Fix early quiescence/termination bug #4550
Conversation
Prior to this commit, the logic to detect quiescence had a race condition in relation to dynamic scheduler scaling and it was possible for the runtime to incorrectly detect quiescence and termninate early if a scheduler thread suspended at just the right time. This commit changes the quiescence logic to keep an accurate track of exactly how many scheduler threads are active at the time the quiescence detection protocol begins so it can ensure that any scheduler threads suspending or unsuspending can no longer cause an incorrect determination that might lead to early termination of the runtime.
ponylang-main
added
the
discuss during sync
|
not included as part of this PR, but stopping the ASIO thread should likely be moved to happen after all the scheduler threads have stopped... maybe worth doing as a follow on thing.. |
SeanTAllen
added
the
changelog - fixed
|
Hi @dipinhora, The changelog - fixed label was added to this pull request; all PRs with a changelog label need to have release notes included as part of the PR. If you haven't added release notes already, please do. Release notes are added by creating a uniquely named file in the The basic format of the release notes (using markdown) should be: Thanks. |
That sounds like a good idea |
| send_msg_all_active(sched->index, SCHED_CNF, sched->ack_token); | ||
| // save the # of active schedulers to expect ACK's from | ||
| sched->ack_count = send_msg_all_active(sched->index, SCHED_CNF, sched->ack_token); | ||
| pony_assert(sched->ack_count > 0); |
There was a problem hiding this comment.
this assertion failed during a CI run.
There was a problem hiding this comment.
yep.. i missed an edge case.. will have a fix soonish...
ponylang-main
removed
the
discuss during sync
Prior to this commit, the logic to detect quiescence had a race condition in relation to dynamic scheduler scaling and it was possible for the runtime to incorrectly detect quiescence and termninate early if a scheduler thread suspended at just the right time.
This commit changes the quiescence logic to keep an accurate track of exactly how many scheduler threads are active at the time the quiescence detection protocol begins so it can ensure that any scheduler threads suspending or unsuspending can no longer cause an incorrect determination that might lead to early termination of the runtime.