[Treebase-guts] [ treebase-Bugs-2953212 ] Hang-up in Search Result tabs

[SourceForge.net <no...@so...>]

Bugs item #2953212, was opened at 2010-02-16 19:48
Message generated for change (Settings changed) made by vgapeyev
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=1126676&aid=2953212&group_id=248804

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: data
Group: None
Status: Open
Priority: 9
Private: No
Submitted By: Vladimir Gapeyev (vgapeyev)
>Assigned to: Nobody/Anonymous (nobody)
Summary: Hang-up in Search Result tabs

Initial Comment:

After Jon and I addressed #2945410 [Proxy Error for taxon->tree search] by increasing Apache timeout to 10 min and #2945391 [New Indices Needed], a new issue transpired on search results: some "click paths" through the result tabs lead to a hang-up.  

To reproduce: 
- Search for "oak", clicking "All Text".
- After this finishes, click on result tabs, avoiding the "Taxa" tab.  I.e., keep clicking on "Matrices", "Trees", and "Tree topologies", in any order, as long as you wish.  They all should be finishing after a short delay. 
- Now, click on "Taxa".  It should finish fairly soon. 
- Now, click on "Matrices". This hangs up forever. 

A bonus observation: If (prior to "Taxa") you continue clicking on "Matrices" several times in a row, it appears you get a new result list each time.  This looks like a bug, unless it is a clever feature.

We spent a couple hours trying to figure out what is going on.  No major revelations, but here are a few observations: 

- When the hang up happens, there is a query process stuck on postgres that does not finish even for an hour. 

- The query process takes about 33% CPU, so something active might be going on during the hang-up. Also, responsiveness of the server to concurrent sessions degrades. 

- After looking up tips on the internets, we tried increasing size of the Tomcat's JDBC connections pool, with no effect. 

- Following a hunch that the switch from the c3p0 connection pool manager to that of Tomcat might have contributed, we rebuilt the WAR from SVN #422, i.e. right before the switch to JNDI.  That one hangs up, too. 

My best guess at the moment is that there is some kind of interaction     between queries on this page that leads to a "busy" deadlock.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=1126676&aid=2953212&group_id=248804