[
https://issues.apache.org/jira/browse/SOLR-5146?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13897763#comment-13897763
]
Shalin Shekhar Mangar commented on SOLR-5146:
---------------------------------------------
Sure Erick. I've been studying the code for transient cores to understand the
potential issues and bottlenecks with using this feature with SolrCloud.
I think we can break it down to four major features:
# Make loadOnStartup=true work with SolrCloud shards - slices are marked with
loadOnStartup=false. All nodes are woken up on a request.
# Make transient replicas work with SolrCloud replication - Leader is always
active, replicas are down. Leader can send 'requestrecovery' to replicas based
on maxDocs/maxTime parameters to make them sync. Maybe we can make peersync and
buffer doc counts configurable.
# Make transient leaders work with SolrCloud - Down leaders are okay. Replicas
may still be up but we won't force leader election. Leaders are woken up only
on a write request.
# Optimize leader election for transient shards - We probably don't want to
force leader election each time a shard wakes up. Instead clusterstate can
remain the truth and leaders can go down. If a shard is woken up again, it can
use the same leader until it goes down. This is far away. We shall focus on
optimization later.
I think an easy win here would be to translate loadOnStartup=false to complete
shards (leader+replicas). I'm going to start building a prototype and see how
easy it turns out to be :)
> Figure out what it would take for lazily-loaded cores to play nice with
> SolrCloud
> ---------------------------------------------------------------------------------
>
> Key: SOLR-5146
> URL: https://issues.apache.org/jira/browse/SOLR-5146
> Project: Solr
> Issue Type: Improvement
> Components: SolrCloud
> Affects Versions: 4.5, 5.0
> Reporter: Erick Erickson
> Assignee: Shalin Shekhar Mangar
> Fix For: 5.0
>
>
> The whole lazy-load core thing was implemented with non-SolrCloud use-cases
> in mind. There are several user-list threads that ask about using lazy cores
> with SolrCloud, especially in multi-tenant use-cases.
> This is a marker JIRA to investigate what it would take to make lazy-load
> cores play nice with SolrCloud. It's especially interesting how this all
> works with shards, replicas, leader election, recovery, etc.
> NOTE: This is pretty much totally unexplored territory. It may be that a few
> trivial modifications are all that's needed. OTOH, It may be that we'd have
> to rip apart SolrCloud to handle this case. Until someone dives into the
> code, we don't know.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
