Open GoogleCodeExporter opened 8 years ago
I would suggest that there be a counter for the values from get_rusage(), and
one for
TOTAL time spent doing I/O. Timing information for the last operation is not
terribly useful from my POV, but accumulated time is.
Of course, you should read "Optimizing Oracle Performance" by Cary Millsap to
learn
what proper instrumentation should look like :-)
Original comment by baron.schwartz
on 16 Nov 2009 at 8:53
Baron,
If we have timing information for the last save, that is graphable more-so than
total
timing information, don't you think? incidentally, Cary's book arrived in the
mail
for me this morning;)
Original comment by ryan.a.l...@gmail.com
on 16 Nov 2009 at 8:55
The last save is a random variable depending on when I ask for the information
and
will not tell anything much. The accumulated time is much more indicative and
will
make sensical graphs, just like CPU stats.
Original comment by baron.schwartz
on 16 Nov 2009 at 9:02
I like at least a few of the above proposals, btw there is already a command to
reset the stats of INFO, that is, CONFIG RESETSTAT.
Original comment by anti...@gmail.com
on 23 Aug 2010 at 3:42
I submitted a pull request to track per-command counts and display them in
INFO: https://github.com/antirez/redis/pull/35
I expect that Salvatore's going to be busy with diskstore for a bit, and I'm
not sure if it'll be suitable for 2.2 inclusion.
I'll work on some more of the suggestions here, like total save times. I'm
starting to think that INFO should take subcommands so that it doesn't get too
unwieldy. Similarly, would want to do CONFIG RESETSTAT CMDCOUNTS or CONFIG
RESETSTATS SAVETIMES as the info space grows.
Original comment by mike.sha...@gmail.com
on 3 Jan 2011 at 7:29
The recently released Radish monitoring tool (http://radishapp.com/) provides a
great example of the type of enhanced INFO output that would be desirable.
Specifically read operations/sec, write operations/sec and per-command (or per
command family) operations/sec would be highly useful to anyone using
monitoring of any sort without having to resort to parsing the MONITOR stream.
Another field that would be useful is memory used by both volatile and
persistent keys.
Original comment by deanwba...@gmail.com
on 1 Jun 2011 at 9:16
Original issue reported on code.google.com by
ryan.a.l...@gmail.com
on 16 Nov 2009 at 6:37