platoneko opened a new pull request, #35473:
URL: https://github.com/apache/doris/pull/35473
## Proposed changes
The current storage path format for remote segment files in Doris is:
`${prefix}/data/${tablet_id}/${rowset_id}_${seg_id}.dat`
There are the following issues with this storage path on HDFS:
- HDFS has a limitation on the number of directory entries in a directory,
typically not exceeding 1 million (dfs.namenode.fs-limits.max-directory-items).
- The HDFS list interface does not have pagination functionality, resulting
in high overhead when listing a large number of files.
- The Recycler needs to support recycling rowsets with an unknown number of
segments, but efficiently deleting all files with the `rowset_id` prefix is
challenging in HDFS.
A more HDFS-friendly path format would be:
`${prefix}/data/${shard_id}/${tablet_id}/${rowset_id}/${seg_id}.dat`
To support the aforementioned path format while maintaining compatibility
with existing data, all existing `StorageResource` / `StorageVault`s are
assumed to use path v0 for reading and writing segment files. When reading or
writing remote segment files, the corresponding file paths need to be obtained
from the `StorageResource`.
## Further comments
If this is a relatively large or complex change, kick off the discussion at
[[email protected]](mailto:[email protected]) by explaining why you
chose the solution you did and what alternatives you considered, etc...
--
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]
