Gracefully handling "MySQL has gone away"

Question

I'm writing a small database adapter in Python, mostly for fun. I'm trying to get the code to gracefully recover from a situation where the MySQL connection "goes away," aka wait_timeout is exceeded. I've set wait_timeout at 10 so I can try this.

Here's my code:

def select(self, query, params=[]):
        try:
            self.cursor = self.cxn.cursor()
            self.cursor.execute(query, params)
        except MySQLdb.OperationalError, e:
            if e[0] == 2006:
                print "We caught the exception properly!"
                print self.cxn
                self.cxn.close()
                self.cxn = self.db._get_cxn()
                self.cursor = self.cxn.cursor()
                self.cursor.execute(query, params)
                print self.cxn

        return self.cursor.fetchall()

Next I wait ten seconds and try to make a request. Here's what CherryPy looks like:

[31/Dec/2009:20:47:29] ENGINE Bus STARTING
[31/Dec/2009:20:47:29] ENGINE Starting database pool...
[31/Dec/2009:20:47:29] ENGINE POOL Connecting to MySQL...
[31/Dec/2009:20:47:29] ENGINE POOL Connecting to MySQL...
[31/Dec/2009:20:47:29] ENGINE POOL Connecting to MySQL...
[31/Dec/2009:20:47:29] ENGINE POOL Connecting to MySQL...
[31/Dec/2009:20:47:29] ENGINE POOL Connecting to MySQL...
[31/Dec/2009:20:47:29] ENGINE Started monitor thread '_TimeoutMonitor'.
[31/Dec/2009:20:47:29] ENGINE Started monitor thread 'Autoreloader'.
[31/Dec/2009:20:47:30] ENGINE Serving on 0.0.0.0:8888
[31/Dec/2009:20:47:30] ENGINE Bus STARTED
We caught the exception properly!   <====================================== Aaarg!
<_mysql.connection open to 'localhost' at 1ee22b0>
[31/Dec/2009:20:48:25] HTTP Traceback (most recent call last):
  File "/usr/local/lib/python2.6/dist-packages/CherryPy-3.1.2-py2.6.egg/cherrypy/_cprequest.py", line 606, in respond
cherrypy.response.body = self.handler()
  File "/usr/local/lib/python2.6/dist-packages/CherryPy-3.1.2-py2.6.egg/cherrypy/_cpdispatch.py", line 25, in __call__
    return self.callable(*self.args, **self.kwargs)
  File "adp.py", line 69, in reports
    page.sources = sql.GetSources()
  File "/home/swoods/dev/adp/sql.py", line 45, in __call__
    return getattr(self.formatter.cxn, parsefn)(sql, sql_vars)
  File "/home/swoods/dev/adp/database.py", line 96, in select
    self.cursor.execute(query, params)
  File "/usr/lib/pymodules/python2.6/MySQLdb/cursors.py", line 166, in execute
    self.errorhandler(self, exc, value)
  File "/usr/lib/pymodules/python2.6/MySQLdb/connections.py", line 35, in defaulterrorhandler
    raise errorclass, errorvalue
OperationalError: (2006, 'MySQL server has gone away')

[31/Dec/2009:20:48:25] HTTP 
Request Headers:
  COOKIE: session_id=e14f63acc306b26f14d966e606612642af2dd423
  HOST: localhost:8888
  CACHE-CONTROL: max-age=0
  ACCEPT: application/xml,application/xhtml+xml,text/html;q=0.9,text/plain;q=0.8,image/png,*/*;q=0.5
  ACCEPT-CHARSET: ISO-8859-1,utf-8;q=0.7,*;q=0.3
  USER-AGENT: Mozilla/5.0 (X11; U; Linux x86_64; en-US) AppleWebKit/532.5 (KHTML, like     Gecko) Chrome/4.0.249.43 Safari/532.5
  CONNECTION: keep-alive
  Remote-Addr: 127.0.0.1
  ACCEPT-LANGUAGE: en-US,en;q=0.8
  ACCEPT-ENCODING: gzip,deflate
127.0.0.1 - - [31/Dec/2009:20:48:25] "GET /reports/1 HTTP/1.1" 500 1770 "" "Mozilla/5.0 (X11; U; Linux x86_64; en-US) AppleWebKit/532.5 (KHTML, like Gecko) Chrome/4.0.249.43 Safari/532.5"

Why doesn't this work?? I clearly catch the exception, regenerate both the connection and the cursor, but it still doesn't work. Is it related to how MySQLdb gets the connections?

Answer 1

Can't see from the code, but my guess would be that the db._get_cxn() method is doing some kind of connection pooling and returning the existing connection object instead of making a new one. Is there not a call you can make on db to flush the existing useless connection? (And should you really be calling an internal _-prefixed method?)

For preventing MySQL has gone away I generally prefer to keep a timestamp with the connection of the last time I used it. Then before trying to use it again I look at the timestamp and close/discard the connection if it was last used more than a few hours ago. This saves wrapping every possible query with a try...except OperationalError...try again.