[syslog-ng] Disk buffer file truncation issue
Raghunath Adhyapak
funduraghu at gmail.com
Sun Jun 14 17:01:18 UTC 2020
Hi,
Here is some more detail on the issue:
Buffer file on syslog-ng machine
root at mymachine:/etc/syslog-ng/disk-buffer# ls -l
total 61580512
-rw------- 1 root root 4096 Jun 14 16:57 syslog-ng-00000.rqf
-rw------- 1 root root 4096 Jun 8 13:00 syslog-ng-00001.rqf
-rw------- 1 root root 63058421272 Jun 14 16:57 syslog-ng-00002.rqf
-rw------- 1 root root 4096 Jun 14 16:56 syslog-ng-00003.rqf
When I run syslog-ng in debug mode, I see the following lines related to
buffer
[2020-06-14T16:43:32.085616] Reliable disk-buffer state loaded;
filename='/etc/syslog-ng/disk-buffer/syslog-ng-00000.rqf',
queue_length='0', size='0'
[2020-06-14T16:43:32.085690] Reliable disk-buffer internal state;
filename='/etc/syslog-ng/disk-buffer/syslog-ng-00000.rqf',
backlog_head='4096', read_head='4096', write_head='4096', backlog_len='0'
[2020-06-14T16:43:32.086555] WARNING: window sizing for tcp sources were
changed in syslog-ng 3.3, the configuration value was divided by the value
of max-connections(). The result was too small, clamping to value of
min_iw_size_per_reader. Ensure you have a proper log_fifo_size setting to
avoid message loss.; orig_log_iw_size='2', new_log_iw_size='100',
min_iw_size_per_reader='100', min_log_fifo_size='50000'
[2020-06-14T16:43:32.086677] Accepting connections; addr='AF_INET(
0.0.0.0:514)'
[2020-06-14T16:43:32.086802] Reliable disk-buffer state loaded;
filename='/etc/syslog-ng/disk-buffer/syslog-ng-00002.rqf',
queue_length='14255828', size='20391794080'
*[2020-06-14T16:43:32.086871] Reliable disk-buffer internal state;
filename='/etc/syslog-ng/disk-buffer/syslog-ng-00002.rqf',
backlog_head='42534306360', read_head='42534307736',
write_head='62926101816', backlog_len='1'*
[2020-06-14T16:43:32.087359] Seeking the journal to the last cursor
position;
cursor='s=3f7a52af1f26424db365ccc75b2cf79d;i=192f;b=6921d51da51b41d5a26d8674232c77d9;m=75ef33f5cc;t=5a80dfded040d;x=b88b960795be54d2'
[2020-06-14T16:43:32.087572] Reliable disk-buffer state loaded;
filename='/etc/syslog-ng/disk-buffer/syslog-ng-00003.rqf',
queue_length='0', size='0'
[2020-06-14T16:43:32.087637] Reliable disk-buffer internal state;
filename='/etc/syslog-ng/disk-buffer/syslog-ng-00003.rqf',
backlog_head='4096', read_head='4096', write_head='4096', backlog_len='0'
I can see that logs are getting sent out to the destination server.
However, the buffer is still piling up.
I am not sure how to get the buffer cleared.
Please advise.
Thanks
Raghu
On Sat, Jun 13, 2020 at 9:09 PM Raghunath Adhyapak <funduraghu at gmail.com>
wrote:
> Hi, I don't have an exact way to reproduce it. However, this is what we
> observe:
> - The buffer file keeps increasing in size.
> - We keep getting syslog messages and there doesn't seem to be any issue
> with the service.
> - When we check the buffer size it grows into gigabytes.
> - At this point, in one case we saw the buffer file size was around 4GB.
> We restarted the service and after the restart, boom, the buffer files got
> cleared immediately (size of buffer file was 4k)
> - Next time when this happened, we checked the contents of the buffer file
> and investigated whether the events were received or not. We found that we
> had received all the events.
>
> Therefore we believe that there was a file truncation issue and restart
> helped to clear it.
>
> Let me know if you need any further details and also on how to prevent
> this from happening
>
> Thanks
> Raghu
>
>
>
> On Tue, Jun 9, 2020, 12:58 Nagy Gábor <gabor.hl at gmail.com> wrote:
>
>> Hi,
>>
>> Can you describe the issue in detail, please? Could you share the
>> reproduction steps?
>>
>> You say there is an issue with disk-buffer truncating after it's emptied,
>> and that is resolved when syslog-ng restart?
>> How do you check if the queue is empty?
>>
>> Regards,
>> Gabor
>>
>>
>>
>>
>>
>> Raghunath Adhyapak <funduraghu at gmail.com> ezt írta (időpont: 2020. jún.
>> 9., K, 0:08):
>>
>>> Correcting subject
>>>
>>> We did some more troubleshooting on this and we found that all logs in
>>> buffer were indeed sent out and that syslog-ng was facing issues in
>>> truncating the file.
>>> This issue got fixed with restart.
>>> However, we are observing that this issue is happening too often.
>>> Would anyone help me understand why this could be happening?
>>>
>>> Thanks
>>> Raghu
>>>
>>> On Mon, May 11, 2020, 19:34 Fabien Wernli <wernli at in2p3.fr> wrote:
>>>
>>>> Hi,
>>>>
>>>> On Mon, May 11, 2020 at 12:23:27PM +0000, László Várady (lvarady) wrote:
>>>> > > 2. Why couldn't syslog-ng resume operations after partition was
>>>> freed up and destination was available?
>>>> >
>>>> > That might be a bug. Once a destination becomes available (set
>>>> time-reopen() to a lower value to check more frequently), syslog-ng should
>>>> send messages out from the disk buffer.
>>>> > Could you reproduce this issue and share the reproduction steps?
>>>>
>>>> I remember having a lot of corrupt disk buffers when disk was full.
>>>> In my case, they caused a segfault on startup, maybe that got "fixed"
>>>> by a
>>>> deletion instead?
>>>>
>>>>
>>>> ______________________________________________________________________________
>>>> Member info: https://lists.balabit.hu/mailman/listinfo/syslog-ng
>>>> Documentation:
>>>> http://www.balabit.com/support/documentation/?product=syslog-ng
>>>> FAQ: http://www.balabit.com/wiki/syslog-ng-faq
>>>>
>>>>
>>> ______________________________________________________________________________
>>> Member info: https://lists.balabit.hu/mailman/listinfo/syslog-ng
>>> Documentation:
>>> http://www.balabit.com/support/documentation/?product=syslog-ng
>>> FAQ: http://www.balabit.com/wiki/syslog-ng-faq
>>>
>>>
>> ______________________________________________________________________________
>> Member info: https://lists.balabit.hu/mailman/listinfo/syslog-ng
>> Documentation:
>> http://www.balabit.com/support/documentation/?product=syslog-ng
>> FAQ: http://www.balabit.com/wiki/syslog-ng-faq
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.balabit.hu/pipermail/syslog-ng/attachments/20200614/0ed08147/attachment.html>
More information about the syslog-ng
mailing list