[Catalyst] Troubleshooting Help Please

[Paul Cory]

All,

Thanks for the help! I'm going to summarize the problem to recap, the  
solutions proposed, and what I found trying to implement them.

First the problem:

>>> I have a Catalyst Web App that requires users to log in before  
>>> they can do
>>> anything (it's a front end for managing mailing list subscriptions).
>>>
>>> The problem is that as the day progresses, and the application  
>>> sees more and
>>> more use, login behavior becomes inconsistent.
>>

>>>  As time passes,
>>> the problem gets progressively worse and it eventually gets to the  
>>> point
>>> where people quit trying because the success rate for logging in  
>>> drops to
>>> 10% or less.
>>>
>>> By fail, I mean you go to the log in screen, put in a valid  
>>> username and
>>> password combination, and then get seamlessly redirected to the  
>>> log in
>>> screen, as if you had just arrived fresh to the app.
>>>
>>> Restarting Apache solves the problem, at least for a while. And  
>>> then things
>>> begin to degrade again...

First, I got this question:

>> I do just have to ask though... why such ancient
>> software?  Apache 1 and MP1?  Ouch.

Because I haven't had the opportunity to set up a test machine with  
Apache 2/mod_perl2 and test all the stuff on our Website that might  
break. I'm a one man show, and definitely in a place where if it ain't  
broke, I don't have time to fix it. :-)

Now, the suggestions were that either I was over running the cache  
file used by Session::Store::FastMmap, or running our of file  
descriptors. In the former case, it was suggested that increase the  
cache size to 100MB, or switch to Session::Store::DBIC.

I monitored file descriptor usage for a while, and it barely changed,  
so that didn't seem to be the cause,

Then I read the documentation and tried raising the cache size for  
Session::Store::FastMmap to 100 MB. Turns out, SLES 10 doesn't like  
that - Apache child processes were aborting all over the place. I  
experimented with the cache size and found that 40MB worked, and 50MB  
and above didn't.

So, after putting that in place, I moved on to swapping out  
Session::Store::FastMmap for Session::Store::DBIC. That actually  
didn't take long at all, and, so far, appears to be working perfectly.

Thanks again for all the help.




Paul Cory
WCPSS Webmaster
pcory at wcpss.net