Re: [ZODB-Dev] "MySQL server has gone away" with RelStorage

2010-10-25 Thread Paul Winkler
On Mon, Oct 25, 2010 at 05:53:05PM +0100, Chris Withers wrote:
> On 25/10/2010 15:57, Paul Winkler wrote:
> > On Sun, Oct 24, 2010 at 05:55:05PM +0100, Chris Withers wrote:
> >> Restarting Zope appeared to fixed the problem, but I'm slightly confused
> >> as I can't find any entries in the logs of the Apache instance that sits
> >> in front of the test instances for the failed requests...
> >
> > Maybe turn on Zope's trace log?
> 
> Okay, but regardless of what happens with Zope, surely I'd expect a row 
> in the Apache log saying what happened?

Depends. If the user saw a 50x error page from the server, then yes.

If the *browser* timed out before receiving a response, then maybe
not.  Like Zope, Apache doesn't put a line in the access log until it
has finished processing the request and sent a response back. If Zope
is more or less hung, and Apache has a long ProxyTimeout configured (I
believe the result is 300 seconds), then you might not see anything
logged.  I am not sure what the default timeouts in various browsers
would be, but my subjective impression is that it's lower than 300
seconds.

That's why I suggested the trace log; you can see exactly what
requests come in as they arrive, and whether & when a response gets
sent back.

-- 

Paul Winkler
http://www.slinkp.com
___
For more information about ZODB, see the ZODB Wiki:
http://www.zope.org/Wikis/ZODB/

ZODB-Dev mailing list  -  ZODB-Dev@zope.org
https://mail.zope.org/mailman/listinfo/zodb-dev


Re: [ZODB-Dev] "MySQL server has gone away" with RelStorage

2010-10-25 Thread Chris Withers
On 25/10/2010 15:57, Paul Winkler wrote:
> On Sun, Oct 24, 2010 at 05:55:05PM +0100, Chris Withers wrote:
>> Restarting Zope appeared to fixed the problem, but I'm slightly confused
>> as I can't find any entries in the logs of the Apache instance that sits
>> in front of the test instances for the failed requests...
>
> Maybe turn on Zope's trace log?

Okay, but regardless of what happens with Zope, surely I'd expect a row 
in the Apache log saying what happened?

I'm just wondering whether I should be suspicious of MySQL + RelStorage 
+ Zope or if I should be looking for some weird Apache issue...

Chris

-- 
Simplistix - Content Management, Batch Processing & Python Consulting
- http://www.simplistix.co.uk
___
For more information about ZODB, see the ZODB Wiki:
http://www.zope.org/Wikis/ZODB/

ZODB-Dev mailing list  -  ZODB-Dev@zope.org
https://mail.zope.org/mailman/listinfo/zodb-dev


Re: [ZODB-Dev] "MySQL server has gone away" with RelStorage

2010-10-25 Thread Paul Winkler
On Sun, Oct 24, 2010 at 05:55:05PM +0100, Chris Withers wrote:
> Restarting Zope appeared to fixed the problem, but I'm slightly confused 
> as I can't find any entries in the logs of the Apache instance that sits 
> in front of the test instances for the failed requests...

Maybe turn on Zope's trace log?

-- 

Paul Winkler
http://www.slinkp.com
___
For more information about ZODB, see the ZODB Wiki:
http://www.zope.org/Wikis/ZODB/

ZODB-Dev mailing list  -  ZODB-Dev@zope.org
https://mail.zope.org/mailman/listinfo/zodb-dev


Re: [ZODB-Dev] "MySQL server has gone away" with RelStorage

2010-10-24 Thread Chris Withers
On 23/10/2010 18:48, Shane Hathaway wrote:
> On 10/23/2010 01:29 AM, Chris Withers wrote:
>> Hi Shane,
>>
>> With the server I'm testing on, when coming back to test after a couple
>> of days working on something else, I get the following logged at Warning:
>>
>> Reconnecting load_conn: (2006, 'MySQL server has gone away')
>>
>> Does this mean the reconnection is being handled by RelStorage or that I
>> need to do some work?
>
> That just means MySQL is timing out connections, which is fine. That
> message was probably followed immediately by an info message saying that
> it reconnected.

Yep, that matches.

>> Also, after this, it seems life the ZODB object cache is emptied.
>> Is that happening or is it just my imagination?
>
> That's possible, but reconnection alone should not cause it.

What could cause a cache to empty like this?

More worryingly, another of the test sites (these sites only get hit in 
patches every few days when people come to do test runs) just timed out 
this morning. I can't see any evidence of the timed out requests in 
either the access log nor anything in the event log.

Restarting Zope appeared to fixed the problem, but I'm slightly confused 
as I can't find any entries in the logs of the Apache instance that sits 
in front of the test instances for the failed requests...

Has anyone seen anything like this?

cheers.

Chris

-- 
Simplistix - Content Management, Batch Processing & Python Consulting
- http://www.simplistix.co.uk
___
For more information about ZODB, see the ZODB Wiki:
http://www.zope.org/Wikis/ZODB/

ZODB-Dev mailing list  -  ZODB-Dev@zope.org
https://mail.zope.org/mailman/listinfo/zodb-dev


Re: [ZODB-Dev] "MySQL server has gone away" with RelStorage

2010-10-23 Thread Shane Hathaway
On 10/23/2010 01:29 AM, Chris Withers wrote:
> Hi Shane,
>
> With the server I'm testing on, when coming back to test after a couple
> of days working on something else, I get the following logged at Warning:
>
> Reconnecting load_conn: (2006, 'MySQL server has gone away')
>
> Does this mean the reconnection is being handled by RelStorage or that I
> need to do some work?

That just means MySQL is timing out connections, which is fine.  That 
message was probably followed immediately by an info message saying that 
it reconnected.

> Also, after this, it seems life the ZODB object cache is emptied.
> Is that happening or is it just my imagination?

That's possible, but reconnection alone should not cause it.

Shane
___
For more information about ZODB, see the ZODB Wiki:
http://www.zope.org/Wikis/ZODB/

ZODB-Dev mailing list  -  ZODB-Dev@zope.org
https://mail.zope.org/mailman/listinfo/zodb-dev


[ZODB-Dev] "MySQL server has gone away" with RelStorage

2010-10-23 Thread Chris Withers
Hi Shane,

With the server I'm testing on, when coming back to test after a couple 
of days working on something else, I get the following logged at Warning:

Reconnecting load_conn: (2006, 'MySQL server has gone away')

Does this mean the reconnection is being handled by RelStorage or that I 
need to do some work?

Also, after this, it seems life the ZODB object cache is emptied.
Is that happening or is it just my imagination?

cheers,

Chris

-- 
Simplistix - Content Management, Batch Processing & Python Consulting
- http://www.simplistix.co.uk
___
For more information about ZODB, see the ZODB Wiki:
http://www.zope.org/Wikis/ZODB/

ZODB-Dev mailing list  -  ZODB-Dev@zope.org
https://mail.zope.org/mailman/listinfo/zodb-dev