Disk-to-heap ratio for Frozen Indices?

The Elasticsearch documentation states...

Because Elasticsearch does not maintain these transient data structures in memory, frozen indices consume much less heap than normal indices. This allows for a much higher disk-to-heap ratio than would otherwise be possible.

With a 30GB heap and shards in the 20-40GB range, I believe the recommendation for non-frozen shards per node would be 600 shards (20 shards/1GB of heap). I have not seen anything documented on what is a reasonable ratio for frozen indices. How many frozen shards can a node with a 30GB heap hold?

Our observation; we have to keep the sum of thawed shard size under 9G but we are OK with 1800 shards (and growing) on cold nodes. We have about 10G (also growing) in frozen space used.

I don't have a stat gathered on the number of frozen vs. thawed shards.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.