kafka-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ali Nazemian <alinazem...@gmail.com>
Subject Re: Kafka disk recommendation for production cluster
Date Thu, 12 Jul 2018 10:14:40 GMT
I meant because there is no actual Raid Controller for running Raid on
cloud, it is basically Software Based Raid [1].

[1] https://en.wikipedia.org/wiki/RAID#Software-based


On Thu, Jul 12, 2018 at 6:29 PM M. Manna <manmedia@gmail.com> wrote:

> From all cases I know so far (including RDBMS storage solution), RAID 10
> has been the choice. Also, the replication you are mentioning is the s/w
> replication nothing to do with RAID 0 setup.
>
>
>
> On 11 July 2018 at 23:59, Ali Nazemian <alinazemian@gmail.com> wrote:
>
> > Thanks. As this proposal is not available for the version of Kafka that
> we
> > are using, we should go with Raid option, but the question is how is it
> > acceptable to use Software Raid0 on a cloud infrastructure?
> >
> > On Thu, 12 Jul. 2018, 08:02 M. Manna, <manmedia@gmail.com> wrote:
> >
> > > Dong Lin's KIP -
> > >
> > > https://cwiki.apache.org/confluence/display/KAFKA/KIP-
> > 112%3A+Handle+disk+failure+for+JBOD
> > >
> > > Should give you some ideas.
> > >
> > > On 11 July 2018 at 14:31, Ali Nazemian <alinazemian@gmail.com> wrote:
> > >
> > > > Hi All,
> > > >
> > > > I was wondering what the disk recommendation is for Kafka cluster? Is
> > it
> > > > acceptable to use RAID0 in the case that replication is 3? We are
> > running
> > > > on a cloud infrastructure and disk failure is addressed at another
> > level,
> > > > so the chance of single disk failure would be very low. Besides, our
> > > > version of Kafka does not fully support JBOD as even a single disk
> > > failure
> > > > can stop a Kafka broker, so we will not get the actual benefit of
> using
> > > > JBOD anyway. However, I am not quite sure how software raid acts in
> > this
> > > > situation as there is no option to use HW Raid on cloud.
> > > >
> > > > Regards,
> > > > Ali
> > > >
> > >
> >
>


-- 
A.Nazemian

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