Elasticsearch bulk queue unbalanced, one node close to full


(左茫) #1

Recently we found one issue: our es cluster had three nodes which storage our product env log, one node of cluster bulk queue size always close to full, other nodes not.

i adjust our send client config file, remove the node and watch it. during some times, i found the node bulk queue size still close to full size.

Who can help solve this problem?

curl -s localhost:9200/_cat/thread_pool/bulk?h=id,host,name,active,queue,queue_size,rejected
node1 xxx bulk 0 0 3000 0
node2 xxx bulk 0 0 3000 0
node3 xxx bulk 9 1525 3000 1287


(Mark Walkom) #2

How are you sending logs to Elasticsearch?


(Christian Dahlqvist) #3

What is the output of the cat nodes API?


(左茫) #4

send client written by ourself, es driver is olivere/elastic.v5


(左茫) #5

sapadmin@ip-192-168-0-170 ~ $ curl localhost:9200/_cat/nodes
xxx 66 99 4 0.28 0.26 0.30 mdi * xxxx
xxx 59 93 9 0.34 0.52 0.60 mdi - xxxx
xxx 57 98 21 1.50 1.49 1.56 mdi - xxxx

all instance type is 8c,30m


(Mark Walkom) #6

Does that client do automatic load balancing?


(Christian Dahlqvist) #7

Are you inserting documents or performing updates? How many shards are you actively indexing into? How are these distributed across the cluster?


(system) #8

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