Correlation between GETS, PUTS and system metrics to identify performance degradation

[rohit kumar <ro...@gmail.com>]

Hi,

I have a question regarding Hbase metrics.
There can be a case when RAM consumption on a hbase node could be very high.
Could this be possibly due to avg_time taken by PUTS/ GETS requests is very
long ?

For example:  Can "compaction queue" metrics be a factor for such time
taking operations?
If yes, then what are other such metrics which could help me to analyze the
impact on GETS/ PUTS operations?

Please help me to understand this correlation because I need to a group of
graphs from hbase metrics to analyze system resource requirements and
predict them in future.


Regards,
Rohit

Re: Correlation between GETS, PUTS and system metrics to identify performance degradation

[Ted Yu <yu...@gmail.com>]

Rohit:
Below are some metrics you should monitor:

compactionQueueLength

flushQueueLength

numCallsInGeneralQueue

numCallsInPriorityQueue

QueueCallTime_mean

QueueCallTime_75th_percentile

QueueCallTime_99th_percentile


Cheers


On Thu, Nov 13, 2014 at 5:05 AM, rohit kumar <ro...@gmail.com> wrote:

> Hi,
>
> I have a question regarding Hbase metrics.
> There can be a case when RAM consumption on a hbase node could be very
> high.
> Could this be possibly due to avg_time taken by PUTS/ GETS requests is very
> long ?
>
> For example:  Can "compaction queue" metrics be a factor for such time
> taking operations?
> If yes, then what are other such metrics which could help me to analyze the
> impact on GETS/ PUTS operations?
>
> Please help me to understand this correlation because I need to a group of
> graphs from hbase metrics to analyze system resource requirements and
> predict them in future.
>
>
> Regards,
> Rohit
>