You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@kafka.apache.org by Jay Kreps <ja...@gmail.com> on 2015/02/14 00:28:38 UTC
Hold off on 0.8.2 upgrades
Hey all,
We found an issue in 0.8.2 that can lead to high CPU usage on brokers with
lots of partitions. We are working on a fix for this. You can track
progress here:
https://issues.apache.org/jira/browse/KAFKA-1952
I would recommend holding off on upgrading to 0.8.2 until we have a fix for
this issue. Sorry for the inconvenience.
-Jay
Re: Hold off on 0.8.2 upgrades
Posted by Jun Rao <ju...@confluent.io>.
Yes, that makes sense. I will try to roll out 0.8.2.1 for vote later today.
Thanks,
Jun
On Wed, Feb 18, 2015 at 4:15 PM, Jay Kreps <ja...@gmail.com> wrote:
> Well, I guess what I was thinking is that since we have the long timeout on
> the vote anyway, no reason not to call the vote now, should anything else
> pop up we can cancel the vote.
>
> -Jay
>
> On Wed, Feb 18, 2015 at 4:04 PM, Jun Rao <ju...@confluent.io> wrote:
>
> > Well, KAFKA-1952 only introduces high CPU overhead if the number of
> > partitions in a fetch request is high, say more than a couple of
> hundreds.
> > So, it may not show up in every installation. For example, if you have
> 1000
> > leader replicas in a broker, but have a 20 node cluster, each replica
> fetch
> > request is only going to include about 50 partitions. Since there is a
> bit
> > of overhead running a release, I was hoping to collect some more feedback
> > from people trying the 0.8.2.0 release who may not be affected by this
> > issue. But I agree that we don't need to wait for too long.
> >
> > Thanks,
> >
> > Jun
> >
> > On Wed, Feb 18, 2015 at 2:13 PM, Jay Kreps <ja...@gmail.com> wrote:
> >
> > > Does it make sense to wait, I don't think people will upgrade without
> the
> > > patched version and I think we should release it to unblock people.
> > >
> > > -Jay
> > >
> > > On Wed, Feb 18, 2015 at 1:43 PM, Jun Rao <ju...@confluent.io> wrote:
> > >
> > > > We have fixed the issue in KAFKA-1952. We will wait for a few more
> days
> > > to
> > > > see if any new issue comes up. After that, we will do an 0.8.2.1
> > release.
> > > >
> > > > Thanks,
> > > >
> > > > Jun
> > > >
> > > > On Fri, Feb 13, 2015 at 3:28 PM, Jay Kreps <ja...@gmail.com>
> > wrote:
> > > >
> > > > > Hey all,
> > > > >
> > > > > We found an issue in 0.8.2 that can lead to high CPU usage on
> brokers
> > > > with
> > > > > lots of partitions. We are working on a fix for this. You can track
> > > > > progress here:
> > > > > https://issues.apache.org/jira/browse/KAFKA-1952
> > > > >
> > > > > I would recommend holding off on upgrading to 0.8.2 until we have a
> > fix
> > > > for
> > > > > this issue. Sorry for the inconvenience.
> > > > >
> > > > > -Jay
> > > > >
> > > >
> > >
> >
>
Re: Hold off on 0.8.2 upgrades
Posted by Jay Kreps <ja...@gmail.com>.
Well, I guess what I was thinking is that since we have the long timeout on
the vote anyway, no reason not to call the vote now, should anything else
pop up we can cancel the vote.
-Jay
On Wed, Feb 18, 2015 at 4:04 PM, Jun Rao <ju...@confluent.io> wrote:
> Well, KAFKA-1952 only introduces high CPU overhead if the number of
> partitions in a fetch request is high, say more than a couple of hundreds.
> So, it may not show up in every installation. For example, if you have 1000
> leader replicas in a broker, but have a 20 node cluster, each replica fetch
> request is only going to include about 50 partitions. Since there is a bit
> of overhead running a release, I was hoping to collect some more feedback
> from people trying the 0.8.2.0 release who may not be affected by this
> issue. But I agree that we don't need to wait for too long.
>
> Thanks,
>
> Jun
>
> On Wed, Feb 18, 2015 at 2:13 PM, Jay Kreps <ja...@gmail.com> wrote:
>
> > Does it make sense to wait, I don't think people will upgrade without the
> > patched version and I think we should release it to unblock people.
> >
> > -Jay
> >
> > On Wed, Feb 18, 2015 at 1:43 PM, Jun Rao <ju...@confluent.io> wrote:
> >
> > > We have fixed the issue in KAFKA-1952. We will wait for a few more days
> > to
> > > see if any new issue comes up. After that, we will do an 0.8.2.1
> release.
> > >
> > > Thanks,
> > >
> > > Jun
> > >
> > > On Fri, Feb 13, 2015 at 3:28 PM, Jay Kreps <ja...@gmail.com>
> wrote:
> > >
> > > > Hey all,
> > > >
> > > > We found an issue in 0.8.2 that can lead to high CPU usage on brokers
> > > with
> > > > lots of partitions. We are working on a fix for this. You can track
> > > > progress here:
> > > > https://issues.apache.org/jira/browse/KAFKA-1952
> > > >
> > > > I would recommend holding off on upgrading to 0.8.2 until we have a
> fix
> > > for
> > > > this issue. Sorry for the inconvenience.
> > > >
> > > > -Jay
> > > >
> > >
> >
>
Re: Hold off on 0.8.2 upgrades
Posted by Jun Rao <ju...@confluent.io>.
Well, KAFKA-1952 only introduces high CPU overhead if the number of
partitions in a fetch request is high, say more than a couple of hundreds.
So, it may not show up in every installation. For example, if you have 1000
leader replicas in a broker, but have a 20 node cluster, each replica fetch
request is only going to include about 50 partitions. Since there is a bit
of overhead running a release, I was hoping to collect some more feedback
from people trying the 0.8.2.0 release who may not be affected by this
issue. But I agree that we don't need to wait for too long.
Thanks,
Jun
On Wed, Feb 18, 2015 at 2:13 PM, Jay Kreps <ja...@gmail.com> wrote:
> Does it make sense to wait, I don't think people will upgrade without the
> patched version and I think we should release it to unblock people.
>
> -Jay
>
> On Wed, Feb 18, 2015 at 1:43 PM, Jun Rao <ju...@confluent.io> wrote:
>
> > We have fixed the issue in KAFKA-1952. We will wait for a few more days
> to
> > see if any new issue comes up. After that, we will do an 0.8.2.1 release.
> >
> > Thanks,
> >
> > Jun
> >
> > On Fri, Feb 13, 2015 at 3:28 PM, Jay Kreps <ja...@gmail.com> wrote:
> >
> > > Hey all,
> > >
> > > We found an issue in 0.8.2 that can lead to high CPU usage on brokers
> > with
> > > lots of partitions. We are working on a fix for this. You can track
> > > progress here:
> > > https://issues.apache.org/jira/browse/KAFKA-1952
> > >
> > > I would recommend holding off on upgrading to 0.8.2 until we have a fix
> > for
> > > this issue. Sorry for the inconvenience.
> > >
> > > -Jay
> > >
> >
>
Re: Hold off on 0.8.2 upgrades
Posted by Jay Kreps <ja...@gmail.com>.
Does it make sense to wait, I don't think people will upgrade without the
patched version and I think we should release it to unblock people.
-Jay
On Wed, Feb 18, 2015 at 1:43 PM, Jun Rao <ju...@confluent.io> wrote:
> We have fixed the issue in KAFKA-1952. We will wait for a few more days to
> see if any new issue comes up. After that, we will do an 0.8.2.1 release.
>
> Thanks,
>
> Jun
>
> On Fri, Feb 13, 2015 at 3:28 PM, Jay Kreps <ja...@gmail.com> wrote:
>
> > Hey all,
> >
> > We found an issue in 0.8.2 that can lead to high CPU usage on brokers
> with
> > lots of partitions. We are working on a fix for this. You can track
> > progress here:
> > https://issues.apache.org/jira/browse/KAFKA-1952
> >
> > I would recommend holding off on upgrading to 0.8.2 until we have a fix
> for
> > this issue. Sorry for the inconvenience.
> >
> > -Jay
> >
>
Re: Hold off on 0.8.2 upgrades
Posted by Jun Rao <ju...@confluent.io>.
We have fixed the issue in KAFKA-1952. We will wait for a few more days to
see if any new issue comes up. After that, we will do an 0.8.2.1 release.
Thanks,
Jun
On Fri, Feb 13, 2015 at 3:28 PM, Jay Kreps <ja...@gmail.com> wrote:
> Hey all,
>
> We found an issue in 0.8.2 that can lead to high CPU usage on brokers with
> lots of partitions. We are working on a fix for this. You can track
> progress here:
> https://issues.apache.org/jira/browse/KAFKA-1952
>
> I would recommend holding off on upgrading to 0.8.2 until we have a fix for
> this issue. Sorry for the inconvenience.
>
> -Jay
>