This repository was archived by the owner on Apr 7, 2025. It is now read-only.
[BEAM-8549] Do not use keyed operator state for checkpoint buffering #29
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
The current buffer logic for items emitted during checkpointing is faulty in the sense that the buffer is partitioned on the output keys of the operator. The key may be changed or even be dropped. Thus, the original key partitioning will not be maintained which will cause checkpointing to fail. An alternative solution would be BEAM-6733 / apache#9652, but this change keeps the current buffering logic in place. The output buffer may now always be redistributed round-robin upon restoring from a checkpoint. Note that this is fine because no assumption can be made about the distribution of output elements of a DoFn operation.
…ted during checkpointing As part of a checkpoint, the current bundle is finalized. When the bundle is finalized, the watermark, currently held back, may also be progressed which can cause the start of another bundle. When a new bundle is started, any to-be-buffered items from the previous bundle for the pending checkpoint may be emitted. This should not happen. This only effects portable pipelines where we have to hold back the watermark due to the asynchronous processing of elements.
If a bundle fails to finalize before creating a checkpoint, it may be swallowed and just considered a checkpointing error. This breaks the execution flow and exactly-once guarantees.
mxm
force-pushed
the
release-2.16.0-lyft-checkpointing
branch
from
573db12 to
9ba8b29
Compare
The default passed from the Kafka consumer wrapper for start_from_timestamp_millis is "None".
tweise
approved these changes
Nov 8, 2019
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
The current buffer logic for items emitted during checkpointing is faulty in the
sense that the buffer is partitioned on the output keys of the operator. The key
may be changed or even be dropped. Thus, the original key partitioning will not
be maintained which will cause checkpointing to fail.
An alternative solution would be BEAM-6733 / apache#9652, but this change keeps the
current buffering logic in place. The output buffer may now always be
redistributed round-robin upon restoring from a checkpoint. Note that this is
fine because no assumption can be made about the distribution of output elements
of a DoFn operation.
Post-Commit Tests Status (on master branch)