You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by Shawn Heisey <ap...@elyograg.org> on 2016/03/09 18:34:22 UTC
Speculating about the removal of the standalone Solr mode
I've been thinking about the fact that standalone and cloud modes in
Solr are very different.
The writing on the wall suggests that Solr will eventually (probably 7.0
minimum) eliminate the standalone mode and always operate with
zookeeper. A "standalone" node would in fact be a single-node cloud
running the embedded zookeeper.
Once zk-as-truth becomes a reality, I can see a few advantages to always
running in cloud mode. The documentation can include one way to
accomplish basic tasks. The CoreAdmin API can be eliminated, and any
required functionality fully merged into the Collections API.
CloudSolrClient will work for all installations. A script that works
for cloud mode will also work for standalone mode, because that's just a
smaller cloud.
I was planning to open an issue to discuss and implement this. If
that's not a good idea, please let me know.
None of my main Solr installations are running in cloud mode, so the
removal of standalone mode will be an inconvenience for me, but I still
think it's the right thing to do in the long term.
Thanks,
Shawn
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org
RE: Speculating about the removal of the standalone Solr mode
Posted by "Dyer, James" <Ja...@ingramcontent.com>.
I would think it unfortunate if this ever happens. Solr in non-cloud mode is simple, easy-to-understand, has few moving parts. Many installations do not need to shard, have real-time-updates, etc. Using the replication handler in "legacy mode" works great for us. The config files are on the filesystem. You need not learn a cli to interact with zookeeper, etc. I would be scared to death running cloud mode in Production if I didn't first obtain an in-depth understanding of zookeeper internals.
I can see if there is a huge burden imposed here and if almost all use-cases require cloud. But as for "api consolidation", there are few api's you need to learn if running non-cloud. So what stops us from focusing apis on the need of cloud installations? And the documentation for non-cloud ought to be simple to maintain, there's so much less to learn and know.
For those of you that work as consultants or for support providers, it may seem that everyone is running cloud mode. But my guess is those who run cloud mode are the ones that cannot get by without your services.
James Dyer
Ingram Content Group
-----Original Message-----
From: Shawn Heisey [mailto:apache@elyograg.org]
Sent: Wednesday, March 09, 2016 11:34 AM
To: dev@lucene.apache.org
Subject: Speculating about the removal of the standalone Solr mode
I've been thinking about the fact that standalone and cloud modes in
Solr are very different.
The writing on the wall suggests that Solr will eventually (probably 7.0
minimum) eliminate the standalone mode and always operate with
zookeeper. A "standalone" node would in fact be a single-node cloud
running the embedded zookeeper.
Once zk-as-truth becomes a reality, I can see a few advantages to always
running in cloud mode. The documentation can include one way to
accomplish basic tasks. The CoreAdmin API can be eliminated, and any
required functionality fully merged into the Collections API.
CloudSolrClient will work for all installations. A script that works
for cloud mode will also work for standalone mode, because that's just a
smaller cloud.
I was planning to open an issue to discuss and implement this. If
that's not a good idea, please let me know.
None of my main Solr installations are running in cloud mode, so the
removal of standalone mode will be an inconvenience for me, but I still
think it's the right thing to do in the long term.
Thanks,
Shawn
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org
Re: Speculating about the removal of the standalone Solr mode
Posted by Rida Benjelloun <ri...@gmail.com>.
Personnaly I Think that is very good Idea. In Constellio we use solr cloud
and zookeeper for sharding. I think that the way you suggest will
simplify and normalise the code.
Regards.
Le 2016-03-09 3:57 PM, "Shawn Heisey" <ap...@elyograg.org> a écrit :
> On 3/9/2016 10:43 AM, Joel Bernstein wrote:
> > From Alfresco's point of view this would be a bad thing. Alfresco uses
> > Solr in stand alone mode and has developed an entire sharding and
> > replication model that fit's the ECM use case. So being forced to have
> > ZooKeeper and Solr Cloud would not be ideal.
>
> I'm aware of the potential pain. Third-party Solr support and the
> documentation that goes with it might require significant changes -- but
> those changes will already be required if those packages want to add
> support for talking to SolrCloud in general.
>
> I firmly believe that "cloud mode only" is the way Solr is headed, and
> that once we reach the other side, Solr will be better, especially
> because of documentation and API consolidation.
>
> My intent would not be to force SolrCloud's built-in sharding on
> everyone. You could still do completely manual sharding and work with
> individual Solr nodes like before. The difference would be that each
> "standalone" Solr node would internally use zookeeper (probably the
> embedded server) to manage itself. We might need to invent a
> "standalone collection" concept that could be used for
> single-shard-single-replica collections, where the core name and the
> collection name are the same, instead of cores named foo_shardN_replicaN.
>
> I myself would feel a lot of the pain you mentioned in relation to
> Alfresco. I'm also manually managing shards on standalone Solr
> instances. I've got a significant investment in a SolrJ application to
> handle these indexes.
>
> The change should not happen before 7.0, and with a major change like
> that on the horizon, the major version *before* the change (such as 6.x)
> should remain the stable branch for quite a while, so everybody has time
> to update and support cloud mode before it becomes mandatory. There
> will be a lot of details to iron out. I hope to be able to help with that.
>
> Thanks,
> Shawn
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
> For additional commands, e-mail: dev-help@lucene.apache.org
>
>
Re: Speculating about the removal of the standalone Solr mode
Posted by Shawn Heisey <ap...@elyograg.org>.
On 3/9/2016 10:43 AM, Joel Bernstein wrote:
> From Alfresco's point of view this would be a bad thing. Alfresco uses
> Solr in stand alone mode and has developed an entire sharding and
> replication model that fit's the ECM use case. So being forced to have
> ZooKeeper and Solr Cloud would not be ideal.
I'm aware of the potential pain. Third-party Solr support and the
documentation that goes with it might require significant changes -- but
those changes will already be required if those packages want to add
support for talking to SolrCloud in general.
I firmly believe that "cloud mode only" is the way Solr is headed, and
that once we reach the other side, Solr will be better, especially
because of documentation and API consolidation.
My intent would not be to force SolrCloud's built-in sharding on
everyone. You could still do completely manual sharding and work with
individual Solr nodes like before. The difference would be that each
"standalone" Solr node would internally use zookeeper (probably the
embedded server) to manage itself. We might need to invent a
"standalone collection" concept that could be used for
single-shard-single-replica collections, where the core name and the
collection name are the same, instead of cores named foo_shardN_replicaN.
I myself would feel a lot of the pain you mentioned in relation to
Alfresco. I'm also manually managing shards on standalone Solr
instances. I've got a significant investment in a SolrJ application to
handle these indexes.
The change should not happen before 7.0, and with a major change like
that on the horizon, the major version *before* the change (such as 6.x)
should remain the stable branch for quite a while, so everybody has time
to update and support cloud mode before it becomes mandatory. There
will be a lot of details to iron out. I hope to be able to help with that.
Thanks,
Shawn
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org
Re: Speculating about the removal of the standalone Solr mode
Posted by Joel Bernstein <jo...@gmail.com>.
>From Alfresco's point of view this would be a bad thing. Alfresco uses Solr
in stand alone mode and has developed an entire sharding and replication
model that fit's the ECM use case. So being forced to have ZooKeeper and
Solr Cloud would not be ideal.
Joel Bernstein
http://joelsolr.blogspot.com/
On Wed, Mar 9, 2016 at 12:34 PM, Shawn Heisey <ap...@elyograg.org> wrote:
> I've been thinking about the fact that standalone and cloud modes in
> Solr are very different.
>
> The writing on the wall suggests that Solr will eventually (probably 7.0
> minimum) eliminate the standalone mode and always operate with
> zookeeper. A "standalone" node would in fact be a single-node cloud
> running the embedded zookeeper.
>
> Once zk-as-truth becomes a reality, I can see a few advantages to always
> running in cloud mode. The documentation can include one way to
> accomplish basic tasks. The CoreAdmin API can be eliminated, and any
> required functionality fully merged into the Collections API.
> CloudSolrClient will work for all installations. A script that works
> for cloud mode will also work for standalone mode, because that's just a
> smaller cloud.
>
> I was planning to open an issue to discuss and implement this. If
> that's not a good idea, please let me know.
>
> None of my main Solr installations are running in cloud mode, so the
> removal of standalone mode will be an inconvenience for me, but I still
> think it's the right thing to do in the long term.
>
> Thanks,
> Shawn
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
> For additional commands, e-mail: dev-help@lucene.apache.org
>
>