Hi, this is a very old topic, for which the answer is that balancing primary shard is not useful, mainly because indexing and replication cost more or less the same.
However, there is a case where it matters: percolation. Our elasticsearch infrastructure is composed of two nodes. Indexing incomming documents works like a charm, but we also have around 100 percolators, some of them involving fuzzy matches on thousands strings. Percolation requires a lot of cpu, a lot more than indexing. There are very few search requests, thus their impact on the cpu is negligeable. We observe that, when all primary shards are gathered on one node (e.g. badly balanced), this node has a very high cpu load, while the other is just quiet. When primary shards are equally balanced, we observe a similar load on the two nodes. #3293 has been updated (I cannot reopen it). Thanks ik -- You received this message because you are subscribed to the Google Groups "elasticsearch" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/af6a8982-6ec2-4535-8648-65fec227b7a0%40googlegroups.com. For more options, visit https://groups.google.com/groups/opt_out.
