When I try to get the cluster state size via GET _nodes/stats in kibana dev tools it gives me this data from the node, which is holding cluster state now
but when I check it on any node using command curl -X GET "https://<address_here>/_cluster/state?pretty" -k -u "<creds_here>" -o cluster_state.json
it gives me cluster_state.json with size of 1013867688 ~ 1GB
and if try to take it without pretty it will be 323170792 ~ 323MB
So this value is nowhere near 11GB that is in serialized_cluster_states
Sure but in what sense? The size on the wire between nodes is the compressed size (average ~5.2MiB according to your numbers). On disk it's stored as a collection of files in ${path.data}/_state with a different format so I expect it's a little larger. Rendered as JSON it's much larger than either of these. Which of these is the "actual size" in the sense in which you're interested?
I don't think this is documented, no, it's only really useful for some fairly deep troubleshooting when read alongside the code itself, it doesn't have much meaning otherwise.
In a matter of size I am interested in the current size of the cluster state, which is located in ${path.data}/_state, but in a matter of content (which is the thing that I am originally looking at and trying to find what's in there currently) I think the second one
And if we are talking about the cluster size itself - is there any documentation (not the unofficial one), where I can find what things can affect the cluster state and is there any way to find out the normal cluster state size (maybe calculate it somehow)?
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.