Best HBase monitoring tool

[Giuseppe Reina <g....@gmail.com>]

Hi all,
   we are using HBase in a couple of our clusters and under heavy loads
when something bad happens we do not have any clues on what happens and
why. Ganglia and Nagios are not helping. We created a dashboard with the
ELK stack using the metrics exposed by the region servers and the master
servers but I'm wondering what's the standard in these cases.

What do you use in your production cluster to monitor HBase and why do you
think it's better than the alternatives?


Kind Regards

RE: Best HBase monitoring tool

["Tapper, Gunnar" <gu...@hp.com>]

Hi Giuseppe et al.



Within HP, we've started to add HBase Monitoring into HP Data Services Manager (HP DSM). In 1.4, you'll see a HBase dashboard, consolidated HBase log messages, and HBase health checks. You'll also see a LOT of the HBase performance metrics landed in tables, which you can query and chart using the Query Whiteboard. We’ll build more visualization of that data as we move forward.



1.4 also introduces 3D representation of performance metrics but we've not yet plugged in the HBase-specific metrics into that visualization.



I'm pondering more HBase functions for the next few releases so I'd love input on what you’d want . For example, our IT department would like to be able to get stats such as number of loads and queries processed per hour, per day, etc. but I've not been able to find a way to get that type of data out of HBase. We've been doing work around the slow-query log data but it's not quite meeting the request.



Anyway, you can download and try a free copy of HP DSM from: https://vertica.hpwsportal.com/#/Product/%7b%22productId%22:%221296%22%7d/Show



I'd love to get feedback on what other function you'd need; I'm attaching a screenshot of HBase monitoring hoping that this mail server will allow it. :)



Thank you,



Gunnar



“People don’t know what they want until you show it to them… Our task is to read things that are not yet on the page.” — Steve Jobs



[cid:image003.jpg@01CFA284.435AF150]



-----Original Message-----
From: Alex Newman [mailto:posix4e@gmail.com]
Sent: Tuesday, July 15, 2014 11:27 AM
To: user
Subject: Re: Best HBase monitoring tool



Giuseppe,



As far as metrics go:

If you want a monitoring solution which doesn't require HBase for course grain metrics I highly recommend graphite. It has become a bit of an industry standard and you may find your ops guy already has it setup. OpenTSDB requires HBase as it's datastore but is an extremely powerful tool for monitoring.



Or are you looking at something more than just metrics?









On Mon, Jul 14, 2014 at 1:57 AM, Giuseppe Reina <g....@gmail.com>> wrote:

> Hi all,

>    we are using HBase in a couple of our clusters and under heavy

> loads when something bad happens we do not have any clues on what

> happens and why. Ganglia and Nagios are not helping. We created a

> dashboard with the ELK stack using the metrics exposed by the region

> servers and the master servers but I'm wondering what's the standard in these cases.

>

> What do you use in your production cluster to monitor HBase and why do

> you think it's better than the alternatives?

>

>

> Kind Regards

Re: Best HBase monitoring tool

[Alex Newman <po...@gmail.com>]

Giuseppe,

As far as metrics go:
If you want a monitoring solution which doesn't require HBase for
course grain metrics I highly recommend graphite. It has become a bit
of an industry standard and you may find your ops guy already has it
setup. OpenTSDB requires HBase as it's datastore but is an extremely
powerful tool for monitoring.

Or are you looking at something more than just metrics?




On Mon, Jul 14, 2014 at 1:57 AM, Giuseppe Reina <g....@gmail.com> wrote:
> Hi all,
>    we are using HBase in a couple of our clusters and under heavy loads
> when something bad happens we do not have any clues on what happens and
> why. Ganglia and Nagios are not helping. We created a dashboard with the
> ELK stack using the metrics exposed by the region servers and the master
> servers but I'm wondering what's the standard in these cases.
>
> What do you use in your production cluster to monitor HBase and why do you
> think it's better than the alternatives?
>
>
> Kind Regards