2

We are seeing both hogging threads and queue length spike. Sometimes at the same time, sometimes not. Hogging threads will go as high as 100, while queue length will regularly spike to 700. I have dug through all the stats/log files I can find and I can't see any correlation with any errors. However, we do see a large number of query timeouts (2-4 thousand per day) between 10-180 seconds. The query timeouts are almost constant.

This is an enterprise level application and we average 700 concurrent users throughout the workday. They report massive slowdowns (10-30 seconds) in page loads almost everyday.

The system is not being taxed. CPU load averages 60%, network utilization is less than 15%. Database server is seeing the same OS stats.

We are running Weblogic 10.3.4 with a JVM heap size of 12GB with a 1:1 ratio old to new gen. Solaris 10 x64 with 64GB RAM. There is a separate Oracle database server with the same OS.

Does anybody here have any experience in dealing with these types of issues?

ericbond007
  • 76
  • 1
  • 6
  • 2
    On the WLS side, look at thread dumps, take 4-5 at 10 seconds apart, see if you see any stuck threads: `kill -3 ` ought to do it. Examining GC logs could be helpful. If query timeouts are constant, looking at what queries need optimizing would be beneficial. – KM. Jul 02 '14 at 16:33
  • 2
    Hi look there is contention among threads ,can you please take thread dump in 5 second interval .10 dump will be fine and try to see if there are blocked threads trying to get monitor in a synchronized block,let me know if i can help further –  Jul 04 '14 at 13:19
  • You should take a look at: EJB pool stats (should you increase some pools?), JDBC pool stats (do you need more connections to DB? are there many waits for free connections?), JVM Heap (do you have enough free heap on the managed servers or do they have to garbage collect all the time?) – fgysin Jul 31 '14 at 11:42

0 Answers0