Re: anyway to throttle nodetool repair?

[Yan Chunlu <sp...@gmail.com>]

so how about disk io?  is there anyway to use ionice to control it?

I have tried to adjust the priority by "ionice -c3 -p [cassandra pid].
 seems not working...

On Wed, Sep 28, 2011 at 12:02 AM, Peter Schuller <
peter.schuller@infidyne.com> wrote:

> > I saw the ticket about compaction throttling, just wonder is that
> necessary
> > to add an option or is there anyway to do repair throttling?
> > every time I run nodetool repair, it uses all disk io and the server load
> > goes up quickly, just wonder is there anyway to make it smoother.
>
> The validating compaction that is part of repair is subject to
> compaction throttling.
>
> The streaming of sstables afterwards is not however. In 1.0 there is
> thottling of streaming:
> https://issues.apache.org/jira/browse/CASSANDRA-3080
>
> --
> / Peter Schuller (@scode on twitter)
>

Re: anyway to throttle nodetool repair?

[Peter Schuller <pe...@infidyne.com>]

> earlier: http://cassandra-user-incubator-apache-org.3065146.n2.nabble.com/how-does-compaction-throughput-kb-per-sec-affect-disk-io-td6831711.html
> might not directly throttle the disk I/O?

Again: Compaction throttling will throttle compaction, which affects
both CPU and I/O for fundamental reasons.

In 1.0 (not earlier) there is throttling of internode streaming, with
similar results.

Without 1.0 it is expected that internode throttling is not throttled.

Also note that compaction and streaming throttling would not
necessarily work well with ionice to begin with; for one thing
depending on situation you may not want to be using the CFQ I/O
scheduler in the kernel to begin with (required for ionice to work);
and for another you may want to throttle the process lower than
available idle I/O bandwidth because of the effects on page cache
eviction.

-- 
/ Peter Schuller (@scode on twitter)

Re: anyway to throttle nodetool repair?

[Yan Chunlu <sp...@gmail.com>]

as I asked earlier:
http://cassandra-user-incubator-apache-org.3065146.n2.nabble.com/how-does-compaction-throughput-kb-per-sec-affect-disk-io-td6831711.html

might not directly throttle the disk I/O?

it would be easy if ionice could work with cassandra. not sure it is because
of jvm or something else, ionice works perfectly with "cp" or "mv" to make
the disk I/O very smooth.

thanks!

On Tue, Oct 11, 2011 at 3:15 PM, Peter Schuller <peter.schuller@infidyne.com
> wrote:

> > so how about disk io?  is there anyway to use ionice to control it?
> > I have tried to adjust the priority by "ionice -c3 -p [cassandra pid].
> >  seems not working...
>
> Compaction throttling (and in 1.0 internode streaming throttling) both
> address disk I/O.
>
> --
> / Peter Schuller (@scode on twitter)
>

Re: anyway to throttle nodetool repair?

[Peter Schuller <pe...@infidyne.com>]

> so how about disk io?  is there anyway to use ionice to control it?
> I have tried to adjust the priority by "ionice -c3 -p [cassandra pid].
>  seems not working...

Compaction throttling (and in 1.0 internode streaming throttling) both
address disk I/O.

-- 
/ Peter Schuller (@scode on twitter)

Re: anyway to throttle nodetool repair?

[Yan Chunlu <sp...@gmail.com>]

I am using commodity hardware so even minor compact make disk io goes 100%
and server load get very high

On Tue, Oct 11, 2011 at 11:19 AM, Yan Chunlu <sp...@gmail.com> wrote:

> so how about disk io?  is there anyway to use ionice to control it?
>
> I have tried to adjust the priority by "ionice -c3 -p [cassandra pid].
>  seems not working...
>
>
> On Wed, Sep 28, 2011 at 12:02 AM, Peter Schuller <
> peter.schuller@infidyne.com> wrote:
>
>> > I saw the ticket about compaction throttling, just wonder is that
>> necessary
>> > to add an option or is there anyway to do repair throttling?
>> > every time I run nodetool repair, it uses all disk io and the server
>> load
>> > goes up quickly, just wonder is there anyway to make it smoother.
>>
>> The validating compaction that is part of repair is subject to
>> compaction throttling.
>>
>> The streaming of sstables afterwards is not however. In 1.0 there is
>> thottling of streaming:
>> https://issues.apache.org/jira/browse/CASSANDRA-3080
>>
>> --
>> / Peter Schuller (@scode on twitter)
>>
>
>