Recently I have found out that on one of the servers my ES create
additional node few days after launch, then one more 2 days later. ES on
this host was alone node not connected to any cluster. I have already found
out that these are worker nodes and number of all nodes on this host is
exact equivalent of max number of nodes per host . They were created
exactly at midnight. From logs I see that it's right before performing
daily reindexing from DB (about 2000 docs, so tiny number).
I've got some questions related to that for which I couldn't google the
answer:
What could cause these nodes to be created (and what usually is the
cause - is there any documentation about that)? Was that reindexing init?
(actual reindexing is logged after creating a node)
Does number of nodes shrink under any circumstances?
I'm wondering ... that host was almost completely idle and only daily
reindexing was some work, but still small considering index's size.
Considering load one node is fairly enough to handle that, so why add
more? If it could be caused just by splitting shards and creating replicas
to have a 'healthy' cluster, why those nodes were added so late? This all
just doesn't add up.
The only time a new node will automatically join a cluster is if it has the
same cluster.name and you are using multicast discovery.
ES does not automatically create nodes for the cluster, there must be
something else happening on your end.
Recently I have found out that on one of the servers my ES create
additional node few days after launch, then one more 2 days later. ES on
this host was alone node not connected to any cluster. I have already found
out that these are worker nodes and number of all nodes on this host is
exact equivalent of max number of nodes per host . They were created
exactly at midnight. From logs I see that it's right before performing
daily reindexing from DB (about 2000 docs, so tiny number).
I've got some questions related to that for which I couldn't google the
answer:
What could cause these nodes to be created (and what usually is the
cause - is there any documentation about that)? Was that reindexing init?
(actual reindexing is logged after creating a node)
Does number of nodes shrink under any circumstances?
I'm wondering ... that host was almost completely idle and only daily
reindexing was some work, but still small considering index's size.
Considering load one node is fairly enough to handle that, so why add
more? If it could be caused just by splitting shards and creating replicas
to have a 'healthy' cluster, why those nodes were added so late? This all
just doesn't add up.
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.