Skip to content

[autoscaler] fix the autoscaling bug for continuously launching failed nodes#11714

Merged
ericl merged 68 commits intoray-project:masterfrom
AmeerHajAli:master
Oct 30, 2020
Merged

[autoscaler] fix the autoscaling bug for continuously launching failed nodes#11714
ericl merged 68 commits intoray-project:masterfrom
AmeerHajAli:master

Conversation

@AmeerHajAli
Copy link
Copy Markdown
Contributor

Previously we considered pending and running nodes based on node tags rather than connected nodes from load metrics. This might be problematic as we cannot rely on node tags. This PR defines running nodes as connected nodes from LoadMetrics and pending nodes as the remaining nodes (non_terminated_nodes-connected nodes + pending queue) Furthermore, this PR treats the nodes that failed to launch as pending, and hence the number of failed nodes that are launched will be bounded by the max allowed pending nodes (in the code the handles the max concurrency).

Unblocks #11615 (merge after merging this please).

Related issue number

Checks

  • I've run scripts/format.sh to lint the changes in this PR.
  • I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
  • Testing Strategy
    • Unit tests
    • Release tests
    • This PR is not tested :(

@ericl ericl added the tests-ok The tagger certifies test failures are unrelated and assumes personal liability. label Oct 30, 2020
@ericl ericl merged commit 7aade46 into ray-project:master Oct 30, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

tests-ok The tagger certifies test failures are unrelated and assumes personal liability.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants