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 Adarsh Hd <hd...@infor.com> on 2018/06/15 10:25:27 UTC

Soft commit impact on replication

Hi All,

Current am using SOLR 5.2.1 on Linux machine.  I have cluster of 5 nodes with master and salve configuration, which gives 5 master nodes and 5slave node. We have enabled only hard commit on master nodes and both soft & hard commit on the slave nodes since the search will happen on slave.
Current we are facing a issue which is causing our slave to replicate full copy from master most of the time , sometimes I get error like unable to clean old index so trigger full copy I could understand that from couple of solr links, but there is lot of time where the full copy is triggered just like with out any warning or any error message could anyone let me know the possibilities that might trigger the full copy replication so often.


Regards
Adarsh

Re: Soft commit impact on replication

Posted by Adarsh_infor <hd...@infor.com>.
Hi Erick,

Thanks for the response. 

First thing we not indexing on Slave.  And we are not re-indexing/optimizing
entire the core in Master node. 

The only warning which I see in the log is "Unable clean the unused index
directory so starting full copy".  
That one i can understand and I don't have issue with that as its normal
behaviour but most of the time. But most of the time it’s just triggers
full-copy without any details in the log.  

And recently in one of the nodes i enabled soft-commit in master nodes and
monitored the corresponding slave node, what i observed is it didn't even
trigger the full-copy not even once for almost 3 consecutive days. So am
wondering do we need to have soft commit enabled in master for replication
to happen smooth if so what’s the dependency there


Thanks 



--
Sent from: http://lucene.472066.n3.nabble.com/Solr-User-f472068.html

Re: Soft commit impact on replication

Posted by Erick Erickson <er...@gmail.com>.
My first guess is that you're indexing to the slave nodes.

Second guess is that you're re-indexing your entire corpus on the master node.

Third guess is that you're optimizing on the master node (don't do this)

What does the slave's log say is the reason? If all the segments on
the master have changed, they'll all need to be copied during
replication so if that's the case that's entirely normal.

And you shouldn't need to commit on the slaves, that should happen as
part of replication.

Best,
Erick

On Fri, Jun 15, 2018 at 3:25 AM, Adarsh Hd <hd...@infor.com> wrote:
> Hi All,
>
> Current am using SOLR 5.2.1 on Linux machine.  I have cluster of 5 nodes with master and salve configuration, which gives 5 master nodes and 5slave node. We have enabled only hard commit on master nodes and both soft & hard commit on the slave nodes since the search will happen on slave.
> Current we are facing a issue which is causing our slave to replicate full copy from master most of the time , sometimes I get error like unable to clean old index so trigger full copy I could understand that from couple of solr links, but there is lot of time where the full copy is triggered just like with out any warning or any error message could anyone let me know the possibilities that might trigger the full copy replication so often.
>
>
> Regards
> Adarsh