New issue
Advanced search Search tips

Issue 916562 link

Starred by 2 users
Status: Available
Owner: ----
Cc:
mar...@chromium.org
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug

Blocked on: View detail
issue 729565
issue 901197
Blocking:
issue 916548



Sign in to add a comment

Swarming: explicitly shed load based on load estimator and pending queue

Project Member Reported by mar...@chromium.org, Dec 19 
Building upon the work of issue 729565, Swarming task creation should deny a task when the task_slice expiration is larger than pending_estimate:

  available_bots = all_bots_that_can_run(req.dimensions)
  pending_tasks = all_tasks_that_can_run_on_bots(available_bots)
  pending_estimate = sum(estimate(req) for req in pending_tasks) / len(available_bots)

plus a safety margin to not shed too aggressively. Note that this may interact in interesting ways with the LIFO scheduler; issue 901197. Marking LIFO as a blocker since this has interesting consequences and both must be sorted out.

In this case the task will immediately be denied with state LOAD_SHED.
Project Member

Comment 1 by bugdroid1@chromium.org, Dec 19 

The following revision refers to this bug:
  https://chromium.googlesource.com/infra/luci/luci-py.git/+/132c74e3ce003b29221bf3e5fc38af8c4f0bc049

commit 132c74e3ce003b29221bf3e5fc38af8c4f0bc049
Author: Marc-Antoine Ruel <maruel@chromium.org>
Date: Wed Dec 19 22:33:36 2018

[swarming] Further proto fine tuning

Lots of small things I realized after the fact. Doing these before piling more
data (TaskResult).

- Assign a bug to every TaskState not implemented yet and comment this.
- Assign a bug to every BotEventType not implemented yet and comment this.
- Add TaskState MISSING_INPUTS and SKIPPED_INTERNAL_FAILURE for further
  disambiguation of failure modes.
- Rename INTERNAL_FAILURE to RAN_INTERNAL_FAILURE to disambiguate with the 2 new
  related failures.
- Fix CASTree.digest back to an hex encoded string. It's 2x larger, but
  much simpler to work with.
- Rename BotInfo.raw to supplemental. The end goal is to have as much
  structured data as possible, but still allow customer-specified
  additional data. Renaming 'raw' to 'supplemental' makes this clearer.
- Create PhysicalEntity, which will enable the clear separation between
  host characteristics and device characteristics.
- Declare TaskProperties.containment, futureproofing the need for task process
  containment.

This change is a breaking change (proto message renumbering); TaskRequest is not
used anywhere yet.

Bug: 757931
Bug: 808836
Bug: 870723
Bug: 905087
Bug: 913978
Bug: 916553
Bug: 916556
Bug: 916557
Bug: 916559
Bug: 916560
Bug: 916562
Bug: 916570
Bug: 916578
Change-Id: Ic1a57d15d028802ad5cf8c6a2f13da15fac662c4
Reviewed-on: https://chromium-review.googlesource.com/c/1384425
Commit-Queue: Marc-Antoine Ruel <maruel@chromium.org>
Reviewed-by: Quinten Yearsley <qyearsley@chromium.org>

[modify] https://crrev.com/132c74e3ce003b29221bf3e5fc38af8c4f0bc049/appengine/swarming/handlers_prpc_test.py
[modify] https://crrev.com/132c74e3ce003b29221bf3e5fc38af8c4f0bc049/appengine/swarming/proto/api/plugin_prpc_pb2.py
[modify] https://crrev.com/132c74e3ce003b29221bf3e5fc38af8c4f0bc049/appengine/swarming/proto/api/swarming.proto
[modify] https://crrev.com/132c74e3ce003b29221bf3e5fc38af8c4f0bc049/appengine/swarming/proto/api/swarming_pb2.py
[modify] https://crrev.com/132c74e3ce003b29221bf3e5fc38af8c4f0bc049/appengine/swarming/proto/api/swarming_prpc_pb2.py
[modify] https://crrev.com/132c74e3ce003b29221bf3e5fc38af8c4f0bc049/appengine/swarming/server/bot_management.py
[modify] https://crrev.com/132c74e3ce003b29221bf3e5fc38af8c4f0bc049/appengine/swarming/server/bot_management_test.py
[modify] https://crrev.com/132c74e3ce003b29221bf3e5fc38af8c4f0bc049/appengine/swarming/server/task_request.py
[modify] https://crrev.com/132c74e3ce003b29221bf3e5fc38af8c4f0bc049/appengine/swarming/server/task_request_test.py
[modify] https://crrev.com/132c74e3ce003b29221bf3e5fc38af8c4f0bc049/appengine/swarming/server/task_result.py
[modify] https://crrev.com/132c74e3ce003b29221bf3e5fc38af8c4f0bc049/appengine/swarming/server/task_result_test.py
►

Sign in to add a comment