I have the following setup:
- 6 shards on a collection called "jobs-09022020"
- 6 TLOG replicas (used only for indexing)
- 6 PULL replicas (used only for searching)
I want to scale the PULL replicas, 6 at a time, and have Solr add replicas
to that new node set when they come up and delete them when that node set
goes down. I have uploaded and verified the following autoscale policy to
the /api/cluster/autoscaling API:
{
"set-trigger": {
"name": "node_added_trigger",
"event": "nodeAdded",
"waitFor": "5s",
"preferredOperation": "ADDREPLICA",
"replicaType": "PULL",
"enabled": true,
"actions": [
{
"name": "compute_plan",
"class": "solr.ComputePlanAction",
"collections": "jobs-09022020"
},
{
"name": "execute_plan",
"class": "solr.ExecutePlanAction"
}
]
},
"set-trigger": {
"name": "node_lost_trigger",
"event": "nodeLost",
"waitFor": "90s",
"preferredOperation": "DELETENODE",
"enabled": true
}
}
However, when I add the new nodes, no new replicas are created. What am I
missing? Here is the output of the /admin/collections?action=COLSTATUS API
(minus the "shards" information":
"jobs-09022020":{
"stateFormat":2,
"znodeVersion":135,
"properties":{
"autoAddReplicas":"false",
"maxShardsPerNode":"1",
"nrtReplicas":"3",
"policy":"search-blue",
"pullReplicas":"2",
"replicationFactor":"3",
"router":{
"field":"id",
"name":"compositeId"
},
"tlogReplicas":"2"
},
"activeShards":6,
"inactiveShards":0,
"schemaNonCompliant":["(NONE)"]
}
My questions:
1. Should I expect to see individual shard replicas coming up as the nodes
come up or do all the nodes have to be available before the shard placement
can happen?
2. What am I missing on the collection to allow for new PULL replicas to
come up via the autoscaling trigger I have above?
Note: this is very similar to Shane Brook's question regarding "Replication
not occurring to newly added SOLRCloud nodes" previously posted to this
list.
--
Drew(i...@gmail.com)
http://wyntermute.dyndns.org/blog/
-- I Drive Way Too Fast To Worry About Cholesterol.
