bitflicker64 opened a new pull request, #3383:
URL: https://github.com/apache/hugegraph/pull/3383
## Purpose of the PR
- part of #3235; no issue of its own
`KvServiceGrpcImpl.get` and `scanPrefix` answer from the leader's local
store. A leader elected a
moment ago may not have applied every committed write yet, so a client can
read a value older than
one it already saw committed: a Server that caches schema from a KV read
would keep a stale copy
past a revision it acknowledged. #3368 put a ReadIndex barrier before PD
reads an id counter; this
PR puts the same barrier, with the same deadline, before the two KV reads.
It was part of #3376 and was reviewed there; it does not depend on that PR
and stands on its own.
Step 2 of the schema cache sync (the record watch with ACK and retry) needs
both merged.
This PR is limited to the ReadIndex barrier before the two KV reads; a bug
that already exists on
master in code this PR did not write is tracked as an issue, not fixed here.
## Main Changes
One commit, three files, 8 production, 205 test and 1 doc lines.
- `KvServiceGrpcImpl.get` and `scanPrefix` call `RaftEngine.waitReadIndex()`
before they read the
store. When the leader cannot confirm and apply a ReadIndex within
`raft.rpc-timeout`, the read
fails with the error in the response header and no value, instead of a
stored value that may be
stale. A follower redirects to the leader as before.
- `hugegraph-pd/docs/configuration.md`: the `raft.rpc-timeout` row says the
deadline also bounds
the KV `get` and `scanPrefix`.
## Verifying these changes
- [ ] Trivial rework / code cleanup without any test coverage. (No Need)
- [ ] Already covered by existing tests, such as *(please modify tests
here)*.
- [x] Need tests and can be verified as follows:
- `KvServiceGrpcImplTest` (new, `hg-pd-service`; it runs with every PD
suite, since `-am`
builds the module): `testGetAndScanReadAfterTheReadIndex` (the
ReadIndex runs before each
read, which then returns the stored value) and
`testFailedReadIndexReturnsNoStoredValue` (a
ReadIndex that times out, as on a leader without a quorum, is answered
with an error and no
value, and the store is not read at all). The raft node is a proxy
that answers each
ReadIndex as jraft does, by setting the result and running the closure.
- Suites on Pune101 (JDK 17, 2026-10-20, under the shared lock) at
9cb4e5582: package and
`editorconfig:check` clean, pd-common 83/0/0/0, pd-core 190/0/0/2,
pd-client 83/0/0/0
against a packaged PD (tests, failures, errors, skipped); the
`hg-pd-service` module ran
29 tests in each suite, master's 27 plus these two.
## Does this PR potentially affect the following parts?
- [ ] Dependencies ([add/update
license](https://hugegraph.apache.org/docs/contribution-guidelines/contribute/#321-check-licenses)
info &
[regenerate_known_dependencies.sh](../install-dist/scripts/dependency/regenerate_known_dependencies.sh))
- [ ] Modify configurations
- [ ] The public API
- [x] Other affects (typed here): a KV `get` or `scanPrefix` on a leader
that cannot confirm a
ReadIndex within `raft.rpc-timeout` (10 s by default) now fails instead of
answering from its
local store; a client retries as it does for any PD error. Each read waits
one ReadIndex round
trip on a healthy leader.
- [ ] Nope
## Documentation Status
Select one option and provide the documentation location when applicable.
- [ ] `Doc - TODO`: required documentation is pending; complete it before
merging.
- [x] `Doc - Done`: documentation is included here or linked below.
- [ ] `Doc - No Need`: no user-visible documentation is affected.
Documentation files in this PR or paired hugegraph-doc PR:
`hugegraph-pd/docs/configuration.md`, the `raft.rpc-timeout` row.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
