process: avoid redundant effort to reap orphan processes #3743
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
ipetkov
added
A-tokio
|
How does this PR affect #3521? |
Darksonn
reviewed
May 6, 2021
|
The mutex in |
* Child processes are akin to global variables within the parent process. Reaping orphaned child processes can be done by any thread (so long as someone does it), but repeated efforts from multiple threads to reap the same process do not add any additional benefit * Previously, each Runtime would register its own SIGCHLD listener and check it for updates during every tick. If a signal was received, then it will attempt to reap the orphan queue. * Signals are also "global resources" meaning that when one signal comes in, all listeners within that process are woken up. This means that if an application has multiple runtimes, each of them will attempt to drain the entire orphan queue any time a SIGCHLD is received (the first one to try will reap whichever orphans have exited, and the other will perform redundant effort). * Now we register a single SIGCHLD listener for the entire global queue itself, which allows us to "dedup" signal notifications even if multiple runtimes tick at the same time and attempt to drive the orphan queue (one will grab a lock and win the race, and everyone else will move on). * Moreover, the single, global SIGCHLD listener will be registered lazily (instead of eagerly on the runtime initialization). This means that applications which do not use child processes will not "pollute" the signal handler space, even if the feature is enabled at compile time
Co-authored-by: Alice Ryhl <alice@ryhl.io>
ipetkov
force-pushed
the
ivan/process-lazy-orphan
branch
from
4859b2f
to
2c5c624
Compare
|
Did that force push have any changes, or was it just a merge of master? (Please try to avoid force pushes — I can't tell what has changed and usually need to review the entire thing again) |
|
@Darksonn sorry I rebased on master because there was an unrelated failure (BSD test failed on some warning for an unused return value from sleep) |
|
No worries, although it's usually enough to merge in master. |
Darksonn
approved these changes
May 14, 2021
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
process. Reaping orphaned child processes can be done by any thread
(so long as someone does it), but repeated efforts from multiple
threads to reap the same process do not add any additional benefit
check it for updates during every tick. If a signal was received, then
it will attempt to reap the orphan queue.
in, all listeners within that process are woken up. This means that if
an application has multiple runtimes, each of them will attempt to
drain the entire orphan queue any time a SIGCHLD is received (the
first one to try will reap whichever orphans have exited, and the
other will perform redundant effort).
itself, which allows us to "dedup" signal notifications even if
multiple runtimes tick at the same time and attempt to drive the
orphan queue (one will grab a lock and win the race, and everyone else
will move on).
lazily (instead of eagerly on the runtime initialization). This means
that applications which do not use child processes will not "pollute"
the signal handler space, even if the feature is enabled at compile
time
Refs #3520