Hi Ilya,
I use Charles Foster implementation of XML:DB API, version 1.2.6 without
any customizations: http://www.cfoster.net/sedna/xmldb/download/
Here is the link again to the connection pool documentation (it's a part of
the above implementation): http://www.cfoster.net/sedna/xmldb/download/
Meanwhile, I've identified the application that makes Sedna wait. At least
restarting the application has unfreezed the database twice already. I'll
analyze the application log along with Sedna event.log shortly. Hopefully,
I'll find the root cause.
Best regards,
Ivan Lagunov
On Thu, Oct 9, 2014 at 3:50 AM, Ilya Taranov <epsi...@socio.msu.ru> wrote:
> I can easily see the scenario, where one of the clients doesn't terminate
> the transaction properly. Unfortunately, current application timeout
> doesn't apply for client hangs. The issue can be easily fixed, however.
>
> Unfortunately, even fix still would not solve your problem, you would
> still have application timeout wait on locks. Probably the most
> straightforward way to find the cause for you would be to log connection
> checkins/checkouts at the client driver level.
>
> Could you please give me a link to the connection pool + driver you
> currently use (including all your changes if possible) so I could help you
> with that?
>
> On Tue, Oct 7, 2014 at 4:37 AM, Ivan Lagunov <lagi...@gmail.com> wrote:
>
>> Hi Ilya,
>>
>> Your suggestion seems to be correct. After I've restarted three Java
>> applications that had connections to Sedna, the database has resumed
>> processing operations. Here is the event.log where Sedna has hanged between
>> 12:35:03 and 13:06:05:
>>
>>
>> https://www.dropbox.com/s/pzafuvzcvfchxwa/event-2014-10-07-13-15-04-018.zip?dl=0
>>
>> I'm not sure which of the applications caused it though. It still worries
>> me as the queries before hanging look quite simple and I cannot understand
>> which client operation could have hanged the whole database.
>>
>> Best regards,
>> Ivan Lagunov
>>
>> On Mon, Oct 6, 2014 at 12:29 PM, Ivan Lagunov <lagi...@gmail.com> wrote:
>>
>>> Hi Ilya,
>>>
>>> Yes, I use a connection pool via Sedna XML:DB API implementation. It's
>>> documented here:
>>>
>>> http://www.cfoster.net/pdf/sedna/xmldb/documentation.pdf#page=10
>>>
>>> Unfortunately, there is no support for active connection timeout. This
>>> could be the cause if there was a query hanging for some reason. Still if
>>> some query hangs forever (that should not as it's in my case) it looks more
>>> like a database bug/deadlock situation.
>>>
>>> I'll try as you suggest next time.
>>>
>>> P.S. Sorry for the second email as I've forgotten to add CC first.
>>>
>>> Best regards,
>>> Ivan Lagunov
>>>
>>> On Fri, Oct 3, 2014 at 10:23 PM, Ilya Taranov <epsi...@socio.msu.ru>
>>> wrote:
>>>
>>>> Ivan, do you have a connection pool? I see one of the se_trn processes
>>>> waiting for socket read.
>>>>
>>>> If you don't have a connection pool, it just might be a client problem.
>>>> I'll take a deeper look anyway, just next time it happens, could you please
>>>> try to shutdown the client thread, if it's even possible.
>>>>
>>>> On Thu, Oct 2, 2014 at 5:30 AM, Ivan Lagunov <lagi...@gmail.com> wrote:
>>>>
>>>>> Hi all,
>>>>>
>>>>> I've been facing major issues with Sedna hanging regularly for the
>>>>> last two weeks (after several months of stable work). I tried to do the
>>>>> complete export-import of data using se_exp but Sedna still hangs
>>>>> regularly. Looking into event.log does not help as queries differ every
>>>>> time. Sometimes it looks like Sedna hangs after some heavy query (with
>>>>> huge
>>>>> response) while in other cases there are no heavy queries to suspect. I've
>>>>> written a script to collect the following logs/dumps you will find in the
>>>>> archives:
>>>>>
>>>>> event.log - the last Sedna event log till the moment DB hangs and is
>>>>> stopped;
>>>>> sedna_procs.txt - list of all Sedna processes (ps -ef);
>>>>> netstat.txt - list of all ESTABLISHED connections;
>>>>> ipcs.txt - results of ipcs -s;
>>>>> <pid>.txt - results of gdb with bt for all threads of a Sedna process
>>>>> (available for all processes at the moment of hanging). This is collected
>>>>> automatically for threads 1-4 of each process.
>>>>>
>>>>> I'm providing two sets of logs/dumps for two separate hanging events:
>>>>>
>>>>> 25th September - seemed to hang without any heavy queries:
>>>>> https://www.dropbox.com/s/2luq53cstc4t3tl/sedna_debug_info1.zip?dl=0
>>>>>
>>>>> 2nd October - hanged after a heavy query:
>>>>> https://www.dropbox.com/s/scmszq9yphyv6xf/sedna_debug_info2.zip?dl=0
>>>>> The heavy query occurs at LOG 02/10/2014 11:10:20
>>>>>
>>>>> If you can find out the cause of the hangings, it'd be the best.
>>>>> Otherwise it'd be nice to hear some recommendations at least on how to
>>>>> avoid these issues/minimize an impact. May be I need to upgrade to some
>>>>> later version but I'm currently using 3.5.615 that is already a night
>>>>> build
>>>>> that we had to upgrade to after some older issue (reported back in
>>>>> March-April of 2013). I'd try some 3.5 or 3.6 version but without knowing
>>>>> the reason there is no guarantee the potential bug is fixed in the later
>>>>> night build.
>>>>>
>>>>> Also it'd be really nice if there was an official release for 3.5 with
>>>>> all the latest bug fixes (I see the last night build is 3.5.627 of 8th May
>>>>> 2013).
>>>>>
>>>>> Best regards,
>>>>> Ivan Lagunov
>>>>>
>>>>>
>>>>> ------------------------------------------------------------------------------
>>>>> Meet PCI DSS 3.0 Compliance Requirements with EventLog Analyzer
>>>>> Achieve PCI DSS 3.0 Compliant Status with Out-of-the-box PCI DSS
>>>>> Reports
>>>>> Are you Audit-Ready for PCI DSS 3.0 Compliance? Download White paper
>>>>> Comply to PCI DSS 3.0 Requirement 10 and 11.5 with EventLog Analyzer
>>>>>
>>>>> http://pubads.g.doubleclick.net/gampad/clk?id=154622311&iu=/4140/ostg.clktrk
>>>>> _______________________________________________
>>>>> Sedna-discussion mailing list
>>>>> Sedna-discussion@lists.sourceforge.net
>>>>> https://lists.sourceforge.net/lists/listinfo/sedna-discussion
>>>>>
>>>>>
>>>>
>>>
>>
>
------------------------------------------------------------------------------
Meet PCI DSS 3.0 Compliance Requirements with EventLog Analyzer
Achieve PCI DSS 3.0 Compliant Status with Out-of-the-box PCI DSS Reports
Are you Audit-Ready for PCI DSS 3.0 Compliance? Download White paper
Comply to PCI DSS 3.0 Requirement 10 and 11.5 with EventLog Analyzer
http://pubads.g.doubleclick.net/gampad/clk?id=154622311&iu=/4140/ostg.clktrk
_______________________________________________
Sedna-discussion mailing list
Sedna-discussion@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sedna-discussion
