In our production environment, we are receiving sporadic but constant 2013
errors from one of our mysql dbs.
The production environment consists of 5 servers running RH 6.2/apache
1.3.27/php 4.0.6, load balanced with a foundry server iron. Each app box
queries a central db box, using pconnect, (rh 8.0/mysql 4.0.12) primarily
for reads (roughly 80% of a script's db transactions) and another db box for
writes (rh 6.2/mysql 4.0.12).
Logic is in place to log any mysql errors on each app box from either mysql
server. Over the past month, well pretty much since logging was enabled, the
'read' server returns quite a few 2013 errors. We log the 'suspect' query
and the queries are valid. Not every query results in an error, but enough
where we get up to 50 an hour from every box. During peak times we average
around 500 qps. The db server in question is pretty robust and hardly ever
carries a load above 1.0. We are using a pretty standard 'huge' version of
my.conf, except for the following changed lines:
skip-innodb
skip-name-resolve
set-variable=thread_stack=256k
set-variable=wait_timeout=60
set-variable=thread_cache_size=40
From scanning previous posts related to errno 2013, we have adjusted
'thread_stack' to above value as well as the 'wait_timeout'. Nothing so far
seems to correct the problem. There seems to be no rhyme or reason as to
what causes the errors, many scripts are executed during the day. In fact
nobody in our office has even experienced the error first-hand, but logs
indicate that it does occur and quite frequently
Thanks in advance.
Kevin A. Miller
