Hi all,
We've been facing some trouble with our Elastic Search installation (v
1.5.1), mostly trying
to bring it back up. Some questions have come up.
This is our situation. We're seeing about 200 unassigned shards, which are
not being
reassigned automatically, which in turn leads our ES to move into a red
status. In this state,
queries are disabled, however we keep storing data.
What we see is that this generates refresh/update mapping tasks which are
never resolved (we think
this is because ES is in a red state). Hoping to solve this, we've been
running _cluster/reroute on primary
shards:
curl -XPOST 'localhost:9290/_cluster/reroute' -d '{
"commands": [
{
"allocate": {
"index": "'$INDEX'",
"shard": '$SHARD',
"node": "'$NODE'",
"allow_primary": true
}
}
]
}'
Commands such as _nodes/_local/{stats, settings, os, processes} will fail
on the master node (hang indefinetly)
We monitor the pending tasks (see _cat/pending_tasks and _cluster/health
output below) and see that
the IMMEDIATE tasks are queued.
However, we're wondering what's the rationale behind the queuing of this
tasks:
Is there a round robin mechanism for the IMMEDIATE or is there some way of
prioritizing the
health state of the cluster over other tasks? Will an IMMEDIATE task
preempt any other (e.g URGENT or HIGH?)
We've noticed that when queuing two IMMEDIATE tasks, the second one may
timeout if the first one
is not resolved fast enough. Is this queue being consumed by a single
thread? If so, anyway to change that?
Thanks in advance!
Tomás
_cat/pending_tasks
% Total % Received % Xferd Average Speed Time Time Time
Current
Dload Upload Total Spent Left
Speed
0 0 0 0 0 0 0 0 --:--:-- --:--:-- --:--:--
0220771 2.1m IMMEDIATE cluster_reroute (api)
220891 29.8s IMMEDIATE cluster_reroute (api)
196772 10.1h HIGH update-mapping
[beer.raw.cl.business.2015-05-04][GDS_search_scans] / node
[v5SxZ7CdRou13tzy-N1DJg], order [1109]
220892 25.9s IMMEDIATE cluster_reroute (api)
196773 10.1h HIGH update-mapping
[beer.raw.pe.business.2015-05-04][GDS_search_scans] / node
[BTYaSC3cT8K_3xHDQYoNXQ], order [419]
196779 10.1h HIGH update-mapping
[beer.raw.ar.business.2015-05-04][prism_retrieve] / node
[iDVZlJycRdeOa1PGB4Oi9Q], order [127]
220893 25.7s IMMEDIATE cluster_reroute (api)
196787 10.1h HIGH refresh-mapping
[beer.raw.pt.business.2015-05-03][[GDS_search_scans]]
196786 10.1h HIGH refresh-mapping
[beer.raw.ca.business.2015-05-03][[GDS_search_scans]]
196774 10.1h HIGH update-mapping
[beer.raw.pe.business.2015-05-04][GDS_search_scans] / node
[Kx-HMg4qQKqepJb1qjjS3A], order [151]
196790 10.1h HIGH refresh-mapping
[beer.raw.ae.search.2015-05-03][[vito]]
196792 10.1h HIGH refresh-mapping
[beer.raw.tr.business.2015-05-03][[GDS_search_scans]]
218944 35.5m URGENT shard-started
([beer.raw.gy.performance.2015-04-07][2], node[BTYaSC3cT8K_3xHDQYoNXQ],
relocating [0clH8MU6Q5Wt8phPRbuTLg], [P], s[INITIALIZING]), reason [after
recovery (replica) from node
[[beer-elastic-s-08][0clH8MU6Q5Wt8phPRbuTLg][beer-elastic-s-08][inet[/10.70.163.240:9300]]{master=false}]]
220894 25.7s IMMEDIATE cluster_reroute (api)
196850 10.1h HIGH refresh-mapping
[beer.raw.fi.business.2015-05-03][[GDS_search_scans]]
196788 10.1h HIGH refresh-mapping
[beer.raw.il.business.2015-05-03][[GDS_search_scans]]
196789 10.1h HIGH refresh-mapping
[beer.raw.nl.business.2015-05-03][[GDS_search_scans]]
Health
_cluster/health?pretty=true"
"cluster_name" : "beer-elastic",
"status" : "red",
"timed_out" : false,
"number_of_nodes" : 20,
"number_of_data_nodes" : 12,
"active_primary_shards" : 32192,
"active_shards" : 64384,
"relocating_shards" : 2,
"initializing_shards" : 14,
"unassigned_shards" : 182,
"number_of_pending_tasks" : 13686
--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/e3f4f449-d359-4ef1-a6a4-a445bb916371%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.