored in the database but am not certain how to convert the data). I do know that it is very difficult to know when a specific user is accessing TrackStudio as users can log in and close the window (and not have to log in again). My guess would be the web-server logs could show an IP for data requests though. I do know that using a deep search filter can take a very long time to execute due to the recursion that's used to find the items on lower levels. From my experience, if you can apply the filter using the simplest terms (like avoiding heavily compounded regular expressions) on the lowest point in the issue heirarchy you'll get best results. Don't abandon hope on my response though as I'm quite sure Maxim/Admin knows his software much better than me.

by **stampy** » Thu Oct 08, 2009 9:35 pm

our version is 3.5.31
oracle database

unfortunately i would not assume that our users would pay that detailed attention to their filters.

we're not even sure that filter usage is th issue here, just a guess so far

by **stampy** » Fri Oct 09, 2009 9:41 pm

is maxim available to comment on this thread?

by **admin** » Sat Oct 10, 2009 8:05 am

How much physical memory do you have ?
Which TS version do you use ?
Which OS ?

by **stampy** » Mon Oct 12, 2009 5:24 pm

our TS version is 3.5.31
oracle database
the server (linux VM) has 4GB of ram i think

by **admin** » Mon Oct 12, 2009 6:22 pm

stampy wrote:our TS version is 3.5.31
oracle database
the server (linux VM) has 4GB of ram i think

Please open jetty file, go to the last lines and edit java startup parameters - add something like "-Xmx1500m" to increase java heap.

Let me know if this occurs again after this change.

by **stampy** » Mon Oct 12, 2009 8:22 pm

we are already using the max value for -Xmx , that is, 75% of the RAM amount

by **admin** » Mon Oct 12, 2009 8:33 pm

stampy wrote:we are already using the max value for -Xmx , that is, 75% of the RAM amount

ok. Please try to reduce filter result set: filter out old tasks, tasks with some handler, etc.

by **stampy** » Tue Oct 13, 2009 12:01 am

you mean go through the filters of all our users to try and might their query results smaller? I cant do that for hundreds of users! Thats why my original question involved trying to understand what filters are getting executed and when - so that maybe we could match filter usage with the server slowdown. We dont even know if the root cause of the slowdown is filter usage, but thats where we're starting.

Support Forum

server slowness

server slowness

Who is online