Re: Solr Operator Autoscaling

[Houston Putman <ho...@apache.org>]

So the SIP for this is
https://cwiki.apache.org/confluence/display/SOLR/SIP-17%3A+Node+Autoscaling+via+Kubernetes,
and you can find more information there.

The features are not released, but they should be available starting with
Solr 9.3 and the Solr Operator v0.8.0.

As for adding new replicas, that's not currently in the design but it's a
future part of the roadmap (and was discussed on the mailing list thread
linked in the SIP. The current design is mainly adding Nodes and moving the
existing replicas around.

I have heard of people using Solr (via the Solr Operator) with HPAs, but it
would be a manual process without this new integration.

- Houston

On Sat, May 27, 2023 at 5:55 PM Aurélien MAZOYER <au...@gmail.com>
wrote:

> Hi Mike and thanks for sharing. It looks indeed quite interesting :-) But
> based on the confluence page and the words used in the abstract ("CAN be
> re-introduced"), it seems that the solr operator is not totally ready yet
> to handle scenarios with auto-scaling (but I would be glad to hear that I
> am wrong :-)). Is there anybody out there who have tried this kind of
> scenario with solr cloud and kubernetes (with or without solr operator), i
> ll really appreciate to have some feedback :-).
>
> Thanks !
>
>
> El sáb, 27 may 2023, 6:20 p. m., Mike Drob <md...@mdrob.com> escribió:
>
> > I don’t know much about this area, but it looks like there will be a
> > session on exactly this at Berlin Buzzwords in a few weeks!
> >
> > https://program.berlinbuzzwords.de/berlin-buzzwords-2023/talk/ELVNYV/
> >
> > On Sat, May 27, 2023 at 10:59 AM Aurélien MAZOYER <
> > aurelien.mazoyer@gmail.com> wrote:
> >
> > > Hello Solr (operator) community!
> > >
> > > As far as I understood from the jira
> > >
> > >
> >
> https://cwiki.apache.org/confluence/display/SOLR/SIP-17%3A+Node+Autoscaling+via+Kubernetes
> > > , this is quite a lot of work ongoing since the AutoScaling
> > > functionality in Solr has been deprecated. However, it is not clear to
> me
> > > if as of now there is a possibility to handle simple scaling scenarios
> > with
> > > the Replica Placement API, Solr Cloud and the Solr Operator. Let's take
> > the
> > > simplest possible one: One collection, one shard, one replica. We would
> > > like to add nodes and Solr replicas if the load increases. If we
> > configure
> > > the HPA with the Solr Operator like explained in the documentation (
> > > https://apache.github.io/solr-operator/docs/local_tutorial), is there
> > also
> > > a way to automatically add a new replica on the pods that are started
> > when
> > > HPA scales the cluster?
> > >
> > > Thanks for your help!
> > >
> > > Aurelien
> > >
> >
>