worker: decode partition records as v2 envelopes with v1 fallback by muhamadazmy · Pull Request #4696 · restatedev/restate

muhamadazmy · 2026-05-05T12:25:20Z

worker: decode partition records as v2 envelopes with v1 fallback

Summary

Switch the partition processor's read path to v2 envelopes. Records
are decoded as v2::Envelope<Raw> first; if the record on disk is
still flexbuffers-encoded v1, we downcast through the new
StorageDecodeError::TypedValueMismatch variant (which now carries
the original Arc) and convert via the v1→v2 compatibility shim.

The decode path no longer relies on Header::dest to decide whether
a record targets this partition. Instead, the processor stores a
KeyFilter derived from its partition key range and checks each
record's Keys via MatchKeyQuery. Deduplication still runs, but
now reads Dedup off the v2 header (converted to the existing
DedupInformation for the dedup table — TODO left to consume Dedup
directly).

apply_record is rewritten around RecordKind: AnnounceLeader and
UpdatePartitionDurability are handled inline; everything else is
forwarded to the state machine as a typed v2 envelope.

Tests under state_machine/ and pp-bench are updated to construct
records via records::<Kind>::new_test(...) instead of the v1
Command::<Variant>(...) enum.

Writers still emit v1 envelopes; only reads go through v2.

Stack created with Sapling. Best reviewed with ReviewStack.

github-actions · 2026-05-06T17:06:01Z

Test Results

8 files ±0 8 suites ±0 4m 36s ⏱️ -27s
50 tests ±0 50 ✅ ±0 0 💤 ±0 0 ❌ ±0
218 runs ±0 218 ✅ ±0 0 💤 ±0 0 ❌ ±0

Results for commit 4bdc261. ± Comparison against base commit d71638a.

♻️ This comment has been updated with latest results.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c83fbaf695

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-05-07T13:53:07Z

+    /// partition key range
+    #[bilrost(tag(4))]
+    pub partition_key_range: Keys,


Preserve decoding of old durability records

When a new processor replays or follows a WAL segment containing Command::UpdatePartitionDurability written by a pre-change node, the flexbuffers/serde V1 payload will not contain this newly added partition_key_range field. Because the field is non-optional and has no #[serde(default)], v1::Envelope::decode fails before the V1-to-V2 fallback can convert the record, so rolling upgrades or replay of untrimmed old durability records can stop the partition processor. Please add a backward-compatible default/compat path for this field.

Useful? React with 👍 / 👎.

AhmedSoliman · 2026-05-08T16:36:00Z

    UnsupportedCodecKind(StorageCodecKind),
+    #[error("Type mismatch. Original value in PolyBytes::Typed({}) does not match requested type", _0.type_name())]
+    #[debug("TypedValueMismatch({})", _0.type_name())]
+    TypedValueMismatch(Arc<dyn StorageEncode>),


Not a huge fan of passing the entire value into the error. Is this really necessary? I suppose one can formulate the error message and return it in a ReString instead?

It's not only for the name. this is actually used to fallback to v1 decoding in some cases.

The problem here is because of record cache, we might try to decode v1 envelope as v2 where the PolyBytes are typed. To make this work, I instead pass the inner value to be handled here https://github.com/restatedev/restate/pull/4696/changes#diff-09e9d092973d97a3155dc5ca9f928dee76e6b85bbea659659c3083da645f7d09R458

AhmedSoliman · 2026-05-11T09:18:56Z

-            keys: payload.record_keys(),
-            payload,
+    fn from(command: C) -> PartialEnvelope {
+        // let payload = payload.into();


AhmedSoliman · 2026-05-11T09:23:09Z

-            let record_lsn = record.lsn;
-            let envelope = Arc::unwrap_or_clone(record.envelope);
+        // todo(azmy): use dedup() directly without first converting to DedupInformation
+        let dedup_information: Option<DedupInformation> = record.envelope.dedup().clone().into();


I'd love if we can address this one pretty soon

AhmedSoliman

Thank you for working on this 🚢

## Summary Switch the partition processor's read path to v2 envelopes. Records are decoded as `v2::Envelope<Raw>` first; if the record on disk is still flexbuffers-encoded v1, we downcast through the new `StorageDecodeError::TypedValueMismatch` variant (which now carries the original Arc) and convert via the v1→v2 compatibility shim. The decode path no longer relies on `Header::dest` to decide whether a record targets this partition. Instead, the processor stores a `KeyFilter` derived from its partition key range and checks each record's `Keys` via `MatchKeyQuery`. Deduplication still runs, but now reads `Dedup` off the v2 header (converted to the existing `DedupInformation` for the dedup table — TODO left to consume Dedup directly). `apply_record` is rewritten around `RecordKind`: AnnounceLeader and UpdatePartitionDurability are handled inline; everything else is forwarded to the state machine as a typed v2 envelope. Tests under state_machine/ and pp-bench are updated to construct records via `records::<Kind>::new_test(...)` instead of the v1 `Command::<Variant>(...)` enum. Writers still emit v1 envelopes; only reads go through v2.

This was referenced May 5, 2026

wal-protocol: introduce v2 Envelope with bilrost encoding and lazy payloads #4695

Merged

wal-protocol: extract v1 Envelope into a dedicated module #4694

Merged

muhamadazmy changed the title ~~Read and use Envelope v2~~ [wip] read and use envelope v2 May 6, 2026

muhamadazmy force-pushed the pr4696 branch 2 times, most recently from 983f7e3 to 4a99ecf Compare May 6, 2026 16:25

muhamadazmy force-pushed the pr4696 branch 2 times, most recently from 5c3811d to e49a760 Compare May 7, 2026 08:41

muhamadazmy changed the title ~~[wip] read and use envelope v2~~ worker: decode partition records as v2 envelopes with v1 fallback May 7, 2026

muhamadazmy force-pushed the pr4696 branch 4 times, most recently from 9bac5c5 to c83fbaf Compare May 7, 2026 13:24

muhamadazmy mentioned this pull request May 7, 2026

worker: write v2 envelopes with typed commands #4700

Draft

muhamadazmy marked this pull request as ready for review May 7, 2026 13:48

muhamadazmy requested a review from AhmedSoliman May 7, 2026 13:48

chatgpt-codex-connector Bot reviewed May 7, 2026

View reviewed changes

muhamadazmy force-pushed the pr4696 branch 13 times, most recently from b3e0c8a to c528d5a Compare May 8, 2026 11:24

muhamadazmy force-pushed the pr4696 branch from c528d5a to d106008 Compare May 8, 2026 16:34

AhmedSoliman reviewed May 8, 2026

View reviewed changes

muhamadazmy force-pushed the pr4696 branch 2 times, most recently from f781f06 to 3b4acac Compare May 10, 2026 16:14

muhamadazmy mentioned this pull request May 10, 2026

wal-protocol: rename Record→Command and unify payload types #4710

Merged

muhamadazmy force-pushed the pr4696 branch 6 times, most recently from d882faa to c1d0d27 Compare May 11, 2026 08:40

AhmedSoliman reviewed May 11, 2026

View reviewed changes

muhamadazmy force-pushed the pr4696 branch from c1d0d27 to 9783a3e Compare May 11, 2026 09:43

AhmedSoliman approved these changes May 11, 2026

View reviewed changes

Comment thread crates/platform/src/storage.rs Outdated

Comment thread crates/worker/src/partition/state_machine/mod.rs

Comment thread crates/worker/src/partition/state_machine/mod.rs

Comment thread tools/pp-bench/src/extract.rs Outdated

muhamadazmy force-pushed the pr4696 branch from 9783a3e to f3d4c5e Compare May 11, 2026 10:11

muhamadazmy force-pushed the pr4696 branch from f3d4c5e to 4bdc261 Compare May 11, 2026 10:29

muhamadazmy merged commit c53757e into main May 11, 2026
73 of 74 checks passed

muhamadazmy deleted the pr4696 branch May 11, 2026 11:03

github-actions Bot locked and limited conversation to collaborators May 11, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

worker: decode partition records as v2 envelopes with v1 fallback#4696

worker: decode partition records as v2 envelopes with v1 fallback#4696
muhamadazmy merged 1 commit into
mainfrom
pr4696

muhamadazmy commented May 5, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented May 6, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot May 7, 2026

Uh oh!

AhmedSoliman May 8, 2026

Uh oh!

muhamadazmy May 8, 2026

Uh oh!

AhmedSoliman May 11, 2026

Uh oh!

AhmedSoliman May 11, 2026

Uh oh!

AhmedSoliman left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

muhamadazmy commented May 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Uh oh!

github-actions Bot commented May 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Test Results

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot May 7, 2026

Choose a reason for hiding this comment

Uh oh!

AhmedSoliman May 8, 2026

Choose a reason for hiding this comment

Uh oh!

muhamadazmy May 8, 2026

Choose a reason for hiding this comment

Uh oh!

AhmedSoliman May 11, 2026

Choose a reason for hiding this comment

Uh oh!

AhmedSoliman May 11, 2026

Choose a reason for hiding this comment

Uh oh!

AhmedSoliman left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

muhamadazmy commented May 5, 2026 •

edited

Loading

github-actions Bot commented May 6, 2026 •

edited

Loading