[pgpool-general: 4457] Re: Pgpool - connection hangs in DISCARD ALL

Tue Feb 16 19:44:53 JST 2016

Hi

Many thanks for the reply and a good news that you are not getting
stuck connection issue after the patch.

Thanks
Best regards
Muhammad Usama

On Fri, Feb 12, 2016 at 9:45 PM, Paweł Ufnalewski <archon at foap.com> wrote:
> Hmm it looks like it's fine now. Right now I only see these in log:
>
> 2016-02-12 17:28:12: pid 8838: LOG:  child process with pid: 27299 exits
> with status 256
> 2016-02-12 17:28:12: pid 8838: LOG:  fork a new child process with pid: 6140
> 2016-02-12 17:30:42: pid 8838: LOG:  child process with pid: 5571 exits with
> status 512
> 2016-02-12 17:30:42: pid 8838: LOG:  fork a new child process with pid: 6720
> 2016-02-12 17:30:43: pid 8838: LOG:  child process with pid: 4444 exits with
> status 512
> 2016-02-12 17:30:43: pid 8838: LOG:  fork a new child process with pid: 6751
> 2016-02-12 17:35:42: pid 8838: LOG:  child process with pid: 6140 exits with
> status 512
> 2016-02-12 17:35:42: pid 8838: LOG:  fork a new child process with pid: 7868
> 2016-02-12 17:40:42: pid 8838: LOG:  child process with pid: 6751 exits with
> status 512
> 2016-02-12 17:40:42: pid 8838: LOG:  fork a new child process with pid: 9018
> 2016-02-12 17:40:42: pid 8838: LOG:  child process with pid: 6720 exits with
> status 512
> 2016-02-12 17:40:42: pid 8838: LOG:  fork a new child process with pid: 9019
>
> Thank you!
>
> Best regards,
> Paweł Ufnalewski
> Infrastructure Architect at Foap.com
>
> W dniu 2016-02-09 o 14:02, Muhammad Usama pisze:
>
>> Hi
>>
>> Many thanks for sharing the pgpool.log, The log shared by you does
>> contains some error messages "ERROR: unable to to flush data to
>> frontend" that have the potential to cause the stuck connection
>> Can you please try out the attached patch if it fix the problem. I am
>> attaching the patches for both 3_5 and 3_4 branches, please use the
>> respective patch as per your setup. Hopefully this should fix the
>> stuck issue.
>>
>> Kind regards
>> Muhammad Usama
>>
>>
>>
>> On Mon, Feb 8, 2016 at 8:49 PM, Paweł Ufnalewski <archon at foap.com> wrote:
>>>
>>> Hi,
>>>
>>>      It looks like it hangs in this places (see attachment). Problem is,
>>> that
>>> developer responsible for app has changed something in code, so
>>> connections
>>> now closes properly from client side (before I got a lot of these errors:
>>>
>>> 2016-02-08 09:33:39: pid 8472: ERROR:  unable to read data from frontend
>>> 2016-02-08 09:33:39: pid 8472: DETAIL:  EOF encountered with frontend)
>>> .
>>>
>>> Best regards,
>>> Paweł Ufnalewski
>>> Infrastructure Architect at Foap.com
>>>
>>> W dniu 2016-02-08 o 09:00, Muhammad Usama pisze:
>>>
>>> Hi
>>>
>>> Thanks in advance for the help. If you could share the pgpool-II log
>>> when the stuck connection happens that would help us in identifiny and
>>> rectifing the problem.
>>>
>>> Thanks
>>> Best regards
>>> Muhammad Usama
>>>
>>>
>>> On Mon, Feb 8, 2016 at 11:36 AM, Paweł Ufnalewski <archon at foap.com>
>>> wrote:
>>>
>>> Hi,
>>>
>>>      just to let you know - I'm having the same problem with 3.4.4
>>> version
>>> (DISCARD ALL appears slower than in 3.4.3 I think, but it still does).
>>> How
>>> can I help to fix this problem?
>>>
>>> Best regards,
>>> Paweł Ufnalewski
>>> Infrastructure Architect at Foap.com
>>>
>>> W dniu 2016-02-01 o 08:44, Muhammad Usama pisze:
>>>
>>> Hi Gerhard
>>>
>>> Many thanks for testing and pointing this out. It's unfortunate that you
>>> are
>>> still getting the stuck connection issue. If it is possible can you
>>> please
>>> share the pgpool-II log for the time when this stuck connection issue
>>> happens. I am more interested in seeing which exact error message that
>>> caused the child process to jump to error handler from where the child
>>> process proceeded to send the DISCARD ALL to backend and eventually got
>>> stuck. Since after many tries we are not able to reproduce this issue, so
>>> log would be really helpful in understanding and fixing the problem.
>>>
>>> Best regards
>>> Muhammad Usama
>>>
>>>
>>> On Sun, Jan 31, 2016 at 9:33 PM, Gerhard Wiesinger <lists at wiesinger.com>
>>> wrote:
>>>
>>> On 28.01.2016 01:10, Tatsuo Ishii wrote:
>>>
>>> On 21.01.2016 20:52, Muhammad Usama wrote:
>>>
>>> Hi
>>>
>>> I am looking into this issue. and unfortunately like Ishii-San I am
>>> also not able to reproduce it. But I found one issue in 3.4 that might
>>> cause the problem. Can you please try the attached patch if it solves
>>> the problem. Also, if the problem still persists, it would be really
>>> helpful if you could share the pgpool-II log.
>>>
>>> I looked at the patch but it includes only logging changes and no
>>> functional changes. Therefore I didn't test it. Do you expect and
>>> behavioral changes to fix it, and why?
>>>
>>> elog() is not only a logging function, but also it plays very
>>> important role including exception handling and error treatments in
>>> pgpool-II. If you are familiar with PostgreSQL internals, you may
>>> notice it (elog() was imported from PostgreSQL source tree).
>>>
>>> Tried version 3.5.0 where the patch is included. Still not working. See
>>> backtrace below.
>>>
>>> Reverting to 3.3.7 which works perfectly.
>>>
>>> Ciao,
>>> Gerhard
>>>
>>> (gdb) back
>>> #0  0x00007fd87fdb6d43 in __select_nocancel () from /lib64/libc.so.6
>>> #1  0x0000564471af16a1 in pool_check_fd (cp=cp at entry=0x564473dfa610) at
>>> protocol/pool_process_query.c:635
>>> #2  0x0000564471af1976 in pool_check_fd (cp=cp at entry=0x564473dfa610) at
>>> protocol/pool_process_query.c:657
>>> #3  0x0000564471b1f67b in pool_read (cp=0x564473dfa610,
>>> buf=buf at entry=0x7ffc1d71bf97, len=len at entry=1) at utils/pool_stream.c:162
>>> #4  0x0000564471af8e6e in read_kind_from_backend
>>> (frontend=frontend at entry=0x564473df3e60,
>>> backend=backend at entry=0x564473df2e00,
>>>      decided_kind=decided_kind at entry=0x7ffc1d71c397 "E") at
>>> protocol/pool_process_query.c:3234
>>> #5  0x0000564471affdc9 in ProcessBackendResponse
>>> (frontend=frontend at entry=0x564473df3e60,
>>> backend=backend at entry=0x564473df2e00, state=state at entry=0x7ffc1d71c41c,
>>>      num_fields=num_fields at entry=0x7ffc1d71c41a) at
>>> protocol/pool_proto_modules.c:2356
>>> #6  0x0000564471af5b15 in pool_process_query (frontend=0x564473df3e60,
>>> backend=0x564473df2e00, reset_request=reset_request at entry=1) at
>>> protocol/pool_process_query.c:302
>>> #7  0x0000564471aed98c in backend_cleanup (backend=<optimized out>,
>>> frontend_invalid=frontend_invalid at entry=0 '\000', frontend=0x564471e09e40
>>> <child_frontend>)
>>>      at protocol/child.c:437
>>> #8  0x0000564471af0637 in do_child (fds=fds at entry=0x564473dee030) at
>>> protocol/child.c:234
>>> #9  0x0000564471ace107 in fork_a_child (fds=0x564473dee030, id=8) at
>>> main/pgpool_main.c:678
>>> #10 0x0000564471aceb6d in reaper () at main/pgpool_main.c:2254
>>> #11 0x0000564471ad322b in PgpoolMain (discard_status=<optimized out>,
>>> clear_memcache_oidmaps=<optimized out>) at main/pgpool_main.c:429
>>> #12 0x0000564471acc7b1 in main (argc=<optimized out>,
>>> argv=0x7ffc1d7219e8)
>>> at main/main.c:310
>>>
>>> #1  0x0000564471af16a1 in pool_check_fd (cp=cp at entry=0x564473dfa610) at
>>> protocol/pool_process_query.c:635
>>> 635                     fds = select(fd+1, &readmask, NULL, &exceptmask,
>>> timeoutp);
>>>
>>> (gdb) print fd
>>> $1 = 8
>>> (gdb) print readmask
>>> $2 = {fds_bits = {256, 0 <repeats 15 times>}}
>>> (gdb) print exceptmask
>>> $3 = {fds_bits = {256, 0 <repeats 15 times>}}
>>> (gdb) print timeoutp
>>> $4 = (struct timeval *) 0x0
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> pgpool-general mailing list
>>> pgpool-general at pgpool.net
>>> http://www.pgpool.net/mailman/listinfo/pgpool-general
>>>
>>>
>>>
>