Is Lucene query provider tunable?

[Evan Moseman]

We are seeing very slow performance that all seem to point to the Lucene search engine using JIRA 6.3.  Our JIRA instance is running on a 32CPUx84GB machine, with 24GB configured for the JVM.  Despite not seeing any indicators that there is a resource shortage on the server, accessing Jira through the web interface is slow.  Browsing a project summary takes 26-30 seconds to respond.  While the searches take that long to return , the load average on the host is 0.77, and there are 0 processes with any IOWait percentage. 

If I do a thread dump there are 16 threads in this same place:

"http-bio-127.0.0.1-8000-exec-307" daemon prio=10 tid=0x00007fe3c0001000 nid=0x21aa runnable [0x00007fe6f7b83000]
java.lang.Thread.State: RUNNABLE
at sun.misc.Unsafe.getByte(Native Method)
at java.nio.DirectByteBuffer.get(DirectByteBuffer.java:249)
at java.nio.ByteBuffer.get(ByteBuffer.java:673)
at java.nio.DirectByteBuffer.get(DirectByteBuffer.java:279)
at org.apache.lucene.store.MMapDirectory$MMapIndexInput.readBytes(MMapDirectory.java:289)
at org.apache.lucene.index.TermBuffer.read(TermBuffer.java:82)
at org.apache.lucene.index.SegmentTermEnum.next(SegmentTermEnum.java:131)
at org.apache.lucene.index.SegmentTermEnum.scanTo(SegmentTermEnum.java:166)
at org.apache.lucene.index.TermInfosReader.get(TermInfosReader.java:273)
at org.apache.lucene.index.TermInfosReader.get(TermInfosReader.java:209)
at org.apache.lucene.index.SegmentReader.docFreq(SegmentReader.java:503)
at org.apache.lucene.search.TermQuery$TermWeight$1.add(TermQuery.java:56)

...

We reindex regularly...

If I turn on Profiling, the Lucene search provider typically uses about 98% of the time.

Any ideas?

[Daniel Wester]

Where is the JIRA home directory hosted and have you look at it's performance?

[Gonchik Tsymzhitov]

Yeah, latency of file system indeed needful :) 

https://community.atlassian.com/t5/Jira-articles/The-story-about-how-to-improve-the-background-indexing-Lucene/ba-p/1568553