On Wed, 20 Jan 2010, Rainer Gerhards wrote:

> David,
>
> thanks for the feedback - and a quick note.
>
> With the new engine, you can do ruleset inclusion (via omruleset[1]). That is
> probably *the* method to handle files that are written to by multiple
> actions. Of course, no need to change if all works in default config. But you
> can gain some extra performance by using buffered mode (for busy files) and
> to use that, you need to have only one action write to each file. This is
> where ruleset inclusion enters the game.

thanks for this, I was thinking about how this could be improved, but this 
looks like it deals with the issue.

on my central box I currently have all the logs written to one file, roll 
that every 5 min, and then at night split this into 45 different files 
based on 100 simplified program names (where I strip out versions so that 
blah-2.3[123] and blah-2.4[123] end up in the same file). I was thinking 
of experimenting to see what happened if I did this in rsyslog instead. 
This is a very good pointer to what I would need to do.

David Lang

> Rainer
>
> [1] http://www.rsyslog.com/doc-omruleset.html
>
>> -----Original Message-----
>> From: [email protected] [mailto:rsyslog-
>> [email protected]] On Behalf Of [email protected]
>> Sent: Wednesday, January 20, 2010 12:26 AM
>> To: rsyslog-users
>> Subject: Re: [rsyslog] rsyslog 5.3.6 (v5-beta) released
>>
>> I've now had v5.3.6 running on my production environemnets since friday
>> with no problems
>>
>> one side effect of the cleanups is that previously when I had multiple
>> filters write to one file I was getting lots of corrupt lines, but the
>> change to have omfile write each transaction rather than just as the
>> buffer filled up seems to have eliminated this (it went from 10's of
>> thousands of corrupted lines/day to none over the weekend and monday,
>> tonight's report will be the acid test to see if it's fully cleaned up)
>>
>> I realize there is still a window for corruption (if two output threads
>> running at the same time both decide they need to write at the same
>> time),
>> but it seems that in practice it's effectively gone.
>>
>> David Lang
>>
>> On Fri, 15 Jan 2010, [email protected] wrote:
>>
>>> On Fri, 15 Jan 2010, Michael Biebl wrote:
>>>
>>>> BTW, I'm actually surprised that you don't encounter those problems
>>>> yourself.
>>>
>>> I'm running 5.3.5 still, I haven't had time to build a new version
>> (hopefully
>>> tomorrow)
>>>
>>> David Lang
>>>
>> _______________________________________________
>> rsyslog mailing list
>> http://lists.adiscon.net/mailman/listinfo/rsyslog
>> http://www.rsyslog.com
> _______________________________________________
> rsyslog mailing list
> http://lists.adiscon.net/mailman/listinfo/rsyslog
> http://www.rsyslog.com
>
_______________________________________________
rsyslog mailing list
http://lists.adiscon.net/mailman/listinfo/rsyslog
http://www.rsyslog.com

Reply via email to