WIP: Backup labels #9

wolcen · 2022-12-16T20:52:26Z

wolcen commented

2022-12-16 20:52:26 +00:00

Tested deploy & backup (have not yet apptempted restore).

wolcen added 5 commits 2022-12-16 20:52:28 +00:00

5ceaf103c2 Correct README re: configuring users

eb0bbb36c5 Correct location of public upload files

5d17dfb4f8 Remove unnecessary mode assignment for config.json

b656e44248 Make backup configuration functional

pre-hook doesn't get a shell context when run by abra, so use sh -c per 3wc

continuous-integration/drone/pr Build is failing Details

a82dc10957 Remove legacy backup configuration

wolcen added 1 commit 2022-12-16 22:02:21 +00:00

continuous-integration/drone/pr Build is failing Details

d1335cacc9 Clean up inside container after backup

decentral1se approved these changes 2022-12-17 08:14:04 +00:00

decentral1se left a comment

LGTM 👍

compose.yml Outdated

						
				@ -34,3 +34,3 @@

				      - internal

				    volumes:

				      - codimd_uploads:/home/hackmd/app/public/uploads

				      - codimd_uploads:/hedgedoc/public/uploads

decentral1se commented

2022-12-17 08:11:46 +00:00

Shall we do codimd_uploads -> hedgedoc_uploads also while we're here? Would be a breaking change on the recipe, people would need to migrate their volumes, but idk how many folks are using this right now? Up to you!

Shall we do `codimd_uploads` -> `hedgedoc_uploads` also while we're here? Would be a breaking change on the recipe, people would need to migrate their volumes, but idk how many folks are using this right now? Up to you!

wolcen commented

2022-12-17 21:19:42 +00:00

I'm not personally sure it's worth breaking people's configs, but there are bigger issues than that, really. I'm a bit concerned about the existing deployments upgrade path - are there possibly upgrade hooks for this (for example, would be great to do a one-time tar/gz of the /hedgedoc/public/uploads if it is in the container)? People's volumes will be pointing to the incorrect folder, backups would not have been working, and just restarting the instance would kill their uploaded files otherwise, if I understand correctly?

Backups were not previously working (for me, at least), but at least this was what the recipe had reported, so hopefully no one was really trusting it for long-term pads (which we do intend to have).

FWIW, all the env vars for hedgedoc use codi's old "CMD_" prefixes, but it would probably be nice to see hedgedoc_uploads in your volume list.

If you'd prefer, I can do them in a different merge/pr, but I'm changing a few more things - I've added a health check for Postgres, and another CMD_ setting that I'll be using - I'm a little unclear how the versioning works at present, but just a note that if it'll require another version, more settings are inbound, one of which is the session secret in order to maintain logins between restarts.

I'm not personally sure it's worth breaking people's configs, but there are bigger issues than that, really. I'm a bit concerned about the existing deployments upgrade path - are there possibly upgrade hooks for this (for example, would be great to do a one-time tar/gz of the /hedgedoc/public/uploads if it is in the container)? People's volumes will be pointing to the incorrect folder, backups would not have been working, and just restarting the instance would kill their uploaded files otherwise, if I understand correctly? Backups were not previously working (for me, at least), but at least this was what the recipe had reported, so hopefully no one was really trusting it for long-term pads (which we do intend to have). FWIW, all the env vars for hedgedoc use codi's old "CMD_" prefixes, but it would probably be nice to see hedgedoc_uploads in your volume list. If you'd prefer, I can do them in a different merge/pr, but I'm changing a few more things - I've added a health check for Postgres, and another CMD_ setting that I'll be using - I'm a little unclear how the versioning works at present, but just a note that if it'll require another version, more settings are inbound, one of which is the session secret in order to maintain logins between restarts.

decentral1se commented

2022-12-18 10:14:40 +00:00

Oh yeh, v reasonable, maybe bundling all your changes into a new major recipe release version is the way to go? The upgrade paths are sometimes a bit ad-hoc yeh, so we have this "release notes" feature 👉 https://docs.coopcloud.tech/maintainers/handbook/#how-do-i-write-version-release-notes 👈 You can write a short guide which will turn up in the shell when people run upgrade and realise they need to do some work before doing the upgrade. Anyway, up to you!

Oh yeh, v reasonable, maybe bundling all your changes into a new major recipe release version is the way to go? The upgrade paths are sometimes a bit ad-hoc yeh, so we have this "release notes" feature 👉 https://docs.coopcloud.tech/maintainers/handbook/#how-do-i-write-version-release-notes 👈 You can write a short guide which will turn up in the shell when people run `upgrade` and realise they need to do some work before doing the upgrade. Anyway, up to you!