indexer-alt: sequential pipeline #20053
Open
+505
−15
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
|
The latest updates on your projects. Learn more about Vercel for Git ↗︎
3 Skipped Deployments
|
8 tasks
8 tasks
## Description
Introduce a new kind of pipeline for indexing that needs commit data in
checkpoint order. This will be used for indexing data that would
previously have gone into `objects` or `objects_snapshot`, where rows
are modified in place, and so can't be committed out-of-order.
Sequential pipelines are split into two parts:
- A `processor` which is shared with the existing concurrent pipeline,
and is responsible for turning checkpoint data into values to be sent
to the database.
- A `committer` which is responsible for batching up prefixes of updates
and sending them to the DB when they are complete (no gaps between the
last write and what has been buffered).
- Although the committer must write out rows in order, it can buffer
the results of checkpoint processed out-of-order.
- It uses the ingestion service's regulator for back-pressure: The
ingestion service is only allowed to run ahead of all sequential
pipelines by its buffer size, which bounds the memory that each
pipeline must use to buffer pending writes.
- Sequential pipelines have different tuning parameters:
- `MIN_BATCH_ROWS`: The threshold for eagerly writing to the DB.
- `MAX_BATCH_CHECKPOINTS`: The maximum number of checkpoints that will
be batched together in a single transaction.
- They guarantee atomicity using DB transactions: All the writes for a
single checkpoint, and the corresponding watermark update are put into
the same DB transaction.
- They support simplifying/merging writes to the DB: If the same object
is modified multiple times across multiple checkpoints, only the
latest write will make it to the DB.
## Test plan
This change is primarily tested by the `sum_obj_types` pipeline
introduced in the next change.
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.
Description
Introduce a new kind of pipeline for indexing that needs commit data in checkpoint order. This will be used for indexing data that would previously have gone into
objectsorobjects_snapshot, where rows are modified in place, and so can't be committed out-of-order.Sequential pipelines are split into two parts:
processorwhich is shared with the existing concurrent pipeline, and is responsible for turning checkpoint data into values to be sent to the database.committerwhich is responsible for batching up prefixes of updates and sending them to the DB when they are complete (no gaps between the last write and what has been buffered).The key design constraints of the sequential pipeline are as follows:
MIN_BATCH_ROWS: The threshold for eagerly writing to the DB.MAX_BATCH_CHECKPOINTS: The maximum number of checkpoints that will be batched together in a single transaction.Test plan
This change is primarily tested by the
sum_obj_typespipeline introduced in the next change.Stack
Release notes
Check each box that your changes affect. If none of the boxes relate to your changes, release notes aren't required.
For each box you select, include information after the relevant heading that describes the impact of your changes that a user might notice and any actions they must take to implement updates.