I'm seeing some strange usage patters on dedicated master nodes in a 5.6.16 Elasticsearch cluster.
Master nodes seem to be having excessive heap pressure during some workloads, causing them to crash with OOM errors. When fetching the tasks for the current master with
GET _tasks?nodes=LI7yjqmNRUiHfHzanCQmeQ&actions=*bulk&detailed
some bulk write tasks appear, but without any details about the index or the exact query.
{
"nodes": {
"LI7yjqmNRUiHfHzanCQmeQ": {
"name": "xxx_master-1",
"transport_address": "10.0.201.159:9300",
"host": "10.0.201.159",
"ip": "10.0.201.159:9300",
"roles": [
"master"
],
"attributes": {
"ml.max_open_jobs": "10",
"ml.enabled": "true"
},
"tasks": {
"LI7yjqmNRUiHfHzanCQmeQ:100703207": {
"node": "LI7yjqmNRUiHfHzanCQmeQ",
"id": 100703207,
"type": "transport",
"action": "indices:data/write/bulk",
"description": "",
"start_time_in_millis": 1632838233217,
"running_time_in_nanos": 256009211,
"cancellable": false
}
}
}
}
}
I have two concerns, why aren't there any details in the response, and why are there bulk write tasks running on a dedicated master? Any ideas?
Thanks,
Nikola