lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "scott.chu" <>
Subject Re(2): Can "Using cp replica and modifying" rather thanADDREPLICA API work?
Date Tue, 31 May 2016 03:43:50 GMT
Thanks for your advice, Erick. I think you point out what I didn't think of and a possible
side effect in the future. I'll go back to the "normal" way  next time I do the same job.

2016/5/31 (週二)
----- Original Message ----- 
From: Erick Erickson 
To: solr-user ; scott(自己) 
Date: 2016/5/31 (週二) 11:12
Subject: Re: Can "Using cp replica and modifying" rather thanADDREPLICA API

Well, that'll work, but you better know _exactly_ what you're doing. 
For instance, you better not be indexing and have committed before 
you start your copy. You better make sure your third node 
is up before you index anything. Etc. Etc. 

Why do you think this "saves time"? Have you measured? Is the 
time savings worth the risk? Because using ADDREPLICA handles all 
the edge cases for you and essentially does what you're talking about 
behind the scenes. And will work even if you're actively indexing. 

Manually copying things seems to be adding additional places for you 
to get it wrong for a pretty rare operation, I'd bet that the time you spend 
manually creating a replica could be better spent letting the ADDREPLICA 
run in the background while you do other more important things. 


On Thu, May 26, 2016 at 7:33 PM, scott.chu <> wrote: 

> On my lab under Windows PC: 
> 2 Solrcloud nodes, 1 collection, named cugna, with numShards=1 and replicationFactor=2,
add index up to 90GB 
> After it worked, I migrate them to CentOS (1 node 1 machine) but I want to add 3rd node
to 3rd machine. I think there's only 1 shard and replicationFactor is only a "startup" parameter,
not a "limitation". So I do these tasks: 
> * Copy node 2's solr to 3rd machine 
> * Go into solr.home 
> * Rename folder 'cugna_shard1_replica2' to 'cugna_shard1_replica3' 
> * Go into 'cugna_shard1_replica3' foler, edit by 
> change 'name' parameter to 'cugna_shard1_replica3' 
> change 'coreNodeName' parametere to 'core_node3' 
> Then Start 3 nodes, they look ok when I go to admin UI to see the cloud diagram. 
> However, I'm wondering if this gonna be ok or if there's something might cause inconsistency
that doesn't show on admin ui? 
> p.s. I did this because I want to save the time to create a new replica. 

> scott.chu, 
> 2016/5/27 (週五) 

已透過 AVG 檢查 - 
版本: 2015.0.6201 / 病毒庫: 4591/12331 - 發佈日期: 05/30/16 
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message