[FLINK-12064] [core, State Backends] RocksDBKeyedStateBackend snapshot uses incorrect key serializer if reconfigure happens during restore

[carp84]

What is the purpose of the change

This PR fixes the issue of RocksDBKeyedStateBackend snapshot using incorrect key serializer if reconfigured during restore in RocksDBKeyedStateBackendBuilder, which is a regression caused by FLINK-10043. We also reenforce the UT cases to cover the issue scenario.

Brief change log

Remove keySerializer from AbstractKeyedStateBackendBuilder and use keySerializerProvider.currentSchemaSerializer() instead in RocksDBKeyedStateBackendBuilder. For the test case part, perform an additional snapshot after state migration.

Verifying this change

This change added tests and can be verified as follows:

• Reenforced StateBackendMigrationTestBase, adding an additional snapshot after state migration to verify the snapshot part.

Does this pull request potentially affect one of the following parts:

• Dependencies (does it add or upgrade a dependency): (no)
• The public API, i.e., is any changed class annotated with @Public(Evolving): (no)
• The serializers: (yes)
  • This change assures to use correct serializer in snapshot
• The runtime per-record code paths (performance sensitive): (no)
• Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Yarn/Mesos, ZooKeeper: (yes)
  • This change prevents possible corruption of snapshot data
• The S3 file system connector: (no)

Documentation

• Does this pull request introduce a new feature? (no)
• If yes, how is the feature documented? (not applicable)

[flinkbot]

Thanks a lot for your contribution to the Apache Flink project. I'm the @flinkbot. I help the community
to review your pull request. We will use this comment to track the progress of the review.

Review Progress

• ✅ 1. The [description] looks good.
  • Approved by @tzulitai [PMC]
• ✅ 2. There is [consensus] that the contribution should go into to Flink.
  • Approved by @tzulitai [PMC]
• ❓ 3. Needs [attention] from.
• ✅ 4. The change fits into the overall [architecture].
  • Approved by @tzulitai [PMC]
• ✅ 5. Overall code [quality] is good.
  • Approved by @tzulitai [PMC]

Please see the Pull Request Review Guide for a full explanation of the review process.

Details

The Bot is tracking the review progress through labels. Labels are applied according to the order of the review items. For consensus, approval by a Flink committer of PMC member is required

Bot commands

The @flinkbot bot supports the following commands:

• @flinkbot approve description to approve one or more aspects (aspects: description, consensus, architecture and quality)
• @flinkbot approve all to approve all aspects
• @flinkbot approve-until architecture to approve everything until architecture
• @flinkbot attention @username1 [@username2 ..] to require somebody's attention
• @flinkbot disapprove architecture to remove an approval you gave earlier

[tzulitai]

Just to clarify:
the problem is that the state backend uses the correct key serializer for state access, but the wrong one is snapshotted in checkpoints, is that correct?

Otherwise, the title / description implies that it is using the wrong key serializer for runtime state access.

[tzulitai]

Thanks for the fix @carp84.

I've verified that the modified tests fail without the fixes to the key serializer access.
+1 to merge, will proceed to merge this to release-1.8 and master.

[tzulitai]

@flinkbot approve all

[carp84]

Thanks for the review! @tzulitai