I seen it so many times; after moving to JRockit JVM I had clean error
logs; after increasing memory such problems disappeared. In most cases
OOME happens when someone executes 'sorted' query. JVM behaviour in
case of catched OOME is very unpredictable... SOLR catches everything
and tries to output (sometimes with great success) HTTP 500 with
XML-formatted stacktrace as content.
Quoting Peter Williams <[EMAIL PROTECTED]>:
Thanks.
Does a message about what went wrong not get written to the log because by
the OOME happens there is not enough memory to actually construct the
message?
Peter
On Mon, Sep 15, 2008 at 2:40 PM, Fuad Efendi <[EMAIL PROTECTED]> wrote:
SOLR main servlet catches all Throwable. In case of very popular OOME with
standard JVM from SUN you will get exactly this behaviour.
==============
http://www.tokenizer.org/bot.html
Quoting Peter Williams <[EMAIL PROTECTED]>:
I am using Solr 1.2.0 with Jetty and I am experiencing some odd failures
of
Solr. Solr seems to just stops listening for new TCP connections. The
Solr
process continues running and log contains nothing suspicious (to me,
anyway) but curl requests against the server fail with "connection
refused"
and `netstat --listening` does not list the Solr port. Restarting Solr
fixes the problem (curl request work, etc). This happens once or more a
day.
Any one have any ideas about how to figure out what is going wrong?
Peter Williams
