🤔 thinking out loud… #ZFS have this self-healing feature to preserve the data… so Office documents, images, videos are safe; even those CSS and HTML files for the static web site.

[joel@gts.tumfatig.net]

thinking out loud… #ZFS have this self-healing feature to preserve the data… so Office documents, images, videos are safe; even those CSS and HTML files for the static web site. But when it comes to databases stuff - think PostgreSQL, MongoDB, Elasticsearch, VictoriaMetrics, etc - those must already have features to guarantee data don’t get corrupted. So isn’t ZFS too much there…

[paul@notnull.space]

@joel
Not sure if it's too much, it's an important link in the chain.
Let ZFS deal with the more hardwarey things (drive failure, bit rot) which it does best, and let the software handle it's data as it likes.
To put it another way let ZFS ensure the data is in a good enough state so the database can operate it's protection mechanisms as and when it needs to.

Then you've also got ZFS mirroring, zraid, snapshots, clones, rollbacks...

[bob_zim@infosec.exchange]

@joel Databases in general trust the storage to be reliable. If it’s not, they can end up in very weird states. With a record size tuned to match the database’s record size, ZFS performs noticeably better than pre-CoW filesystems, so there’s very little reason not to use it.