Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
HBase, mail # dev - Design review: Secondary index support through coprocessors


Copy link to this message
-
Re: Design review: Secondary index support through coprocess
Wei Tan 2014-01-22, 19:38
Hi James, I guess as long as you have cross-region RPC this cyclic
deadlock can occur? If you batch the RPC, it may be far less likely to
occur because the number of RPC is far fewer. Do I understand you
correctly?
Thanks,
Wei

---------------------------------
Wei Tan, PhD
Research Staff Member
IBM T. J. Watson Research Center
http://researcher.ibm.com/person/us-wtan

From:   James Taylor <[EMAIL PROTECTED]>
To:     "[EMAIL PROTECTED]" <[EMAIL PROTECTED]>,
Date:   01/22/2014 02:01 PM
Subject:        Re: Design review: Secondary index support through
coprocess

FYI, the Phoenix mutable secondary indexing implementation won't make
cyclical RPC calls like this. We do a single batched mutation from the RS
of the data table to the RS of the index table to do the index
maintenance.
But the situation Jesse described is possible, though we have not
encountered this in our testing. As Jesse mentioned, the fix is to
have a custom
RPC handler for index writes that has a higher priority than the other
ones.

Thanks,
James
On Wed, Jan 22, 2014 at 9:06 AM, Vladimir Rodionov
<[EMAIL PROTECTED]>wrote:

> Deadlocks are possible because  cross region RPCs create cyclic
> dependencies in HBase cluster.
>
> RS1-> RS2->RS3->RS1, where -> is PRC call
>
> now imagine that last call from RS3 to RS1 is blocked because there no
> more available handler threads to process it.
>
> Best regards,
> Vladimir Rodionov
> Principal Platform Engineer
> Carrier IQ, www.carrieriq.com
> e-mail: [EMAIL PROTECTED]
>
> ________________________________________
> From: Wei Tan [[EMAIL PROTECTED]]
> Sent: Wednesday, January 22, 2014 7:51 AM
> To: [EMAIL PROTECTED]
> Subject: RE: Design review: Secondary index support through coprocess
>
> Why cross-RS RPC is going to cause deadlocks? It is a matter of logic
> incorrectness, or resource outage? Say, if we set the #handler to be
> large, logically deadlock still occurs?
> Best regards,
> Wei
>
>
>
>
> From:   Vladimir Rodionov <[EMAIL PROTECTED]>
> To:     "[EMAIL PROTECTED]" <[EMAIL PROTECTED]>,
> Date:   01/20/2014 03:00 PM
> Subject:        RE: Design review: Secondary index support through
> coprocess
>
>
>
> >>Yes, the coprocessors potentially cross RS boundaries.
>
> The open path to the disaster. Inter region RPCs in coprocessors may
> result in periodic cluster - wide deadlocks
>
>
> Best regards,
> Vladimir Rodionov
> Principal Platform Engineer
> Carrier IQ, www.carrieriq.com
> e-mail: [EMAIL PROTECTED]
>
> ________________________________________
> From: James Taylor [[EMAIL PROTECTED]]
> Sent: Monday, January 20, 2014 11:39 AM
> To: [EMAIL PROTECTED]
> Subject: Re: Design review: Secondary index support through coprocess
>
> Yes, the coprocessors potentially cross RS boundaries. No, the index is
> not
> co-located with the main table. Take a look at the link I sent as that
> should be able to answer a lot of questions.
>
> Thanks,
> James
>
>
> On Mon, Jan 20, 2014 at 11:03 AM, Michael Segel
> <[EMAIL PROTECTED]>wrote:
>
> > James,
> >
> > Ok…
> >
> > Its been a while since we talked about this…
> >
> > While the index is in a separate table, is that table being split and
> > collocated with the main table?
> >
> > If you’re using the coprocessor to maintain the index, that would
imply
> > you’re crossing RS boundaries if your index is truly orthogonal.
> >
> > Is this what you’re doing?
> >
> > On Jan 20, 2014, at 11:32 AM, James Taylor <[EMAIL PROTECTED]>
> wrote:
> >
> > > Mike,
> > > Yes, you're mistaken:
> > > - secondary indexes in Phoenix are orthogonal to the base table.
> They're
> > in
> > > a separate table (
> > > http://phoenix.incubator.apache.org/secondary_indexing.html).
> > > - Phoenix has joins. They're in our master branch with a release
> > scheduled
> > > for next month
> > > - numeric strings? Not a use case for indexing numeric data? Have
you
> > ever
> > > seen a number used as an ID?
> > > Thanks,
> > > James
much
only
query
were
and
Again,
indexing
in
that
scalable
different?
ngdata)?
much
impl
this
would
for
support
of
and
the
https://issues.apache.org/jira/secure/attachment/12621909/SecondaryIndex%20Design_Updated_2.pdf

or
message
be
If
form,
please