Collect separate statistics for failed tasks

[losipiuk]

Description

Collect separate statistics for failed tasks.
Show aggregated CPU/Scheduled time and cumulative memory for failed tasks in Web UI.

Is this change a fix, improvement, new feature, refactoring, or other?

improvement

Is this a change to the core query engine, a connector, client library, or the SPI interfaces? (be specific)

query engine/UI

Related issues, pull requests, and links

fixes: #10734
improves on: #10754

Documentation

(x) No documentation is needed.
( ) Sufficient documentation is included in this PR.
( ) Documentation PR is available with #prnumber.
( ) Documentation issue #issuenumber is filed, and can be handled later.

Release notes

( ) No release notes entries required.
(x) Release notes entries required with the following suggested text:

# UI
* Add CPU time, scheduled time, and cumulative memory statistics regarding failed tasks in a query to web UI. ({issue}`10754`)

# General
* Add CPU time, scheduled time, and cumulative memory statistics regarding failed tasks in a query to query-completion event. ({issue}`10734`)

[losipiuk]

This is initial work towards #10734.

An alternative approach is in #11317 but I like this one better. It is a bit more code but it is more straightforward as we are only adding fields we need to. The #11317 added whole QueryStats/BasicQueryStats instance to count stats for a subset of tasks but many fields in the object only make sense if we are in the context of whole query.

[losipiuk]

@arhimondr please skim if it looks like a good direction to you.

[arhimondr]

nit: maybe "wasted" (here and in other places)?

[losipiuk]

I personally like failed more as it is more precise and does not add emotional meaning to json field. But I do not feel super strongly. @martint / @findepi opinion?

[findepi]

i'd prefer "failed" too.

btw do you plan anything like speculative execution, where work could become "wasted" without being "failed"?

[findepi]

eg "cancelled"

[losipiuk]

Yeah - that will come up. But even then "wasted" is wrong term really. It has negative connotation, and if we are up to kill some tasks to speed up the execution of whole query it should not be seen as we are doing something wrong. And "wasting" resources feels wrong. So maybe some other term - neither "failed" nor "wasted" - but nothing comes to my mind.

[martint]

FWIW, I think "failed" is ok. The failure could be intrinsic (cluster / resource / system issues) or induced (task was forced to abort). In either case, it failed to complete.