You are viewing a plain text version of this content. The canonical link for it is here.
Posted to solr-user@lucene.apache.org by Amrit Sarkar <sa...@gmail.com> on 2017/01/28 08:21:21 UTC
Regarding embedded ZK with Solr
I would like to understand how the embedded ZK works with Solr. If Xg
memory is allocated to the Solr installation and we spin up the SolrCloud
with embedded ZK; what part/percentage of the X is allocated to the ZK or
is it shared?
If that is known, how can I change the memory settings for the embedded ZK?
Amrit Sarkar
Search Engineer
Lucidworks, Inc.
415-589-9269
www.lucidworks.com
Twitter http://twitter.com/lucidworks
LinkedIn: https://www.linkedin.com/in/sarkaramrit2
Re: Regarding embedded ZK with Solr
Posted by Shawn Heisey <ap...@elyograg.org>.
On 1/28/2017 1:21 AM, Amrit Sarkar wrote:
> I would like to understand how the embedded ZK works with Solr. If Xg
> memory is allocated to the Solr installation and we spin up the SolrCloud
> with embedded ZK; what part/percentage of the X is allocated to the ZK or
> is it shared?
>
> If that is known, how can I change the memory settings for the embedded ZK?
It's all one JVM, so it's all one shared heap. I don't know of any way
within Solr or Zookeeper for it to be configured to specify how much of
the heap it is allowed to use. Both pieces of software will allocate
whatever they need to do their jobs, assuming the JVM has been
configured with a large enough heap.
You should not be using the embedded zookeeper except for test or demo
purposes. A production SolrCloud cluster should be configured with
separate zookeeper processes, running on at least three servers. When
using the embedded zookeeper, it goes down anytime that instance of Solr
does.
Thanks,
Shawn
Re: Regarding embedded ZK with Solr
Posted by Amrit Sarkar <sa...@gmail.com>.
Hi Jan,
Nevermind you already answered that in the mailing list. Thank you again.
Amrit Sarkar
Search Engineer
Lucidworks, Inc.
415-589-9269
www.lucidworks.com
Twitter http://twitter.com/lucidworks
LinkedIn: https://www.linkedin.com/in/sarkaramrit2
On Mon, Jan 30, 2017 at 1:21 PM, Amrit Sarkar <sa...@gmail.com>
wrote:
> Hi Jan,
>
> Thank you for responding. I understand we should not use embedded ZK in
> production. As three components share the same heap, I am assuming there is
> no way to control the component individual consumption rate?
>
> Amrit Sarkar
> Search Engineer
> Lucidworks, Inc.
> 415-589-9269
> www.lucidworks.com
> Twitter http://twitter.com/lucidworks
> LinkedIn: https://www.linkedin.com/in/sarkaramrit2
>
> On Sun, Jan 29, 2017 at 4:35 AM, Jan Høydahl <ja...@cominvent.com>
> wrote:
>
>> Hi,
>>
>> The embedded ZK is not intended for production use, please use a
>> standalone cluster.
>> Embedded ZK will share JVM heap with Jetty and Solr, so if any of them
>> cause OOM, it will affect all.
>>
>> --
>> Jan Høydahl, search solution architect
>> Cominvent AS - www.cominvent.com
>>
>> > 28. jan. 2017 kl. 09.21 skrev Amrit Sarkar <sa...@gmail.com>:
>> >
>> > I would like to understand how the embedded ZK works with Solr. If Xg
>> > memory is allocated to the Solr installation and we spin up the
>> SolrCloud
>> > with embedded ZK; what part/percentage of the X is allocated to the ZK
>> or
>> > is it shared?
>> >
>> > If that is known, how can I change the memory settings for the embedded
>> ZK?
>> >
>> > Amrit Sarkar
>> > Search Engineer
>> > Lucidworks, Inc.
>> > 415-589-9269
>> > www.lucidworks.com
>> > Twitter http://twitter.com/lucidworks
>> > LinkedIn: https://www.linkedin.com/in/sarkaramrit2
>>
>>
>
Re: Regarding embedded ZK with Solr
Posted by Amrit Sarkar <sa...@gmail.com>.
Hi Jan,
Thank you for responding. I understand we should not use embedded ZK in
production. As three components share the same heap, I am assuming there is
no way to control the component individual consumption rate?
Amrit Sarkar
Search Engineer
Lucidworks, Inc.
415-589-9269
www.lucidworks.com
Twitter http://twitter.com/lucidworks
LinkedIn: https://www.linkedin.com/in/sarkaramrit2
On Sun, Jan 29, 2017 at 4:35 AM, Jan Høydahl <ja...@cominvent.com> wrote:
> Hi,
>
> The embedded ZK is not intended for production use, please use a
> standalone cluster.
> Embedded ZK will share JVM heap with Jetty and Solr, so if any of them
> cause OOM, it will affect all.
>
> --
> Jan Høydahl, search solution architect
> Cominvent AS - www.cominvent.com
>
> > 28. jan. 2017 kl. 09.21 skrev Amrit Sarkar <sa...@gmail.com>:
> >
> > I would like to understand how the embedded ZK works with Solr. If Xg
> > memory is allocated to the Solr installation and we spin up the SolrCloud
> > with embedded ZK; what part/percentage of the X is allocated to the ZK or
> > is it shared?
> >
> > If that is known, how can I change the memory settings for the embedded
> ZK?
> >
> > Amrit Sarkar
> > Search Engineer
> > Lucidworks, Inc.
> > 415-589-9269
> > www.lucidworks.com
> > Twitter http://twitter.com/lucidworks
> > LinkedIn: https://www.linkedin.com/in/sarkaramrit2
>
>
Re: Regarding embedded ZK with Solr
Posted by Jan Høydahl <ja...@cominvent.com>.
Hi,
The embedded ZK is not intended for production use, please use a standalone cluster.
Embedded ZK will share JVM heap with Jetty and Solr, so if any of them cause OOM, it will affect all.
--
Jan Høydahl, search solution architect
Cominvent AS - www.cominvent.com
> 28. jan. 2017 kl. 09.21 skrev Amrit Sarkar <sa...@gmail.com>:
>
> I would like to understand how the embedded ZK works with Solr. If Xg
> memory is allocated to the Solr installation and we spin up the SolrCloud
> with embedded ZK; what part/percentage of the X is allocated to the ZK or
> is it shared?
>
> If that is known, how can I change the memory settings for the embedded ZK?
>
> Amrit Sarkar
> Search Engineer
> Lucidworks, Inc.
> 415-589-9269
> www.lucidworks.com
> Twitter http://twitter.com/lucidworks
> LinkedIn: https://www.linkedin.com/in/sarkaramrit2