@deepjoy@hachyderm.io

· 6mo ago

Shoebox is live. S3-compatible API for your local filesystem.

docker run -it -v ~/Photos:/data -p 9000:9000 ghcr.io/deepjoy/shoebox /data

Files stay where they are. Each directory becomes an S3 bucket. Works with rclone, AWS CLI, any S3 SDK. Credentials auto-generated and printed on startup.

SQLite metadata. MIT licensed. Or `cargo install shoebox`.

github.com/deepjoy/shoebox

#selfhosted #selfhosting #homelab #rustlang #opensource #S3 #NAS

hachyderm.io

Hachyderm.io

11

0

1

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

Pointed Shoebox at a directory: 106,715 files, 71.4 GB. Found 18,847 duplicate files across 581 folders. 8.2 GB reclaimable. Files untouched at their original paths.

Built this to find duplicate photos. That required an S3 server. Then a terminal couldn't show me which photos matched, so I built a webapp too.

One curl command enables CORS and your browser connects to your running instance. MIT licensed, ~10MB binary.

Looking for testers running NAS/homelab setups (btrfs, ext4, ZFS, Docker volumes). Reply or DM your setup.

github.com/deepjoy/shoebox
deepjoy.github.io/shoebox-webapp

#selfhosted #selfhosting #homelab #rustlang #opensource

hachyderm.io

Hachyderm.io

3

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

Replying to

@cthulku Yeah that's totally normal. HDDs spin up for any read/write, and background tasks (indexing, health checks, container logging) keep them chattering even when you're not doing anything obvious. Biggest quality-of-life upgrade I ever did was gradually mixing SSDs into my array. Even replacing one or two drives made a noticeable difference in noise. SSD prices have dropped enough that it's actually a reasonable strategy now. You don't have to do it all at once, just swap one in next time a drive dies or you feel like spending $40. Rubber feet help too but they're treating the symptom, not the cause.

1

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

@andrew@mastodon.furrow.me I keep trying to move to Podman and it keeps sending me back to Docker. It's like a bad relationship where the other person is technically better for you but can't remember your birthday.

One day we'll all be rootless and free. Today is not that day.

1

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

@phoenix@chaos.social Love the framing of "get your hands dirty, not run the next Let's Encrypt." That's exactly the right approach. Understanding what's actually happening under the hood makes debugging so much less painful when certs inevitably break at 2am on a Saturday.

1

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

Replying to

@ernestdeleon I get the impulse but I went the other direction. Everything in my house is connected, it's just all local. Home Assistant, Zigbee devices, wired PoE cameras. Nothing phones home, nothing needs the internet to work. Modem dies? Everything still runs. I keep spares for anything that's a bottleneck, router, Zigbee hub, extra cameras. Swap and go. The problem isn't smart, it's cloud. Once you cut that out, connected stuff is actually great. Took me a few years of pain to get there though.

1

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

Replying to

@bplein@bvp.me Authentik. I run it behind Traefik on my homelab and it handles OIDC/SAML for everything - NextCloud, Forgejo, etc. The UI is polished enough that non-techies can handle login flows and password resets without you getting a support call. It also has a mailcow integration: https://integrations.goauthentik.io/chat-communication-collaboration/mailcow/ Fair warning: back up your secrets religiously. I once deleted my auth server secrets and got locked out of everything. That's the downside of centralizing auth - it becomes the one thing you absolutely cannot lose.

integrations.goauthentik.io

Integrate with mailcow | authentik

1

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

Replying to

@opennomad@mastodon.social This is great. I run a bunch of scheduled tasks on a bare-minimum Debian box and systemd timers have always felt heavier than they need to be compared to cron. A single-file wrapper with natural language schedules is a really nice touch.

1

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

Replying to

@davidbures That Hetzner timeout is almost certainly a port thing. Their storage boxes use port 23 for SFTP, not the standard 22. If your SSH config or Borg remote is pointing at 22 it'll just hang forever. That said if you're already shopping around, restic is great. Way less fiddly than Borg for remote targets, and no weird YAML. Just restic init and restic backup. Done. Although best form of backup is to have some, don't sleep on it.

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

Replying to

@thomasjsn @gohugoio @healthchecks_io I have @uptime_kuma and Dashy running on my server and I never look at either of them. The journal entries are what would actually make a status page useful. That DNS certificate issue is exactly the kind of thing I'd forget about and then waste hours rediscovering. Might need to steal this idea.

0

2

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

Replying to

@thomasjsn Yeah the "if one actually uses it" is the key part 😄 I'm great at setting up tools and terrible at actually sticking with them. jrnl looks nice though, low friction is exactly what it takes to make a habit stick.

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

@courtney@pdx.social 10 minutes and a cup of coffee is honestly the most effective VPS debugging strategy I've found. Just enough time for someone else's problem to stop being your problem.

mod_evasive is solid. If you're on nginx instead of Apache at some point, rate limiting is built in and pretty painless to configure. But for Apache that's the right tool.

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

Replying to

@autkin@fosstodon.org Been there. Running a home server setup means you always have almost the right cable somewhere in a drawer. Hope the UPS battery held long enough for the print job!

0

2

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

Replying to

@autkin@fosstodon.org yup, the sentiment resonates since I live in an area that has power outages from time to time. The last time I had a btrfs scrub running. The only way I could make sure that was not killed was by shutting down all my Poe devices that were also on the ups.

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

Replying to

@rune@social.intothecloud.net Nice! I've been running Home Assistant for a while but never thought to connect it through MCP like this. The inventory query is what gets me though. I have a whole rack of drives and random hardware I lose track of. HomeBox might be the move.

0

2

0

DJ Majumdar @deepjoy@hachyderm.io

· 7mo ago

Replying to

@rune@social.intothecloud.net Glue apps is exactly the right word for it. The accuracy problem is real. I know if I set something like that up for my hardware I'd fall behind on updating it within a week. HA assist might actually be the trick though.

0