Elasticsearch.md

_cat interface

Es provides various useful info under /_cat/ namespaces like that:

~ᛯ curl  localhost:9200/_cat/nodes
dev14.non.3dart.com 127.0.0.1 47 42 2.04 d m dev14 [ ex-cluste ]
dev13.non.3dart.com 127.0.0.1 71 48 0.77 d m dev13 [ ex-cluste ]
dev09.non.3dart.com 127.0.0.1 57 49 1.31 d m dev09 [ ex-cluste ]
dev11.non.3dart.com 127.0.0.1 63 50 0.86 d * dev11 [ ex-cluste ]
dev10.non.3dart.com 127.0.0.1 56 47 1.08 d m dev10 [ ex-cluste ]
dev12.non.3dart.com 127.0.0.1 56 42 0.68 d m dev12 [ ex-cluste ]

Adding '?v' will add colum names + padding, like that:

~ᛯ curl  localhost:9200/_cat/nodes?v
host                   ip        heap.percent ram.percent load node.role master name
dev14.non.3dart.com 127.0.0.1           46          42 1.78 d         m      dev14 [ ex-cluste ]

• /_cat/ - index of all cats
• /_cat/allocation - disk space/shard allocation summary
• /_cat/shards - list of all shards and their replicas
• /_cat/shards/{index} - list of shards belonging to index
• /_cat/master - master node
• /_cat/nodes - all nodes
• /_cat/indices - health and summary of all indices
• /_cat/indices/{index} - health and summary of one indice
• /_cat/segments - internal data about lucene stuff
• /_cat/segments/{index} - internal data about lucene stuff
• /_cat/recovery - state of recovery
• /_cat/recovery/{index} - state of recovery for one index
• /_cat/health - health + some basic data
• /_cat/pending_tasks - tasks
• /_cat/aliases - aliases
• /_cat/aliases/{alias} - one alias
• /_cat/thread_pool - count of active threads in pool
• /_cat/plugins - installed plugins

useful commands

• curl -XPOST 'http://localhost:9200/_cluster/nodes/_local/_shutdown?delay=60s' - notify cluster you are shutting down then exit after 60 seconds. instead of local you can specify nodes separated by , , attributes like rack:2 or keywords like _master or _all. * can be used too on any of those for example ra*:2*
• curl -XPUT localhost:9200/_cluster/settings -H 'Content-Type: application/json' -d '{"transient":{"cluster.routing.allocation.enable": "none"}}' - disable reallocating shards. Useful if you want to restart one of nodes without too much reshuffling
• curl -XPUT localhost:9200/_cluster/settings -H 'Content-Type: application/json' -d '{"transient":{"cluster.routing.allocation.enable": "all"}}' - re-enable reallocation
• curl -XPUT localhost:9200/_cluster/settings -H 'Content-Type: application/json' -d '{"transient":{"cluster.routing.allocation.cluster_concurrent_rebalance": "6"}}' - change concurrent rebalance limit for whole cluster
• curl -XPUT localhost:9200/_cluster/settings -H 'Content-Type: application/json' -d '{"transient":{"cluster.routing.allocation.node_concurrent_recoveries": 15}}' - change concurrent recoveries on node
• curl -XPOST localhost:9200/_cluster/reroute?retry_failed - retry failed shard reallocations
• curl 'http://localhost:9200/_cat/shards?v&h=index,node,shard,prirep,state,unassigned.reason - display reason why shard is unassigned
• curl -XGET "localhost:9200/_cluster/allocation/explain" - explain in details
• curl -XPUT localhost:9200/_cluster/settings -H 'Content-Type: application/json' -d '{"transient":{"cluster.routing.allocation.exclude._name":"node1,node2"}}' - exclude node1 and node2 from allocation, basically a soft node decommision. also works with _ip and _host and with globs.
• curl -XPUT localhost:9200/_cluster/settings -H 'Content-Type: application/json' -d '{"transient":{"cluster.routing.allocation.exclude._name":null}}' - remove exclude rule for node name
• curl -XPUT localhost:9200/_cluster/settings -H 'Content-Type: application/json' -d '{"transient":{"cluster.*":null}}' - wildcards work for resetting too
• curl -XPOST "http://localhost:9200/syslog-2018.08.06/_forcemerge?max_num_segments=1" - force full merge (removes deleted records) on index
• curl -XPOST localhost:9200/_cluster/reroute -H 'Content-Type: application/json' -d '{"commands":[{"move":{"index":"index-name","shard":6,"from_node":"node1","to_node":"node2"}}]}' - force shard relocation
• `curl -XPUT "localhost:9200/indexname/_settings?pretty" -H 'Content-Type: application/json' -d'{"index" : {"number_of_replicas" : 1}}'

change number of shards

for a in $(curl -s http://localhost:9200/_cat/shards |awk '{print $1}'|perl -ne 'if (length($_) > 20) {print $_}'|uniq|sort) ; do curl -XPUT "localhost:9200/$a/_settings?pretty" -H 'Content-Type: application/json' -d'
{
    "index" : {
        "number_of_replicas" : 0
    }
 }
' & sleep 0.4 ;done