Skip to content

Conversation

@tonyxuqqi
Copy link

  • use flushed last applied index in gc

Signed-off-by: qi.xu tonxuqi@outlook.com

What problem does this PR solve?

Problem Summary:
TiKV may fail to restart after crash because GC may delete data that has not been synced to disk when disable_kv_wal is true.

What is changed and how it works?

Proposal:

What's Changed:
When GC, try to read the truncated index from SST directly to figure out the last index.

Related changes

tonyxuqqi and others added 3 commits September 15, 2021 06:35
* use flushed last applied index in gc

Signed-off-by: qi.xu <tonxuqi@outlook.com>
Add a new config: server.max_snapshot_file_raw_size (size before compression) to limit the write amplification in big region scenario.
For scenarios when older version TiKv co-exists (like rolling upgrade), it will rely on feature gate and still use single snapshot file for backward compatibility.

Signed-off-by: tonyxuqqi <tonyxuqi@outlook.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant