[collectd] Bug in utils_cache.c

[Paul Sadauskas]

Yeah, thats the exact message I get in the log, so its probably the same bug.

I added the continue in my patch, and it prevented it form crashing.
However, now I see in my log:

[2009-07-07 13:55:53] uc_check_timeout: apollo/uptime/uptime is
missing but ``uninteresting''
[2009-07-07 13:55:53] uc_check_timeout: apollo/users/users is missing
but ``uninteresting''
[2009-07-07 13:55:56] uc_check_timeout:
apollo/tcpconns-53-local/tcp_connections-TIME_WAIT is missing but
``uninteresting''

And no output after that, and its been two hours. I guess I need to
figure out what makes something uninteresting, and why its happening.


On Tue, Jul 7, 2009 at 2:11 PM, Mariusz Gronczewski<xani666 at gmail.com> wrote:
> I think i found another one, after compiling today's git snapshot.
> When i fire collectd after few (5-30)s i get
> collectd: utils_cache.c:55: cache_compare: Assertion `(a != ((void
> *)0)) && (b != ((void *)0))' failed.
>
> this is server which gathers data from several other servers so prob.
> on less busy machine it will take more time to trigger.
> Let me know if u want additional data.
>
> Mariusz.
>
> 2009/7/7 Paul Sadauskas <psadauskas at gmail.com>:
>> tokkee on IRC & I think we found a bug with utils_cache.c. The
>> uc_check_timeout function is missing a continue after the
>> "uninteresting" service check, that causes a key to be null.
>>
>> Here's a quick patch:
>>
>> diff --git a/src/utils_cache.c b/src/utils_cache.c
>> index 5e62aa4..ea46750 100644
>> --- a/src/utils_cache.c
>> +++ b/src/utils_cache.c
>> @@ -345,6 +345,7 @@ int uc_check_timeout (void)
>>       sfree (keys[i]);
>>       sfree (key);
>>       cache_free (ce);
>> +      continue;
>>     }
>>
>>     /* If we get here, the value is ``interesting''. Query the record from the
>>
>>
>> I have a backtrace showing the log and error message here:
>> http://gist.github.com/142274
>>
>> I'm not sure if this is all that's needed, or even tested it. It takes
>> ~24 hours for this bug to show itself.
>>
>> Let me know if you need any more info.
>>
>> Paul
>>
>> _______________________________________________
>> collectd mailing list
>> collectd at verplant.org
>> http://mailman.verplant.org/listinfo/collectd
>>
>