Embed Notice
HTML Code
Corresponding Notice
- Embed this notice@Jain @ivesen The seaweedfs fsck isn't too hard to operate but we're not talking about a reliable thing like xfs. You probably can't use seaweedfs for entirely critical data and you *likely* want to replicate on a filesystem/block layer that has snapshots to have "backups" of volumes just in case, even if the volumes are replicated / and or erasure coded.
It should write pretty evenly across volume servers *over time* as they fill up but it is not doing it for all small bulks of files. Maybe it just writes it into *one* of your 25gb or w/e volumes and then replicates it to another drive as you asked. Amuch simple(r) scheme than having the rule engines+databases ceph runs but it makes actual hw usage overall a lot more modest and predictable.