You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hbase.apache.org by la...@apache.org on 2015/12/31 09:04:19 UTC
Compact after manual flush broken in 1.0+?
Normally HBase triggers a compaction after a manual flush (of a region or table).
Looks like in 1.0 and later the code is still there but not actually exercised, because FlushTableSubprocedure calls Region.flush() directly and ignores the return code.
Was that by design? If so, we can remove the flushResult logic. But it looks like this is an oversight and we want to trigger the compaction still.
Comments?
-- Lars
Re: Compact after manual flush broken in 1.0+?
Posted by Vladimir Rodionov <vl...@gmail.com>.
>> Normally HBase triggers a compaction after a manual flush (of a region
or table).
Yes, I found this while working on FIFOCompaction test. Manual flush does
not trigger compaction. But wait, we can manually trigger compaction if we
need :).
-Vlad
On Wed, Jan 6, 2016 at 1:23 PM, Jean-Marc Spaggiari <jean-marc@spaggiari.org
> wrote:
> Unable to find it. Feel free to open.
>
> JMS
>
> 2015-12-31 12:36 GMT-05:00 Jean-Marc Spaggiari <je...@spaggiari.org>:
>
> > I think there is one. I will search later today...
> > Le 2015-12-31 11:35 AM, "Ted Yu" <yu...@gmail.com> a écrit :
> >
> >> A brief search turned up no hit.
> >>
> >> Please go ahead with new JIRA.
> >>
> >> On Thu, Dec 31, 2015 at 8:32 AM, <la...@apache.org> wrote:
> >>
> >> > Cool :)
> >> > Is there a jira already? Otherwise I'll file one.
> >> > -- Lars
> >> >
> >> > From: Jean-Marc Spaggiari <je...@spaggiari.org>
> >> > To: dev <de...@hbase.apache.org>
> >> > Cc: lars hofhansl <la...@apache.org>
> >> > Sent: Thursday, December 31, 2015 6:10 AM
> >> > Subject: Re: Compact after manual flush broken in 1.0+?
> >> >
> >> > Indeed. I already informed the list about that months ago ;)
> >> >
> >> > 2015-12-31 9:09 GMT-05:00 Ted Yu <yu...@gmail.com>:
> >> >
> >> > > bq. it looks like this is an oversight
> >> > >
> >> > > I agree.
> >> > >
> >> > > On Thu, Dec 31, 2015 at 12:04 AM, <la...@apache.org> wrote:
> >> > >
> >> > > > Normally HBase triggers a compaction after a manual flush (of a
> >> region
> >> > or
> >> > > > table).
> >> > > > Looks like in 1.0 and later the code is still there but not
> actually
> >> > > > exercised, because FlushTableSubprocedure calls Region.flush()
> >> directly
> >> > > and
> >> > > > ignores the return code.
> >> > > >
> >> > > > Was that by design? If so, we can remove the flushResult logic.
> But
> >> it
> >> > > > looks like this is an oversight and we want to trigger the
> >> compaction
> >> > > still.
> >> > > > Comments?
> >> > > > -- Lars
> >> > > >
> >> > > >
> >> > >
> >> >
> >> >
> >> >
> >> >
> >>
> >
>
Re: Compact after manual flush broken in 1.0+?
Posted by Jean-Marc Spaggiari <je...@spaggiari.org>.
Unable to find it. Feel free to open.
JMS
2015-12-31 12:36 GMT-05:00 Jean-Marc Spaggiari <je...@spaggiari.org>:
> I think there is one. I will search later today...
> Le 2015-12-31 11:35 AM, "Ted Yu" <yu...@gmail.com> a écrit :
>
>> A brief search turned up no hit.
>>
>> Please go ahead with new JIRA.
>>
>> On Thu, Dec 31, 2015 at 8:32 AM, <la...@apache.org> wrote:
>>
>> > Cool :)
>> > Is there a jira already? Otherwise I'll file one.
>> > -- Lars
>> >
>> > From: Jean-Marc Spaggiari <je...@spaggiari.org>
>> > To: dev <de...@hbase.apache.org>
>> > Cc: lars hofhansl <la...@apache.org>
>> > Sent: Thursday, December 31, 2015 6:10 AM
>> > Subject: Re: Compact after manual flush broken in 1.0+?
>> >
>> > Indeed. I already informed the list about that months ago ;)
>> >
>> > 2015-12-31 9:09 GMT-05:00 Ted Yu <yu...@gmail.com>:
>> >
>> > > bq. it looks like this is an oversight
>> > >
>> > > I agree.
>> > >
>> > > On Thu, Dec 31, 2015 at 12:04 AM, <la...@apache.org> wrote:
>> > >
>> > > > Normally HBase triggers a compaction after a manual flush (of a
>> region
>> > or
>> > > > table).
>> > > > Looks like in 1.0 and later the code is still there but not actually
>> > > > exercised, because FlushTableSubprocedure calls Region.flush()
>> directly
>> > > and
>> > > > ignores the return code.
>> > > >
>> > > > Was that by design? If so, we can remove the flushResult logic. But
>> it
>> > > > looks like this is an oversight and we want to trigger the
>> compaction
>> > > still.
>> > > > Comments?
>> > > > -- Lars
>> > > >
>> > > >
>> > >
>> >
>> >
>> >
>> >
>>
>
Re: Compact after manual flush broken in 1.0+?
Posted by Jean-Marc Spaggiari <je...@spaggiari.org>.
I think there is one. I will search later today...
Le 2015-12-31 11:35 AM, "Ted Yu" <yu...@gmail.com> a écrit :
> A brief search turned up no hit.
>
> Please go ahead with new JIRA.
>
> On Thu, Dec 31, 2015 at 8:32 AM, <la...@apache.org> wrote:
>
> > Cool :)
> > Is there a jira already? Otherwise I'll file one.
> > -- Lars
> >
> > From: Jean-Marc Spaggiari <je...@spaggiari.org>
> > To: dev <de...@hbase.apache.org>
> > Cc: lars hofhansl <la...@apache.org>
> > Sent: Thursday, December 31, 2015 6:10 AM
> > Subject: Re: Compact after manual flush broken in 1.0+?
> >
> > Indeed. I already informed the list about that months ago ;)
> >
> > 2015-12-31 9:09 GMT-05:00 Ted Yu <yu...@gmail.com>:
> >
> > > bq. it looks like this is an oversight
> > >
> > > I agree.
> > >
> > > On Thu, Dec 31, 2015 at 12:04 AM, <la...@apache.org> wrote:
> > >
> > > > Normally HBase triggers a compaction after a manual flush (of a
> region
> > or
> > > > table).
> > > > Looks like in 1.0 and later the code is still there but not actually
> > > > exercised, because FlushTableSubprocedure calls Region.flush()
> directly
> > > and
> > > > ignores the return code.
> > > >
> > > > Was that by design? If so, we can remove the flushResult logic. But
> it
> > > > looks like this is an oversight and we want to trigger the compaction
> > > still.
> > > > Comments?
> > > > -- Lars
> > > >
> > > >
> > >
> >
> >
> >
> >
>
Re: Compact after manual flush broken in 1.0+?
Posted by Ted Yu <yu...@gmail.com>.
A brief search turned up no hit.
Please go ahead with new JIRA.
On Thu, Dec 31, 2015 at 8:32 AM, <la...@apache.org> wrote:
> Cool :)
> Is there a jira already? Otherwise I'll file one.
> -- Lars
>
> From: Jean-Marc Spaggiari <je...@spaggiari.org>
> To: dev <de...@hbase.apache.org>
> Cc: lars hofhansl <la...@apache.org>
> Sent: Thursday, December 31, 2015 6:10 AM
> Subject: Re: Compact after manual flush broken in 1.0+?
>
> Indeed. I already informed the list about that months ago ;)
>
> 2015-12-31 9:09 GMT-05:00 Ted Yu <yu...@gmail.com>:
>
> > bq. it looks like this is an oversight
> >
> > I agree.
> >
> > On Thu, Dec 31, 2015 at 12:04 AM, <la...@apache.org> wrote:
> >
> > > Normally HBase triggers a compaction after a manual flush (of a region
> or
> > > table).
> > > Looks like in 1.0 and later the code is still there but not actually
> > > exercised, because FlushTableSubprocedure calls Region.flush() directly
> > and
> > > ignores the return code.
> > >
> > > Was that by design? If so, we can remove the flushResult logic. But it
> > > looks like this is an oversight and we want to trigger the compaction
> > still.
> > > Comments?
> > > -- Lars
> > >
> > >
> >
>
>
>
>
Re: Compact after manual flush broken in 1.0+?
Posted by la...@apache.org.
Cool :)
Is there a jira already? Otherwise I'll file one.
-- Lars
From: Jean-Marc Spaggiari <je...@spaggiari.org>
To: dev <de...@hbase.apache.org>
Cc: lars hofhansl <la...@apache.org>
Sent: Thursday, December 31, 2015 6:10 AM
Subject: Re: Compact after manual flush broken in 1.0+?
Indeed. I already informed the list about that months ago ;)
2015-12-31 9:09 GMT-05:00 Ted Yu <yu...@gmail.com>:
> bq. it looks like this is an oversight
>
> I agree.
>
> On Thu, Dec 31, 2015 at 12:04 AM, <la...@apache.org> wrote:
>
> > Normally HBase triggers a compaction after a manual flush (of a region or
> > table).
> > Looks like in 1.0 and later the code is still there but not actually
> > exercised, because FlushTableSubprocedure calls Region.flush() directly
> and
> > ignores the return code.
> >
> > Was that by design? If so, we can remove the flushResult logic. But it
> > looks like this is an oversight and we want to trigger the compaction
> still.
> > Comments?
> > -- Lars
> >
> >
>
Re: Compact after manual flush broken in 1.0+?
Posted by Jean-Marc Spaggiari <je...@spaggiari.org>.
Indeed. I already informed the list about that months ago ;)
2015-12-31 9:09 GMT-05:00 Ted Yu <yu...@gmail.com>:
> bq. it looks like this is an oversight
>
> I agree.
>
> On Thu, Dec 31, 2015 at 12:04 AM, <la...@apache.org> wrote:
>
> > Normally HBase triggers a compaction after a manual flush (of a region or
> > table).
> > Looks like in 1.0 and later the code is still there but not actually
> > exercised, because FlushTableSubprocedure calls Region.flush() directly
> and
> > ignores the return code.
> >
> > Was that by design? If so, we can remove the flushResult logic. But it
> > looks like this is an oversight and we want to trigger the compaction
> still.
> > Comments?
> > -- Lars
> >
> >
>
Re: Compact after manual flush broken in 1.0+?
Posted by Ted Yu <yu...@gmail.com>.
bq. it looks like this is an oversight
I agree.
On Thu, Dec 31, 2015 at 12:04 AM, <la...@apache.org> wrote:
> Normally HBase triggers a compaction after a manual flush (of a region or
> table).
> Looks like in 1.0 and later the code is still there but not actually
> exercised, because FlushTableSubprocedure calls Region.flush() directly and
> ignores the return code.
>
> Was that by design? If so, we can remove the flushResult logic. But it
> looks like this is an oversight and we want to trigger the compaction still.
> Comments?
> -- Lars
>
>