Bug #21611: preemption notices do not appear in crunch-run.txt - Arvados

Custom queries

All assigned issues
All issues assigned for grooming
My issues for grooming
My issues for grooming (no story pts)
Prioritized open issues

Actions

Copy link

Bug #21611

closed

preemption notices do not appear in crunch-run.txt

Added by Peter Amstutz 8 months ago. Updated 3 months ago.

Status:

Resolved

Priority:

Normal

Assigned To:

Tom Clegg

Category:

Crunch

Target version:

Development 2024-05-22 sprint

Story points:

Release:

Arvados 3.0

Release relationship:

Auto

Description

I've looked at a number of containers now that have been preempted and none of them have crunch-run.txt updated to say that it received a preemption notice even though it is supposed to.

Subtasks 1 (0 open — 1 closed)

Task #21734: Review 21611-log-chunk-delay

Resolved

Tom Clegg

05/15/2024

Actions

Related issues

	Related to Arvados - Bug #21833: Regression: stdout and stderr in log collection missing line-by-line timestamps	Duplicate	Tom Clegg				Actions
	Related to Arvados - Bug #21834: Restore timestamps in container stdout.txt and stderr.txt	Resolved	Tom Clegg				Actions

Issue # Delay: days Cancel

History
Notes
Property changes
Associated revisions

Actions

Copy link

Updated by Peter Amstutz 8 months ago

Target version changed from Future to Development 2024-04-24 sprint

Actions

Copy link

Updated by Peter Amstutz 8 months ago

Description updated (diff)

Actions

Copy link

Updated by Peter Amstutz 8 months ago

Target version changed from Development 2024-04-24 sprint to Development 2024-05-08 sprint

Actions

Copy link

Updated by Peter Amstutz 8 months ago

Target version changed from Development 2024-05-08 sprint to Development 2024-04-24 sprint

Actions

Copy link

Updated by Peter Amstutz 8 months ago

Target version changed from Development 2024-04-24 sprint to Development 2024-05-08 sprint

Actions

Copy link

Updated by Peter Amstutz 7 months ago

Description updated (diff)
Subject changed from crunch-run updates copy of container.json in log collection when a container ends and/or runtime_status is updated to preemption notices do not appear in crunch-run.txt
Tracker changed from Feature to Bug

Actions

Copy link

Updated by Peter Amstutz 7 months ago

Assigned To set to Tom Clegg

Actions

Copy link

Updated by Tom Clegg 7 months ago

Status changed from New to In Progress

I suspect the "write log, then save log collection" is doing the opposite of what we want, because

"write log entry" just writes the message to the throttled-logging buffer, not yet to the log collection
"save log collection" saves the log collection and resets the auto-flush timer, minimizing the chance auto-flush will happen before the preempted instance shuts down

It's possible something else is going on too, but either way, we should rearrange the logging pipeline so the log collection gets updated immediately instead of after the "group logs into chunks" step. If nothing else, that will reduce latency for showing logs in workbench.

Actions

Copy link

Updated by Tom Clegg 7 months ago

21611-log-chunk-delay @ 8a5db7b48c1fb11423110490267fea17161f7674 -- developer-run-tests: #4207 (flaky fuse test, see #21660)

21611-log-chunk-delay @ 8a5db7b48c1fb11423110490267fea17161f7674 -- developer-run-tests: #4208 (Something is already running on port 38402.)

21611-log-chunk-delay @ 8a5db7b48c1fb11423110490267fea17161f7674 -- developer-run-tests: #4209

Removes all the "buffer logs into chunks and send them to POST /arvados/v1/logs" code that was preventing the existing "flush logs immediately" code from working as intended (see #note-8 above).

All agreed upon points are implemented / addressed.
- ✅
Anything not implemented (discovered or discussed during work) has a follow-up story.
- N/A
Code is tested and passing, both automated and manual, what manual testing was done is described
- ✅ updated preemption-warning test case to check that the container record is promptly updated with a log PDH that mentions the preemption warning message
Documentation has been updated.
- N/A
Behaves appropriately at the intended scale (describe intended scale).
- N/A
Considered backwards and forwards compatibility issues between client and server.
- N/A
Follows our coding standards and GUI style guidelines.
- ✅

This will also have the side effect of reducing logging latency in workbench. Previously LogBytesPerEvent/LogSecondsBetweenEvents (default 4K/5s) were introducing a store/wait/forward delay even when LimitLogBytesPerJob was zero.

Actions

Copy link

#10

Updated by Peter Amstutz 7 months ago

Target version changed from Development 2024-05-08 sprint to Development 2024-05-22 sprint

Actions

Copy link

#11

Updated by Brett Smith 6 months ago

Tom Clegg wrote in #note-9:

21611-log-chunk-delay @ 8a5db7b48c1fb11423110490267fea17161f7674 -- developer-run-tests: #4209

Removes all the "buffer logs into chunks and send them to POST /arvados/v1/logs" code that was preventing the existing "flush logs immediately" code from working as intended (see #note-8 above).

LGTM. My one nit is I think the configuration keys in the upgrade notes would look better in monospace. (I wish we had a documentation style guide to help us keep consistent on stuff like this.)

Actions

Copy link

#12