David Steele

``` sftp > cd > ls ```

> Could it be a pgbackrest bug? Sure, that's possible, but if we can't reproduce it we don't have any chance of fixing it. How big is the database that...

> With my previous observations, so far, I wonder whether this might be related to the number of files in the backup? Interesting idea -- I just did a test...

> Delete Y Data Directory (PGDATA) and all WAL records on VMy. It would be more efficient to add `--delta` to the restore: https://pgbackrest.org/user-guide.html#restore/option-delta > Recopy Y PostgreSQL Configuration Files...

This is likely happening because retention is never being met. We hold old archive until retention is met to make sure replicas don't get starved for data after the first...

> set the full retention to what you what to keep. This is just temporary if you want to remove the old WAL. Otherwise the old WAL will go away...

We would be open to an fcntl() implementation. The goal is to be Posix-compliant but this must have slipped through since it is so widely supported.

> Would you please review these lock.c and lock.h modified files: We will review your changes and adjust tests if you submit a pull request -- see https://github.com/pgbackrest/pgbackrest/pulls. Thanks!

Can you submit a PR with a proposed fix for this?

Query ID might be interesting but I don't think exec time is appropriate for auditing. But even for query ID I'd want to get more feedback from people who would...