hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stack <st...@duboce.net>
Subject Re: hbase:meta blocks in L1?
Date Tue, 19 Jul 2016 16:30:49 GMT
Addressed by: HBASE-16237 Blocks for hbase:meta table are not cached in L1
cache
M

On Fri, Jul 15, 2016 at 1:06 AM, Lars George <lars.george@gmail.com> wrote:

> I just checked on 1.2.1, using the "as JSON by File", then searching
> for the cached file (the hash). And yes, the hbase:meta blocks are in
> L2 once you enable it. So the missing setCacheDataInL1(true) for the
> metaTableDescriptor() method is causing this.
>
> On Fri, Jul 15, 2016 at 2:08 AM, Ted Yu <yuzhihong@gmail.com> wrote:
> > I looked at related code in branch-1.
> > HColumnDescriptor#isCacheDataInL1() uses default value of
> > HColumnDescriptor.DEFAULT_CACHE_DATA_IN_L1 which is false
> >
> > I think setCacheDataInL1(true) call is needed for meta HTableDescriptor
> >
> > On Thu, Jul 14, 2016 at 12:57 PM, Nick Dimiduk <ndimiduk@gmail.com>
> wrote:
> >
> >> IIRC, the configuration handling and IoC of multi-level block cache
> changed
> >> significantly around 1.2. This was to facilitate an overhaul and overall
> >> improvement of how cache layers are built, defined, and configured.
> >> Structures may be quite different between master and various
> incarnations
> >> of 1.x branches. Please examine each independently, don't assume what
> holds
> >> on one branch applies to another.
> >>
> >> On Thu, Jul 14, 2016 at 12:35 PM, Ted Yu <yuzhihong@gmail.com> wrote:
> >>
> >> > Interesting.
> >> >
> >> > Do you want to log a JIRA ?
> >> >
> >> > I checked HTableDescriptor.java in branch-1.1 to branch-1 where
> >> > setCacheDataInL1(true) is missing.
> >> >
> >> > The fix would be a one line change.
> >> >
> >> > On Thu, Jul 14, 2016 at 11:43 AM, Lars George <lars.george@gmail.com>
> >> > wrote:
> >> >
> >> > > Ah wait, this was fixed in
> >> > > https://issues.apache.org/jira/browse/HBASE-15467, but is not right
> >> > > (if I read this proper) in the 1.x line. It could mean that
> hbase:meta
> >> > > is cached in L2 only.
> >> > >
> >> > > On Thu, Jul 14, 2016 at 7:37 PM, Ted Yu <yuzhihong@gmail.com>
> wrote:
> >> > > > Looking at FSTableDescriptors#createMetaTableDescriptor() , both
> >> column
> >> > > > families have the following:
> >> > > >
> >> > > >                 .setCacheDataInL1(true),
> >> > > >
> >> > > > Is there something else you were looking for ?
> >> > > >
> >> > > > Cheers
> >> > > >
> >> > > > On Thu, Jul 14, 2016 at 6:59 AM, Lars George <
> lars.george@gmail.com>
> >> > > wrote:
> >> > > >
> >> > > >> Hey,
> >> > > >>
> >> > > >> HBASE-11364 did add `.setCacheDataInL1(true)`, but the later
> change
> >> to
> >> > > >> use `HTableDescriptor.metaTableDescriptor()` changes the
layout
> of
> >> the
> >> > > >> `hbase:meta` to omit the L1 cache setting. Is this on purpose
or
> an
> >> > > >> oversight?
> >> > > >>
> >> > > >> In other words, are the meta blocks currently never cached
in L1?
> >> > > >>
> >> > > >> Lars
> >> > > >>
> >> > >
> >> >
> >>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message