I¹m looking at dashboard page on all 4 nodes and seeing Physical Memory 92% compared with ~41-44%
And JVM-Memory 52.9% compared to 23-28% The reason I mentioned slave is that on the core overview page there is An entry for Slave (Searching) that doesn¹t appear on any of the other nodes Cheers, David On 24/03/2014 14:47, "Shawn Heisey" <s...@elyograg.org> wrote: >On 3/24/2014 7:15 AM, David Flower wrote: >> We have a 4 node cluster with a collection thats sharded into 2 and each >> shard having a master and a slave for redundancy however 1 node has >>decied >> to use twice the ram that the others are using within the cluster >> >> The only difference we can spot between the node is that the one with >>the >> ram usage is saying its a slave while all the other are reporting that >> they are masters > >If you are using SolrCloud, then there are no masters and no slaves. >Each shard has a leader, but that is not a permanent role. > >The master and slave designations that you see on the Replication tab >have zero meaning in SolrCloud unless a replication happens to be >happening right at that moment. In SolrCloud, replication is only used >at node startup, and only if it's required. The master/slave roles are >decided at the moment of replication and are not changed until another >replication becomes necessary. > >When you say it's using twice the RAM, what *precisely* are you looking >at which tells you this? Due to Solr using MMap for file access, some >of the numbers reported by the operating system will look bad but will >not indicate a problem. > >Thanks, >Shawn >
