[check_postgres] threshold measurement unit for --action=hot_standby_delay

[karfunckel-checkpostgres at yahoo.de]

Hallo,
thanks for your fast reply. Bytes is indeed not an easy unit for this case. If I upload a file into DB there would be only one row lag for a few "seconds". This is more than under standable. But at the same time a lot of "bytes" lag, which are possibly the same when both servers would be oout of sync for a long time with only small but important changes. 
 
Well, lets wait for 9.1... :-)
Thanks,
Karfunckel

--- Guillaume Lelarge <guillaume at lelarge.info> schrieb am Fr, 22.7.2011:

> Von: Guillaume Lelarge <guillaume at lelarge.info>
> Betreff: Re: [check_postgres] threshold measurement unit for --action=hot_standby_delay
> An: karfunckel-checkpostgres at yahoo.de
> CC: check_postgres at bucardo.org
> Datum: Freitag, 22. Juli, 2011 22:14 Uhr
> On Fri, 2011-07-22 at 08:33 +0100, karfunckel-checkpostgres at yahoo.de
> wrote:
> > Hallo, 
> > can anyone explain me the meaning of the values in
> --warning/--critical when using it with
> --action=hot_standby_delay? What is the "measurement unit"?
> Are the results in seconds, MB, ...?
> > 
> 
> Bytes.
> 
> > In the Documentation I only could find "The valid
> options for this option depends on the action used" and
> "...the delta between xlog...".
> > The knowing of the "measurement unit" is important for
> the decision what values should be used for the threshold. 
> > 
> > Example: The lag of 1 second oder 1 file differenz
> would be more than ok for me, but the lag of 1 minute would
> be not ok.
> > 
> > And the results I receive are so different that I dont
> kwon how to interpret them:
> > First result: time=0.01 replay_delay=86040;10;20
> receive_delay=74800;10;20
> > a second later: time=0.01 replay_delay=0;10;20
> receive_delay=0;10;20
> > 
> 
> First number is the number of bytes waiting to be
> replayed/received on
> the slave, second number is the warning value, third number
> is the
> critical value.
> 
> Number of bytes is not an easy number to work with. At
> least 9.1 will
> have a more interesting unit for the lag (time unit).
> 
> 
> -- 
> Guillaume
>   http://blog.guillaume.lelarge.info
>   http://www.dalibo.com
> 
>