On Thu, Sep 13, 2012 at 2:47 AM, Kathleen Ting <kathleen@apache.org> wrote:
> Hi Liwei,
>
> Could JVM warm-up account for the additional time required to getData?
>
I don't think it's something related to JVM. The test program works(gets
better performance) well with the other zookeeper cluster.
>
> Or do you have log4j logging set to something other than INFO?
> DEBUG/TRACE can slow things down.
>
no, the log setting is INFO.
>
> Regards, Kathleen
>
>
> On Wed, Sep 12, 2012 at 3:33 AM, sun liwei <sliveysun@gmail.com> wrote:
> > The version of zookeeper is 3.3.4.
> >
> > On Wed, Sep 12, 2012 at 6:07 PM, sun liwei <sliveysun@gmail.com> wrote:
> >
> >> Here is what I haveļ¼
> >>
> >>
> >> - zookeeper cluster with 3 severs
> >> - 1700+ kafka topics (that means there are 1700+ children under path
> >> '/brokers/topics'), these topics are written frequently by kafka.
> >> - log file size is 67108880 byte, snapshot file size is 4192072
> >> byte, new log/snapshot files are created every one or two minutes
> >>
> >> The problem is that it takes more than 40ms to getData of a single
> topic.
> >> I have another zookeeper cluster and the getData takes no more than 2ms
> to
> >> getData of a single path. This cluster doesn't server kafka.
> >>
> >> Any body knows why the performance is so bad?
> >>
> >> Many thanks and best regards.
> >>
> >> Liwei
> >>
>
|