Howdy;
My app’s volumes list is littered with deleted and migrated volume states:
❯ fly volumes list
ID STATE NAME SIZE REGION ZONE ATTACHED VM CREATED AT
vol_1g67340gj0prydxw created ephemera_cache 25GB ewr 262d b038c6a8 2 months ago
vol_w1q85vgnplp4zdxe created ephemera_cache 20GB ord 4f52 c55be546 9 hours ago
vol_0nylzre1kno4qmkp created ephemera_cache 20GB ord d6bf 4968e5d5 9 hours ago
vol_52en7r1lq994k6yx created ephemera_cache 25GB sea 4df6 33bc3775 2 months ago
vol_70zy6r7qmj9vdjng created ephemera_cache 25GB sea bec5 cc0b41f5 2 months ago
vol_ke628r630onrwmnp created ephemera_cache 25GB dfw e163 4b2955d7 2 months ago
vol_0nylzre83n1rqmkp deleted ephemera_cache 25GB ewr 429d 2 months ago
vol_jlgz1vp7lkv78m30 created ephemera_cache 25GB yyz c150 23c7f693 10 months ago
vol_2n0l9vl6pjgv635d created ephemera_cache 25GB dfw ccb3 deab349c 2 months ago
vol_3xme149mgzrowpln migrated ephemera_cache 25GB ord 3c24 10 months ago
vol_ke628r6lzqrwmnpy migrated ephemera_cache 25GB ord 3c24 10 months ago
vol_mjn924ogpqr03lq7 migrated ephemera_cache 25GB dfw b7f2 10 months ago
I am unable to affect any of the volumes in migrated or deleted states without getting errors, so they hang around like bad pennies when managing volumes.
@kurt has said previously that migrated volumes are not being charged, which is great, but now I’m running into this when trying to create new volumes in some regions:
❯ fly volumes create ephemera_cache --region ewr --size 20
Error failed creating volume: Can't create volume, application is already using 9 of 9 zones in ewr
Related to the volume shenanigans – I was using volume anchor scaling, or I was but I can’t find that in the docs anymore. Is that still the blessed path for picking your regions if you have a volume-backed app? I switched to auto-scaling with a equal min-max to control it at the moment.
Honestly if I could just declare the regions, counts and volume size/name all in my fly.toml it would move from magical to utility, in the best way possible. I realize there would still be decisions or defaults around how permanent a backing volume is during a re-release (e.g. re-use, or drop the old and re-create, etc), but that feels like a hill you’ve already climbed with existing volume behaviors – re-use when it’s there with the same size, drop & create when the size changes, create when there’s no matching volume. Heck, put confirmations in the CLI if you’re feeling saucy.
Wrapping up:
- Zombie volumes – what do?
- Anchor scaling – old & busted?