Changxin Miao

No need to fire useless delete requests when slice is still in stage dir

1

Avoid slow and expensive object deletion if the block is still in staging

Write should fail fast if file is already in error state, avoiding resource consumption caused by infinite retries from bad clients. If this change poses a risk, we may consider...

I didn't find anywhere that deletes `uuid` from `trackerUUIDs`, which may causes memory leak for a long-running pinger. And it may have a performance issue when searching for `uuid` in...

When stage disk hangs for a long time, user will get an EIO finally. We can avoid this fatal error by applying `PutTimeout` to stage writings too. https://github.com/juicedata/juicefs/blob/fd191eadc403a30441a3aee291a8c8622b2aa32a/pkg/vfs/writer.go#L360-L362

[Feature] - Cache backend files/blobs/objects

4

**Describe the solution you'd like** Would you consider caching the backend objects in the gateway? This will benefit repeated read scenarios a lot. Another use case is to warm-up the...

In a mixed read-write workload, write requests are easily affected by read requests, even if the backend object storage does not have such issue. This is because when the buffer...

**What would you like to be added**: Trash files are flattened into hourly buckets, and `CleanupTrashBefore` deletes those entries one at a time sequentially. It's better to have it use...