[Management] List / Terminate Running Tasks

[AlonaNadler]

It was mentioned by a user, occasionally users run very heavy queries without being aware e.g. run a query for all data for the last 7 days, they want a way for the administrator to be able to terminate it.

Currently, there is no way to terminate the query. ES added an ID to all task which allows you to terminate specific task elastic/elasticsearch#27764
Initially, I raised that request with the Discover team but it makes much more sense to have that as an admin tool to allow the administrator to kill heavy queries.
This includes a page with a list of running tasks which then the admin can decide to terminate.

@elastic/kibana-monitoring

[pickypg]

This seems more like a management task to me. The tasks API does not report the entire query (at least not currently), and monitoring -- under best practices with a separate, dedicated monitoring cluster -- implies that there's nothing that you can do with the knowledge beyond trying to not repeat it.

However, if Management provides a Task listing, then a user could pick and choose between tasks live when it's most relevant. Longer term, I think that query performance makes perfect sense (either in combination with APM or by reading from the slowlog with Filebeat), but the most immediate need is a list / button in Management in my opinion.

[AlonaNadler]

Following a meeting with the Elastic product team, support frequently need to help users and figure which query slow the cluster or slow specific dashboard,
Going a step forward, If we have a task listing page it will help to have a specific Kibana ID or any other metadata which can help correlate a task to what triggered it in Kibana e.g specific visualization/dashboard

[elasticmachine]

Pinging @elastic/kibana-stack-management (Team:Stack Management)

[cjcenizal]

See #42621 for screenshots of a prototype of this UI.

See #18411 for original discussion.

[cjcenizal]

Relates to #44193

[matschaffer]

nice to see at least @cjcenizal is still thinking of this. _tasks?detailed is still basically the first place I look when trying to figure out why an ES cluster is slow even today.

[mbarretta]

Still a common ask in the field

[Aubermean]

+1 on this