simplify supervision propagation; unhandled events always cause failure #773
+392
−96
Conversation
This file contains hidden or 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
meta-cla
bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D79702385 |
Summary: Adds Proc::instance() which returns an actor instance and its corresponding handler. This allows the user to create a regular actor without any message handlers. The returned `Instance` provides all the normal capabilities, including sending and receiving messages, being able to spawn and manage child actors, etc. This is the foundation for a kind of "script mode" actor. Differential Revision: D79685752
mariusae
added a commit
to mariusae/monarch-1
that referenced
this pull request
Aug 6, 2025
…re (meta-pytorch#773) Summary: Currently (local) supervision can propagate events without also killing an intermediate actor. This is 1) wrong; and 2) complicated. Instead, we treat an unhandled supervision event as an actor failure, and then reduce the propagation paths to one: that of an actor failing. In order to retain accurate attribution, we add a "caused_by" field to the actor supervision events. Differential Revision: D79702385
mariusae
force-pushed
the
export-D79702385
branch
from
1a9c216 to
836dce5
Compare
|
This pull request was exported from Phabricator. Differential Revision: D79702385 |
mariusae
force-pushed
the
export-D79702385
branch
from
836dce5 to
e3bc0c7
Compare
mariusae
added a commit
to mariusae/monarch-1
that referenced
this pull request
Aug 6, 2025
…re (meta-pytorch#773) Summary: Currently (local) supervision can propagate events without also killing an intermediate actor. This is 1) wrong; and 2) complicated. Instead, we treat an unhandled supervision event as an actor failure, and then reduce the propagation paths to one: that of an actor failing. In order to retain accurate attribution, we add a "caused_by" field to the actor supervision events. Reviewed By: shayne-fletcher Differential Revision: D79702385
mariusae
added a commit
to mariusae/monarch-1
that referenced
this pull request
Aug 6, 2025
…re (meta-pytorch#773) Summary: Currently (local) supervision can propagate events without also killing an intermediate actor. This is 1) wrong; and 2) complicated. Instead, we treat an unhandled supervision event as an actor failure, and then reduce the propagation paths to one: that of an actor failing. In order to retain accurate attribution, we add a "caused_by" field to the actor supervision events. Reviewed By: shayne-fletcher Differential Revision: D79702385
mariusae
force-pushed
the
export-D79702385
branch
from
e3bc0c7 to
b514443
Compare
|
This pull request was exported from Phabricator. Differential Revision: D79702385 |
mariusae
added a commit
to mariusae/monarch-1
that referenced
this pull request
Aug 6, 2025
…re (meta-pytorch#773) Summary: Pull Request resolved: meta-pytorch#773 Currently (local) supervision can propagate events without also killing an intermediate actor. This is 1) wrong; and 2) complicated. Instead, we treat an unhandled supervision event as an actor failure, and then reduce the propagation paths to one: that of an actor failing. In order to retain accurate attribution, we add a "caused_by" field to the actor supervision events. Reviewed By: shayne-fletcher Differential Revision: D79702385
mariusae
force-pushed
the
export-D79702385
branch
from
b514443 to
23b866c
Compare
…re (meta-pytorch#773) Summary: Pull Request resolved: meta-pytorch#773 Currently (local) supervision can propagate events without also killing an intermediate actor. This is 1) wrong; and 2) complicated. Instead, we treat an unhandled supervision event as an actor failure, and then reduce the propagation paths to one: that of an actor failing. In order to retain accurate attribution, we add a "caused_by" field to the actor supervision events. Reviewed By: shayne-fletcher Differential Revision: D79702385
|
This pull request was exported from Phabricator. Differential Revision: D79702385 |
mariusae
force-pushed
the
export-D79702385
branch
from
23b866c to
7e7f89a
Compare
dulinriley
pushed a commit
to dulinriley/monarch
that referenced
this pull request
Aug 6, 2025
Summary: Pull Request resolved: meta-pytorch#644 1.10.1 (March 5th, 2025) Fixed Fix memory leak when using to_vec with Bytes::from_owner (meta-pytorch#773) 1.10.0 (February 3rd, 2025) Added Add feature to support platforms without atomic CAS (meta-pytorch#467) try_get_* methods for Buf trait (meta-pytorch#753) Implement Buf::chunks_vectored for Take (meta-pytorch#617) Implement Buf::chunks_vectored for VecDeque<u8> (meta-pytorch#708) Fixed Remove incorrect guarantee for chunks_vectored (meta-pytorch#754) Ensure that tests pass under panic=abort (meta-pytorch#749) Reviewed By: cjlongoria Differential Revision: D78948561 fbshipit-source-id: bb755e25088f9d77f0d5a9ae3027cadc51ea586a
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.
Summary:
Currently (local) supervision can propagate events without also killing an intermediate actor.
This is 1) wrong; and 2) complicated.
Instead, we treat an unhandled supervision event as an actor failure, and then reduce the propagation paths to one: that of an actor failing.
In order to retain accurate attribution, we add a "caused_by" field to the actor supervision events.
Differential Revision: D79702385