【meta】【copyfilerange】chunk slices append should check slices exists like meta write operation

[penglaixy99]

What would you like to be added:

with copyfileranges operations, before append, chunk slices value should check whether the appending slice already exist in the chunk slice value.

Why is this needed:

we have met a problem, some applications (e.x vscode) use copyfilerange operations, application then operater into some exception state, then send the same copyfilerange commands for a long time. Then the dst inodes one chunk slice increased to more than 8MB, kv value 8MB! It's so big and tikv server IO utils increased to 100%. All the meta service latency increase!

[jiefenghuang]

For writes passed in by the client, we do not perform slice deduplication during the write operation. This avoids unnecessary metadata reads. Instead, when a certain condition is met (for example, when the number of slices becomes too large), we trigger a compact operation after reads and writes.

The CopyFileRange currently does not trigger this check automatically — of course, we could consider adding it to all write-type operations (such as CopyFileRange, Truncate, etc.). But since this scenario is somewhat special, for now you can trigger the compaction manually using the compact tool.

[lidaohang]

1