Hi,

I have a table, with about 200 000 raws, keys is string[64] (ish) and
value is string[512].

It's splitted over 16 regions located on 7 regionservers.

So it's not a big table, and there is a lot of horsepower behind it.

I asked a major_compaction few hours ago. Let's say, about 5 hours
ago. and it's still compacting! But all servers activities seems to be
null. CPU usage is almost 0.

There is nothing on the master logs.

How can I see what's going on? Is there a way to see the compaction queue?

JM

You're on hbase 0.94.3 , right ?

When was the last time major compaction ran ?

Compaction is region server activity so you should be able to find some clue in region server log.

Cheers

On Jan 1, 2013, at 11:42 AM, Jean-Marc Spaggiari <[EMAIL PROTECTED]> wrote:

> Hi,
>
> I have a table, with about 200 000 raws, keys is string[64] (ish) and
> value is string[512].
>
> It's splitted over 16 regions located on 7 regionservers.
>
> So it's not a big table, and there is a lot of horsepower behind it.
>
> I asked a major_compaction few hours ago. Let's say, about 5 hours
> ago. and it's still compacting! But all servers activities seems to be
> null. CPU usage is almost 0.
>
> There is nothing on the master logs.
>
> How can I see what's going on? Is there a way to see the compaction queue?
>
> JM

Yes, I'm running on 0.94.3. The last major compaction ran yesterday.
It's almost daily. That's why I was surprise it took so long. I mean,
I'm only compacting regions who moved. So it should be pretty quick.
But was not the case. It took about 6h to complete. Strange. Maybe
something went wrong when I stopped/started hbase.

also, there was almost no activity on the network nor on the CPUs. I
will have to add disks monitoring on Ganglia to see if I was limited
by IOs...

I looked at the regions logs and everything was fine. It was showing
some compaction information every few seconds.

JM

2013/1/1, Ted <[EMAIL PROTECTED]>:
> You're on hbase 0.94.3 , right ?
>
> When was the last time major compaction ran ?
>
> Compaction is region server activity so you should be able to find some clue
> in region server log.
>
> Cheers
>
> On Jan 1, 2013, at 11:42 AM, Jean-Marc Spaggiari <[EMAIL PROTECTED]>
> wrote:
>
>> Hi,
>>
>> I have a table, with about 200 000 raws, keys is string[64] (ish) and
>> value is string[512].
>>
>> It's splitted over 16 regions located on 7 regionservers.
>>
>> So it's not a big table, and there is a lot of horsepower behind it.
>>
>> I asked a major_compaction few hours ago. Let's say, about 5 hours
>> ago. and it's still compacting! But all servers activities seems to be
>> null. CPU usage is almost 0.
>>
>> There is nothing on the master logs.
>>
>> How can I see what's going on? Is there a way to see the compaction
>> queue?
>>
>> JM
>

bq. It took about 6h to complete.

If the above behavior is reproducible, we should investigate more deeply.

Thanks for sharing.

On Tue, Jan 1, 2013 at 6:14 PM, Jean-Marc Spaggiari <[EMAIL PROTECTED]
> wrote:

> Yes, I'm running on 0.94.3. The last major compaction ran yesterday.
> It's almost daily. That's why I was surprise it took so long. I mean,
> I'm only compacting regions who moved. So it should be pretty quick.
> But was not the case. It took about 6h to complete. Strange. Maybe
> something went wrong when I stopped/started hbase.
>
> also, there was almost no activity on the network nor on the CPUs. I
> will have to add disks monitoring on Ganglia to see if I was limited
> by IOs...
>
> I looked at the regions logs and everything was fine. It was showing
> some compaction information every few seconds.
>
> JM
>
> 2013/1/1, Ted <[EMAIL PROTECTED]>:
> > You're on hbase 0.94.3 , right ?
> >
> > When was the last time major compaction ran ?
> >
> > Compaction is region server activity so you should be able to find some
> clue
> > in region server log.
> >
> > Cheers
> >
> > On Jan 1, 2013, at 11:42 AM, Jean-Marc Spaggiari <
> [EMAIL PROTECTED]>
> > wrote:
> >
> >> Hi,
> >>
> >> I have a table, with about 200 000 raws, keys is string[64] (ish) and
> >> value is string[512].
> >>
> >> It's splitted over 16 regions located on 7 regionservers.
> >>
> >> So it's not a big table, and there is a lot of horsepower behind it.
> >>
> >> I asked a major_compaction few hours ago. Let's say, about 5 hours
> >> ago. and it's still compacting! But all servers activities seems to be
> >> null. CPU usage is almost 0.
> >>
> >> There is nothing on the master logs.
> >>
> >> How can I see what's going on? Is there a way to see the compaction
> >> queue?
> >>
> >> JM
> >
>

Do you have 100% data locality ? How is the network trafic ?
On Wed, Jan 2, 2013 at 7:07 AM, Ted Yu <[EMAIL PROTECTED]> wrote:

> bq. It took about 6h to complete.
>
> If the above behavior is reproducible, we should investigate more deeply.
>
> Thanks for sharing.
>
> On Tue, Jan 1, 2013 at 6:14 PM, Jean-Marc Spaggiari <
> [EMAIL PROTECTED]
> > wrote:
>
> > Yes, I'm running on 0.94.3. The last major compaction ran yesterday.
> > It's almost daily. That's why I was surprise it took so long. I mean,
> > I'm only compacting regions who moved. So it should be pretty quick.
> > But was not the case. It took about 6h to complete. Strange. Maybe
> > something went wrong when I stopped/started hbase.
> >
> > also, there was almost no activity on the network nor on the CPUs. I
> > will have to add disks monitoring on Ganglia to see if I was limited
> > by IOs...
> >
> > I looked at the regions logs and everything was fine. It was showing
> > some compaction information every few seconds.
> >
> > JM
> >
> > 2013/1/1, Ted <[EMAIL PROTECTED]>:
> > > You're on hbase 0.94.3 , right ?
> > >
> > > When was the last time major compaction ran ?
> > >
> > > Compaction is region server activity so you should be able to find some
> > clue
> > > in region server log.
> > >
> > > Cheers
> > >
> > > On Jan 1, 2013, at 11:42 AM, Jean-Marc Spaggiari <
> > [EMAIL PROTECTED]>
> > > wrote:
> > >
> > >> Hi,
> > >>
> > >> I have a table, with about 200 000 raws, keys is string[64] (ish) and
> > >> value is string[512].
> > >>
> > >> It's splitted over 16 regions located on 7 regionservers.
> > >>
> > >> So it's not a big table, and there is a lot of horsepower behind it.
> > >>
> > >> I asked a major_compaction few hours ago. Let's say, about 5 hours
> > >> ago. and it's still compacting! But all servers activities seems to be
> > >> null. CPU usage is almost 0.
> > >>
> > >> There is nothing on the master logs.
> > >>
> > >> How can I see what's going on? Is there a way to see the compaction
> > >> queue?
> > >>
> > >> JM
> > >
> >
>

--
Adrien Mogenet
06.59.16.64.22
http://www.mogenet.me

When I'm doing the compaction it's because one region moved to a new
server, and that's what was never ending.

I was not able to reproduce that. I will continue to keep an eye on
this and will provide logs and details if it happens again.

2013/1/2, Adrien Mogenet <[EMAIL PROTECTED]>:
> Do you have 100% data locality ? How is the network trafic ?
>
>
> On Wed, Jan 2, 2013 at 7:07 AM, Ted Yu <[EMAIL PROTECTED]> wrote:
>
>> bq. It took about 6h to complete.
>>
>> If the above behavior is reproducible, we should investigate more deeply.
>>
>> Thanks for sharing.
>>
>> On Tue, Jan 1, 2013 at 6:14 PM, Jean-Marc Spaggiari <
>> [EMAIL PROTECTED]
>> > wrote:
>>
>> > Yes, I'm running on 0.94.3. The last major compaction ran yesterday.
>> > It's almost daily. That's why I was surprise it took so long. I mean,
>> > I'm only compacting regions who moved. So it should be pretty quick.
>> > But was not the case. It took about 6h to complete. Strange. Maybe
>> > something went wrong when I stopped/started hbase.
>> >
>> > also, there was almost no activity on the network nor on the CPUs. I
>> > will have to add disks monitoring on Ganglia to see if I was limited
>> > by IOs...
>> >
>> > I looked at the regions logs and everything was fine. It was showing
>> > some compaction information every few seconds.
>> >
>> > JM
>> >
>> > 2013/1/1, Ted <[EMAIL PROTECTED]>:
>> > > You're on hbase 0.94.3 , right ?
>> > >
>> > > When was the last time major compaction ran ?
>> > >
>> > > Compaction is region server activity so you should be able to find
>> > > some
>> > clue
>> > > in region server log.
>> > >
>> > > Cheers
>> > >
>> > > On Jan 1, 2013, at 11:42 AM, Jean-Marc Spaggiari <
>> > [EMAIL PROTECTED]>
>> > > wrote:
>> > >
>> > >> Hi,
>> > >>
>> > >> I have a table, with about 200 000 raws, keys is string[64] (ish)
>> > >> and
>> > >> value is string[512].
>> > >>
>> > >> It's splitted over 16 regions located on 7 regionservers.
>> > >>
>> > >> So it's not a big table, and there is a lot of horsepower behind it.
>> > >>
>> > >> I asked a major_compaction few hours ago. Let's say, about 5 hours
>> > >> ago. and it's still compacting! But all servers activities seems to
>> > >> be
>> > >> null. CPU usage is almost 0.
>> > >>
>> > >> There is nothing on the master logs.
>> > >>
>> > >> How can I see what's going on? Is there a way to see the compaction
>> > >> queue?
>> > >>
>> > >> JM
>> > >
>> >
>>
>
>
>
> --
> Adrien Mogenet
> 06.59.16.64.22
> http://www.mogenet.me
>