[Gllug] Crashing server diagnostics - Resolved

[Andy Farnsworth]

Andy Farnsworth wrote:
> Andy Farnsworth wrote:
>   
>> Karanbir Singh wrote:
>>   
>>     
>>> Andy Farnsworth wrote:
>>>   
>>>     
>>>       
>>>> After tracking down a bit more of the error, it seems that it is 
>>>> asterisk is not shutting down before the reboot and the zaptel driver 
>>>> won't unload.  If I manually shutdown asterisk first, the reboot works 
>>>>     
>>>>       
>>>>         
>>> this rings a bell somewhere...
>>>
>>> what version of asterisk and what kernel are you using ? I recall there was a 
>>> zaptel driver issue in the 4.3 / 4.2 days. Also what sort of i/o subsys do you 
>>> have there ?
>>>   
>>>     
>>>       
>> Asterisk version is from Trixbox v2.2 install CD.  Fairly recent, but 
>> not positive which version and my machine is currently running memtest 
>> so I cannot check it right now.
>>
>> The box is running 2.5 Gb RAM, an Athlon XP running at 1.6 Ghz, a brand 
>> new Seagate 500 Gb SATA HD.  Interestingly, I have never been conscious 
>> of the bandwidth available to the various caches and memtest is showing 
>> the following for this fairly old system:
>>
>> L1 Cache: 128k 10209MB/s
>> L2 Cache: 256k  3250MB/s
>> Memory: 2560M 1015MB/s
>>
>> This seems to indicate that on the Athlon XP platform, you have a 
>> Gigabye of bandwidth to RAM, 3 Gigabyte bandwidth to the L2 Cache and 10 
>> Gigabytes of bandwidth to the L1 cache.  Since PCI (32 bit/ 33 Mhz) 
>> gives you about a Gigabit of bandwidth total to the periferals, this is 
>> fairly quick.  I might have to run this CD on my desktop machine which 
>> is an Athlon 64 and see what it reports in the way of Memory bandwidth 
>> there.
>>
>> Andrew
>>   
>>     
>
> Well memtest86 reported a memory error on the box, I have reduced the 
> RAM and I am rerunning the test.  Thanks for that suggestion.
>
> Just for completness I ran it on my 2Ghz Athlon 64 and got the following 
> memory bandwidths:
>
> L1 Cache: 128k 16478MB/s
> L2 Cache: 256k  3926MB/s
> Memory: 2560M 2014MB/s
>
> Andrew
>   
Just thought I would update this post with my final conclusions.

There were multiple symptoms and it appears there were multiple causes 
which is what made this so difficult to diagnose.

Symptom:    Random Crashes, both under load and idle, no particular time 
scale (minutes to days)
Cause:         Bad RAM - memtest86+ indicated it was at approximately 
the 510Mb level
Solution:      Remove my 512 Mb stick of RAM leaving a matched pair of 
1Gb sticks

Symptom:    Kernel Panic on shutdown / reboot
Cause:         Zaptel driver not releasing due to Asterisk not shutting 
down prior to shutdown / reboot
Solution:      Shutdown Asterisk prior to reboot / shutdown

Andrew

-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug