kafka-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jun Rao <...@confluent.io>
Subject Re: [VOTE] KIP-188 - Add new metrics to support health checks
Date Fri, 08 Sep 2017 16:09:33 GMT
Hi, Raijini,

Thanks for the explanation. We can leave those as they are then.

Jun

On Thu, Sep 7, 2017 at 3:23 AM, Rajini Sivaram <rajinisivaram@gmail.com>
wrote:

> Hi all,
>
> I have added one more metric to KIP-188 to show the current status of
> broker's ZooKeeper connections. Please let me know if you have any
> concerns.
>
> Hi Jun,
>
> I was wondering which is a better group for FetchMessageConversionsPerSec,
> now that we have MessageConversionsTimeMs at the request level. I don't
> have a strong opinion either way, but at the moment, as a topic metric,
> xxxMessageConversionsPerSec is along with MessagesInPerSec,
> TotalFetchRequestsPerSec etc. which are all kind of related and are all
> rate metrics. If we move it to the request level, we will
> have MessageConversionsTimeMs and MessageConversionsPerSec together, but
> will lose the topic grouping. Since all the other metrics at request level
> are time histograms, perhaps it is better to leave MessageConversionsPerSec
> along with the other topic rate metrics?
>
> I had added a ZooKeeperClient wrapper for my initial PR, but I will rebase
> on Onur's code when it is ready. Thank you!
>
> Many thanks,
>
> Rajini
>
> On Wed, Sep 6, 2017 at 10:52 PM, Jun Rao <jun@confluent.io> wrote:
>
> > Hi, Raijini,
> >
> > Thanks for the KIP. +1. Just a minor comment.
> >
> > Since we only measure MessageConversionsTimeMs at the request type level,
> > is it useful to collect the following metrics at the topic level?
> >
> > *MBean*:
> > kafka.server:type=BrokerTopicMetrics,name=FetchMessageConversionsPerSec,
> > topic=([-.\w]+)
> >
> > *MBean*:
> > kafka.server:type=BrokerTopicMetrics,name=ProduceMessageConversionsPerSe
> > c,topic=([-.\w]+)
> >
> >
> > Also, for the ZK latency metric, Onur added a new ZookeeperClient wrapper
> > and is in the middle of converting existing zkClient usage to the new
> > wrapper. So, we probably want to add the latency metric in the new
> wrapper.
> >
> > Jun
> >
> > On Thu, Aug 24, 2017 at 10:50 AM, Rajini Sivaram <
> rajinisivaram@gmail.com>
> > wrote:
> >
> > > Hi all,
> > >
> > > I would like to start the vote on KIP-188 that adds additional metrics
> to
> > > support health checks for Kafka Ops. Details are here:
> > >
> > > https://cwiki.apache.org/confluence/display/KAFKA/KIP-
> > > 188+-+Add+new+metrics+to+support+health+checks
> > >
> > > Thank you,
> > >
> > > Rajini
> > >
> >
>

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