When calling getUpdatesSince get IO error: while stat a file for size

[pipamc]

Note: Please use Issues only for bug reports. For questions, discussions, feature requests, etc. post to dev group: https://www.facebook.com/groups/rocksdb.dev

Now I use rocksdb with v5.18.3 and use getUpdatesSince API to sync data to downstream. But I got error
IO error: while stat a file for size: /data2/data/k/k_instance_1166609718592933888/data/10.3.63.9:10090/archive/183996.log: No such file or directory
and I check the directory, the file does not exist any more. After I read the code, I found there is one bad case when getUpdatesSince get all the files but WAL rotated occurs.

Expected behavior

No such error.

Actual behavior

When calling getUpdatesSince, I got this error.

Steps to reproduce the behavior

upstream write a lot of data, and calling getUpdatesSince will get this error.

[riversand963]

Just to confirm, you are using recycle_log_file_num = 1? Seems the race condition you described is not handled in wal manager code.
Would you be willing to contribute a fix together with a unit test?

[pipamc]

Just to confirm, you are using recycle_log_file_num = 1? Seems the race condition you described is not handled in wal manager code.
Would you be willing to contribute a fix together with a unit test?

no, I just use default settings, recycle_log_file_num = 0