pcp
[Top] [All Lists]

Re: Possible pmmgr issue?

To: "Frank Ch. Eigler" <fche@xxxxxxxxxx>, Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Subject: Re: Possible pmmgr issue?
From: Nathan Scott <nathans@xxxxxxxxxx>
Date: Thu, 13 Feb 2014 20:29:26 -0500 (EST)
Cc: pcp developers <pcp@xxxxxxxxxxx>
Delivered-to: pcp@xxxxxxxxxxx
In-reply-to: <20140214002744.GI11820@xxxxxxxxxx>
References: <1952377955.6159460.1392281163287.JavaMail.zimbra@xxxxxxxxxx> <1444843200.6174759.1392282098405.JavaMail.zimbra@xxxxxxxxxx> <20140213125811.GG11820@xxxxxxxxxx> <1040057732.6827059.1392333686610.JavaMail.zimbra@xxxxxxxxxx> <2147120813.6837227.1392334988862.JavaMail.zimbra@xxxxxxxxxx> <20140214002744.GI11820@xxxxxxxxxx>
Reply-to: Nathan Scott <nathans@xxxxxxxxxx>
Thread-index: XH9tRX8F2NWQYfXK4eQMMZiMHDh0zQ==
Thread-topic: Possible pmmgr issue?

----- Original Message -----
> Hi -
> 
> > File:etc-pcp-pmlogger.tgz   8 KB
> > File:etc-pcp-pmmgr.tgz              1 KB
> > File:var-log-pcp-pmlogger.tgz       7501 KB
> > File:var-log-pcp-pmmgr.tgz  115240 KB
> 
> The two pmlogger.log files identify the apples & oranges directly.
> 
> pmmgr's pmlogconf:
> "Auto-generated by pmlogconf on:  Friday 14 February 09:11:41 EST 2014"
> "254 metrics ... logged every 60 sec: 44340 bytes or 60.89 Mbytes/day"
> 
> other pmlogconf:
> "Auto-generated by pmlogconf on:  Mon Jul 22 12:48:51 EST 2013"
> "100 metrics ... logged every 60 sec: 4496 bytes or 6.17 Mbytes/day"
> 

Yep very good point, interesting - those are not default log sizes I'd
expected (for either!).  Ohhh, its found and enabled postgres logging!
And the other config is really out-dated.

Re-creating the pmlogger log conf, and same config generated now.  I
should revisit the frequency with which pmlogger_daily auto-generates,
not clear that's working as planned.

OOC, any thoughts on the question of merging logs during the middle of
the day?  Restarting pmcd appears to be an expensive operation with a
pmmgr in play without some strategy there?  (I can restart pmcd/pmmgr
a whole lot more quickly than those logs can be merged)


And, yeah, wow - hmm, thats alot of logged data now, I wonder if we're
going overboard with some of the logging configs.  :|

Yet another issue - the daily logs should be compressed after a few
days.  It seems the daily script has stopped some of its functions?
Rotation is happening but cull &| compress appears to not be anymore.
Oh here we go... running daily script by hand, verbose mode reveals:

pmlogger_merge: Warning: archive "20140123.09.55" is empty and will be skipped
Input archives to be merged:
        20140123.00.10
        20140123.16.25
pmlogextract: Error: __pmLogRead[log 20140123.00.10]: Corrupted record in a PCP 
archive log
pmlogextract: Error occurred at byte offset 2690960 into a file of 2691072 
bytes.
The last record, and the remainder of this file will not be extracted.
Archive "20140123" not created.
pmlogger_merge: Directory: /var/log/pcp/pmlogger/verge
pmlogger_merge: Failed: pmlogextract  20140123.00.10 20140123.16.25 20140123
pmlogger_merge: Trying to continue, although output archive may be corrupted.
Output archive files:
pmlogger_merge: Error: file "20140123.meta" not created
pmlogger_merge: Error: file "20140123.index" not created
pmlogger_merge: Error: file "20140123.0" not created
Merged output archive 20140123 ...
pmdumplog: Cannot open archive "20140123": No such file or directory
Skip culling and compression ...


Hmm, file corruption.  Oh man, that never happens! ;)  (/me ducks and
runs).  And an empty log file in the mix too.

In terms of the cull/compress though, the last line is key ... it
seems overly drastic - but appears to be working as planned, Ken?
Self-correction would be good, once the problem archives have all
scrolled past their use-by date?

BTW, the option to compress is also missing in pmmgr, I realise now;
another one for that list.  Not easily implementable either, it would
seem, with one merge log - probably need libpcp to be able to do this
directly - bump log format, add code to do in-line compression...?
Or "hack" it, and inflate/deflate around the merges...?  That defeats
the point of having the one continuous log handy though, I think, so
it really would have to be in libpcp.  (a good feature anyway, IMO)

cheers.

--
Nathan

<Prev in Thread] Current Thread [Next in Thread>