how elasticsearch cluster plan?
master data ingest coordinating
the four nodes . how about Proportion about the fours nodes ?
(as master:20% data: 60% ingest:10% coord:10%)
can give a suggested reference
thanks so much!!!~~~~~~~~~~~~~~~
With 4 nodes I'd keep all the default values.
Just because you can define dedicated node types does not mean you should. For clusters with 3-5 data nodes it is often fine to keep the default configuration (master eligible, holds data and supports ingest) for all the nodes. Make sure you set minimum_master_nodes correctly.
Once you reach a larger cluster size, you generally always want exactly 3 dedicated master nodes, even for very large clusters. The proportion of other node types depend on use case and requirements.
thanks
if have 100 nodes about 20T+ data one day
how about as master client ingest and coord ?
You will generally have 3 dedicated master nodes even for large clusters. When it comes to dedicated ingest nodes it depends in how much you ingest, how long you keep the data and how complex your pipelines are. There is no fixed ratio compared to the number of data ondes. Coordinating nodes may be useful, but are not necessarily required. I have seen large clusters only using a coordinating node per Kibana instance. Whether you actually need them at all and how many you need will therefore depend on the use case as well.
about bulk use tcp 9300 http 9200 to write data ,witch is better about tcp and http?
is as one kibana connect to one coording nodes?
Yes, deploying each Kibana instance together with a coordinating node is a common practice. For bulk indexing I would recommend using the HTTP interface as it has largely similar performance characteristics to the transport protocol.
in es 5.6.x...
the Java High Level REST Client is better than 5.4.3?
As far as I know the Java high-level REST client is only compatible with Elasticsearch 5.6 onwards.
This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.