Cos

Indeed It's more critical to File Systems and Databases for having SPOF
compared to computing engines.
>> >> might crash in 5 months or in 2 minutes.
If the failure is caused by a hardware issue, then it's unlikely to have
No.2 going down in just 2 minutes. Although most of the time in our case
they're software issues, which is why there's a case that No.2 goes down
immediately after No.1 dies.
Due to that, I see the HA feature in the way that it at least solve part of
the question.
Probably those gossip, no master implementations are more robust, I don't
have any experience though.


Faraaz

Thanks for the answer and looking for the announcement :)

2016-01-27 11:04 GMT+08:00 RJ Nowling <[email protected]>:

> Oops... meant QFS, not Apache Apex.  Sorry!  But the question about having
> a maintainer still stands.  :)
>
> On Tue, Jan 26, 2016 at 9:02 PM, RJ Nowling <[email protected]> wrote:
>
> > Hi Faraaz,
> >
> > Is there someone who would be willing to take responsibility for
> > maintaining the Apache Apex packages in Bigtop?
> >
> > We generally like to find a maintainer for each package and add them to
> > our maintainers list.  This way we know who to contact if a build fails
> and
> > blocks a release.  If there is no maintainer (and the build has
> problems),
> > we use that as grounds for removing the package from Bigtop.
> >
> > Thanks!
> > RJ
> >
> > On Tue, Jan 26, 2016 at 3:05 PM, Faraaz Sareshwala <
> > [email protected]> wrote:
> >
> >> Thanks for the warm welcome guys :).
> >>
> >> I agree that smaller patches will be easier to parse and review. Right
> >> now, I have Debian and RPM packaging complete. I have created a JIRA
> ticket
> >> (https://issues.apache.org/jira/browse/BIGTOP-2283) which I will use to
> >> track the integration.
> >>
> >> I’ll start sending in pull requests as I complete each of the tasks
> >> necessary to get this integration done. I’d love all of your guys’
> guidance
> >> along the way so that the integration can be as correct to bigtop’s
> >> standards as possible.
> >>
> >> Also, regarding the single point of failure sidenote discussed in
> another
> >> thread, Quantcast is actively working on updating qfs to support a
> >> distributed metaserver so that we can avoid the single point of failure.
> >> It’s not ready for release yet but we will announce it when it is :).
> >>
> >>
> >> Faraaz
> >>
> >> On 1/26/16, 11:33 AM, "jay vyas" <[email protected]> wrote:
> >>
> >> >Please add Your name to the MAINTAINERS.txt file in the PR for "QFS" !
> >> >
> >> >On Tue, Jan 26, 2016 at 2:24 PM, Konstantin Boudnik <[email protected]>
> >> wrote:
> >> >
> >> >> On Wed, Jan 27, 2016 at 01:05AM, Evans Ye wrote:
> >> >> > Oh, if I made you uncomfortable,  you know I didn't mean that :)
> >> >>
> >> >> Oh believe me - it takes a lot more than that to make me
> uncomfortable
> >> ;)
> >> >> Besides, you gave me an opportunity to speak up about the SPOFs in
> the
> >> >> Hadoop
> >> >> ecosystem. Which I can do for hours!
> >> >>
> >> >> > I just want to know more about the design because I'm always
> >> interested
> >> >> in
> >> >> > the architecture things, especially when I see something different.
> >> >> > Even Spark just added HA in their master in recent release last
> >> year. So
> >> >> > it's totally fine.
> >> >> >
> >> >> > And the following is just  for the discussion:
> >> >> >
> >> >> >> Just a side node: as you know literally _all_ components in
> >> Hadoop-based
> >> >> >> stack have SPOF.
> >> >> >
> >> >> > I think it's more related to the maturity of a component, but
> >> somehow it
> >> >> > will get there to eliminate the SPOF just like what HDFS, YARN,
> >> Oozie,
> >> >> Spark
> >> >> > did.  For operations, two SPOFs is good enough for them to have a
> >> good
> >> >> sleep
> >> >> > at night ;)
> >> >>
> >> >> It might be ok for Oozie, but not for a file system. Because once
> your
> >> >> master
> >> >> has failed over to the standby you don't know how long that one has.
> It
> >> >> might
> >> >> crash in 5 months or in 2 minutes. That's why if you're an admin for
> a
> >> >> production critical cluster you'll get paged immediately. Perhaps 3
> >> >> o'clock in
> >> >> the morning.
> >> >>
> >> >> I spent last three years working on the active-active software for
> >> mission
> >> >> critical systems, so I would love to share some architectural
> thoughts
> >> with
> >> >> you, if you're interested.
> >> >>
> >> >> Cos
> >> >>
> >> >> > 2016-01-26 12:08 GMT+08:00 Jay Vyas <[email protected]>:
> >> >> >
> >> >> > > Awesome!  It's great to hear about another HDFS alternative :)
> >> >> > >
> >> >> > > , We tried pretty hard to make bigtop more "HCFS" compliant, by
> >> adding
> >> >> the
> >> >> > > init hcfs tooling and so on, but never completed making the
> puppet
> >> >> recipes
> >> >> > > add support for other file systems.
> >> >> > >
> >> >> > >
> >> >> > > Would be very interesting to see if the quant cast team can take
> >> over
> >> >> on
> >> >> > > the HCFS integration so that bigtop is a home for more Than just
> >> one
> >> >> under
> >> >> > > filesystem .
> >> >> > >
> >> >> > > Let us know what your planning on working on, looking forward to
> >> it...!
> >> >> > >
> >> >> > >
> >> >> > >
> >> >> > >
> >> >> > > > On Jan 25, 2016, at 9:06 PM, Faraaz Sareshwala <
> >> >> > > [email protected]> wrote:
> >> >> > > >
> >> >> > > > Hello everyone in the Bigtop community!
> >> >> > > >
> >> >> > > > My name is Faraaz Sareshwala and I am an engineer working at
> >> >> Quantcast
> >> >> > > to integrate QFS (https://quantcast.github.io/qfs) into the
> Bigtop
> >> >> > > project. I just wanted to send a small note out to introduce
> >> myself to
> >> >> the
> >> >> > > community and announce that we are working on an integration.
> Feel
> >> >> free to
> >> >> > > reach out to me if you have any questions about QFS, our
> >> integration,
> >> >> or
> >> >> > > Quantcast itself.
> >> >> > > >
> >> >> > > > At this point, I am nearly complete integrating QFS. I plan to
> >> >> submit a
> >> >> > > patch through github and JIRA later in the week or early next
> week
> >> >> once I
> >> >> > > get a chance to test everything is working correctly. I’ve read
> the
> >> >> wiki on
> >> >> > > contribution to the project, but if you guys have any other tips
> or
> >> >> > > guidance for me, I’d really appreciate that.
> >> >> > > >
> >> >> > > > Patch coming soon!
> >> >> > > >
> >> >> > > > Take care,
> >> >> > > > Faraaz
> >> >> > > >
> >> >> > >
> >> >>
> >> >
> >> >
> >> >
> >> >--
> >> >jay vyas
> >>
> >
> >
>

Reply via email to