Solr 4.9 - configs and collections

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Solr 4.9 - configs and collections

Abhi Basu
Running on MS HDInsight and Solr 4.9.What is the BKM for creation, update,
delete of configurations and collections?

I do the following:

1. First I create the zk config:
sudo zkcli.sh -cmd upconfig -zkhost zknode
<http://zk1-esohad.tzun3mpncofedp04lr3ird23xc.jx.internal.cloudapp.net/>:2181
-confdir /home/sshuser/ems-collection-49/conf/ -confname ems-collection

2. Then I create the collection:
curl '
http://headnode0:8983/solr/admin/collections?action=CREATE&name=ems-collection&numShards=2&replicationFactor=2&maxShardsPerNode=1
'

This works the first time. When I change the zk config, do I run the same
command #1? Also, do I do a reload:

curl '
http://headnode0:8983/solr/admin/collections?action=RELOAD&name=ems-collection
'

Or, do I need to delete and recreate the collection?


Very familiar with CDH solrctl commands that make life easier by only
having one command for this. Any help is appreciated.

Thanks,

Abhi

--
Abhi Basu
Reply | Threaded
Open this post in threaded view
|

Re: Solr 4.9 - configs and collections

Erick Erickson
Yes, use the same command to upload a config. Yes, you need to reload
a collection.

Lots of this functionality is in the newer bin/solr scripts, 4.9 is
3.5 years old.

Best,
Erick

On Mon, Mar 26, 2018 at 7:43 AM, Abhi Basu <[hidden email]> wrote:

> Running on MS HDInsight and Solr 4.9.What is the BKM for creation, update,
> delete of configurations and collections?
>
> I do the following:
>
> 1. First I create the zk config:
> sudo zkcli.sh -cmd upconfig -zkhost zknode
> <http://zk1-esohad.tzun3mpncofedp04lr3ird23xc.jx.internal.cloudapp.net/>:2181
> -confdir /home/sshuser/ems-collection-49/conf/ -confname ems-collection
>
> 2. Then I create the collection:
> curl '
> http://headnode0:8983/solr/admin/collections?action=CREATE&name=ems-collection&numShards=2&replicationFactor=2&maxShardsPerNode=1
> '
>
> This works the first time. When I change the zk config, do I run the same
> command #1? Also, do I do a reload:
>
> curl '
> http://headnode0:8983/solr/admin/collections?action=RELOAD&name=ems-collection
> '
>
> Or, do I need to delete and recreate the collection?
>
>
> Very familiar with CDH solrctl commands that make life easier by only
> having one command for this. Any help is appreciated.
>
> Thanks,
>
> Abhi
>
> --
> Abhi Basu
Reply | Threaded
Open this post in threaded view
|

Re: Solr 4.9 - configs and collections

Shawn Heisey-2
In reply to this post by Abhi Basu
On 3/26/2018 8:43 AM, Abhi Basu wrote:
> Running on MS HDInsight and Solr 4.9.What is the BKM for creation, update,
> delete of configurations and collections?

I have no idea what a BKM is.  I will cover the update of configuration
below.

> I do the following:
>
> 1. First I create the zk config:
> sudo zkcli.sh -cmd upconfig -zkhost zknode
> <http://zk1-esohad.tzun3mpncofedp04lr3ird23xc.jx.internal.cloudapp.net/>:2181
> -confdir /home/sshuser/ems-collection-49/conf/ -confname ems-collection

Exactly what you've got configured there for the zkhost parameter is
difficult to decipher because it looks like the hsotname got replaced
with a URL by your mail client.  But I think you've only got one ZK
server there.  Usually there are at least three of them.  The command
actually only needs one, but the zkHost string usually has at least
three.  It's generally a good idea to use the same string for zkcli that
you use for Solr itself, so it works even when a server is down.

> 2. Then I create the collection:
> curl '
> http://headnode0:8983/solr/admin/collections?action=CREATE&name=ems-collection&numShards=2&replicationFactor=2&maxShardsPerNode=1
> '
>
> This works the first time. When I change the zk config, do I run the same
> command #1? Also, do I do a reload:

Yes, if you want to change an existing config and then make it active,
you re-upload the config and then reload any affected collection. 
Deleting and recreating the collection is not something you would want
to do unless you plan to completely rebuild it anyway -- deleting the
collection will also delete all the index data.  If that's what you
WANT, then deleting and recreating the collection is a good way to make
it happen.  Many config updates *do* require a reindex, and some changes
will also require completely deleting the index directories before
building it again.

> Very familiar with CDH solrctl commands that make life easier by only
> having one command for this. Any help is appreciated.

If you're using CDH, you'll want to talk to Cloudera for help.  They
customize their Solr install to the point where they're the only ones
who know how to use it properly.

Thanks,
Shawn

Reply | Threaded
Open this post in threaded view
|

Re: Solr 4.9 - configs and collections

Abhi Basu
Thanks for the explanations, very helpful.

One more question, what is the sequence to delete the collection? If I use
the rest api to delete the collection, then when I go to create it again, I
sometimes get an error message saying shard already present. How to clean
up the underlying directories on all the nodes?

Thanks,

Abhi

On Mon, Mar 26, 2018 at 6:22 PM, Shawn Heisey <[hidden email]> wrote:

> On 3/26/2018 8:43 AM, Abhi Basu wrote:
> > Running on MS HDInsight and Solr 4.9.What is the BKM for creation,
> update,
> > delete of configurations and collections?
>
> I have no idea what a BKM is.  I will cover the update of configuration
> below.
>
> > I do the following:
> >
> > 1. First I create the zk config:
> > sudo zkcli.sh -cmd upconfig -zkhost zknode
> > <http://zk1-esohad.tzun3mpncofedp04lr3ird23xc.jx.internal.cloudapp.net/
> >:2181
> > -confdir /home/sshuser/ems-collection-49/conf/ -confname ems-collection
>
> Exactly what you've got configured there for the zkhost parameter is
> difficult to decipher because it looks like the hsotname got replaced
> with a URL by your mail client.  But I think you've only got one ZK
> server there.  Usually there are at least three of them.  The command
> actually only needs one, but the zkHost string usually has at least
> three.  It's generally a good idea to use the same string for zkcli that
> you use for Solr itself, so it works even when a server is down.
>
> > 2. Then I create the collection:
> > curl '
> > http://headnode0:8983/solr/admin/collections?action=
> CREATE&name=ems-collection&numShards=2&replicationFactor=
> 2&maxShardsPerNode=1
> > '
> >
> > This works the first time. When I change the zk config, do I run the same
> > command #1? Also, do I do a reload:
>
> Yes, if you want to change an existing config and then make it active,
> you re-upload the config and then reload any affected collection.
> Deleting and recreating the collection is not something you would want
> to do unless you plan to completely rebuild it anyway -- deleting the
> collection will also delete all the index data.  If that's what you
> WANT, then deleting and recreating the collection is a good way to make
> it happen.  Many config updates *do* require a reindex, and some changes
> will also require completely deleting the index directories before
> building it again.
>
> > Very familiar with CDH solrctl commands that make life easier by only
> > having one command for this. Any help is appreciated.
>
> If you're using CDH, you'll want to talk to Cloudera for help.  They
> customize their Solr install to the point where they're the only ones
> who know how to use it properly.
>
> Thanks,
> Shawn
>
>


--
Abhi Basu