hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ted Dunning <tdunn...@maprtech.com>
Subject Re: HBase jmx stats
Date Tue, 24 May 2011 17:21:56 GMT
In case anybody wants estimates of medians, Mahout has some easily
extractable code to compute medians and first and third quartiles without
keeping lots of data around.  As a side effect, it computes averages and
standard deviations as well.

I don't think that such a small thing as this warrants an entire dependency
and would be happy to extract the code in question if anybody wants this.
 Just point me at a JIRA where I should stick it.

On Tue, May 24, 2011 at 10:17 AM, Stack <stack@duboce.net> wrote:

> Tim:
>
> You should get a copy of Lars' book.  Then you'd know what the below
> are (smile).
>
> Quoting:
>
> "A commonly used metric in HBase is called time varying rate, which
> not only tracks the number of events, but also how long each event
> took to complete.  A TVR exposes four values.   [Below] shows the
> values and their meanings.
>
> Value Name             Short        Description
> Number Operations  NumOps    The actual number of events since the last
> poll
> Minimum Time        MniTime    The shortest time reported for an event
> to complete
> Maximum Time       MaxTime   The longest time reported for an event to
> complete
> Average Time         AvgTime    The average time for completing
> events...sum of times per event divided by number of events"
>
> Our metrics need a bit of a revamp so file issues on stuff that seems off
> Tim.
>
> St.Ack
>
> On Mon, May 23, 2011 at 10:39 AM, Tim Sell <trsell@gmail.com> wrote:
> > I am collecting hbase stats from jmx.
> >
> > Here's an example of the compaction stats from one of the region
> servers...
> >
> > compactionQueueSize     0
> > compactionSizeAvgTime   19349729
> > compactionSizeMaxTime   566512933
> > compactionSizeMinTime   0
> > compactionSizeNumOps    8
> > compactionTimeAvgTime   83
> > compactionTimeMaxTime   1300
> > compactionTimeMinTime   0
> > compactionTimeNumOps    8
> >
> > What does NumOps of compactionSize and compactionTime mean? is that
> > the total number of compactions ever?
> >
> > Also, having a time suffix for compactionSize seems confusing. It
> > seems an artifact of this:
> >
> > protected final PersistentMetricsTimeVaryingRate compactionTime =
> >    new PersistentMetricsTimeVaryingRate("compactionTime", registry);
> > protected final PersistentMetricsTimeVaryingRate compactionSize =
> >    new PersistentMetricsTimeVaryingRate("compactionSize", registry);
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message