[InterMine Dev] web app restarts

Joe Carlson jwcarlson at lbl.gov
Fri May 8 01:28:06 BST 2015


Hi Julie and gang

We have just deployed our latest phytozome build based on intermine 1.4. 
This is our first public release using Hikari.

Our hopes were that going to hikari would solve some of the problem 
we've been seeing about tomcat restarts. We've traded emails about this 
in the past where we see that we have to restart every couple of hours 
when under load. (We run a 'are you alive' cron job every 3 minuts and 
force a restart if we don't get a response.)

At the time I think we had attributed it to the postgres connection and 
we looking forward to the hikari pooling. It behaved well in internal 
use, but now that we're public we continue to see the restarts.

I'm trying to do a little forensics to see what might be causing them. 
I'm seeing "OutOfMemoryError: Java heap space", typically after a call 
to retrieving the genomic sequence of a region 
(service/regions/sequence). I had noticed that you had removed this 
service in 1.4. I restored it since we're making use of it to deliver 
sequence to our main web portal. Did you remove this because you had 
seen it as being problematic?

At this point, I'm not absolutely sure this is the source of the 
restarts but I'm very suspicious of 
org.intermine.bio.web.export.GenomicRegionSequenceExporter. There is a 
static map of entire chromosomes that is being stored. The substring is 
retrieved by calling substring on elements of this map. This may work 
for smaller mines but we have enough sequence in our database that I 
suspect this is part of our problem.

Was this web service removed deliberately? Is there something to replace 
it? As I recall, the other sequence retrieval services I found only 
retrieved the sequence for specific features and not chromosome slices.

Thanks,

Joe Carlson



More information about the dev mailing list