Settings → Storage: choose where Kiri keeps images, containers and volumes (for example an external drive)

Open
#355 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
20/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Active
Tech stack
rust

Research direction

Start with crates/kiri-image/src/store.rs at Store::default_root() and the referenced Linux agent and Windows WSL setup paths to map how each platform locates its data. The issue also points to daemon/mod.rs, setup.rs, and the related issues #122, #123, and #345. Before implementation, the five open design questions need decisions; done means a cross-platform settings flow can safely move existing data, handle unavailable storage, and preserve current defaults.

Written by the indexing model from the issue text.

Description

enhancement
What you're trying to do

My Mac's internal disk is small, and images, build layers and volumes add up quickly. I'd like to keep Kiri's data on an external SSD, for example /Volumes/External/kiri, without losing what I already have. OrbStack and Docker Desktop both have a setting for this.

Today the only way is KIRI_DATA_DIR, which is undocumented for users, has to reach the LaunchAgent somehow, and doesn't move anything: Kiri starts empty in the new folder.

What you'd like Kiri to do

An optional Settings → Storage tab, the same on macOS, Windows and Linux:

  • Shows where the data is, how much it uses and how much is free there.
  • Change… picks a folder, checks it, moves the data, and restarts the engine. Reset to default moves it back.
  • Users who never open it see no change at all: no new file, same paths.
What I'd propose (open to any of it)

Where the choice is kept. A small ~/.kiri/storage.json with the folder. It goes beside vm.json and never inside the data folder itself. Store::default_root() (crates/kiri-image/src/store.rs:133) would look at KIRI_DATA_DIR first, then this file, then today's default, so Kiri Dev and anyone using the variable are unaffected. ~/.kiri itself stays in the home folder, as daemon/mod.rs:49 intends.

Checks before moving. The path must be absolute and not inside the current folder, or the other way round. The folder must be empty or already hold Kiri data, and writable (a probe file is written and removed). The free space must cover what the data actually uses, plus a margin. FAT and exFAT are refused, because they can't hold a sparse 64 GiB data.img.

The move.

  1. Stop containers, noting which were running, and retire the VM.
  2. Copy into <target>/.kiri-moving.
  3. Verify the copy: blobs by their digest, everything else by size and content.
  4. Rename into place and write a marker file.
  5. Write storage.json atomically. That write is the commit point.
  6. Restart the engine and start the containers that were running.

If anything fails before step 5, only .kiri-moving is removed and Kiri carries on where it was. The old data is kept until Kiri has started from the new place; then the tab offers Remove old data. Progress is polled the way Settings → Import does it.

A drive that isn't there. If the configured folder has no marker (the drive isn't mounted), the daemon doesn't create a fresh tree there. It stays up with a clear "storage unavailable" error and a banner (Retry, Choose another folder, Use default), rather than exiting.

Things I found while looking into it (main at 260cf66, macOS 27.2, Apple silicon)
  • std::fs::copy fills in sparse files across volumes. A 2 GiB sparse file using 16 KB took 2 GiB at the destination when copied to another volume. On the same volume, and with macOS cp, it stayed at 16 KB. A naive move of vm/data.img would therefore write all 64 GiB, so the copy needs to skip holes (SEEK_DATA/SEEK_HOLE).
  • A missing folder gives an unhelpful error. With KIRI_DATA_DIR on a volume that isn't mounted, kiri daemon exits 1 with only Permission denied (os error 13). The LaunchAgent's KeepAlive {SuccessfulExit=false} would restart it over and over.
  • Nothing stops two daemons sharing a data folder. If they use different sockets, both can open the same data.img; there is no lock. A lock file held by the daemon would close that gap.
  • Kiri Lite's links need repointing. ~/.kiri/bin/docker and cliPluginsExtraDirs hold absolute paths into the data folder (setup.rs:523-546), so a move would have to update them.
  • Linux keeps data in two places. The agent's /var/lib/kiri is fixed (kiri-init/src/linux/mod.rs:34 and three other places), and the host-side metadata follows default_root(). Both would have to move.
  • Windows keeps the WSL distro in %LOCALAPPDATA%\Kiri\wsl (kiri-win/src/setup.rs:532). Moving it would mean moving the distro's VHDX.
  • Related: #122/#123 (trim) decide how much a move has to copy, and #345 is why the move stops containers before copying.

I ran a source build with KIRI_DATA_DIR on an external APFS drive (USB, ownership disabled). The VM, pulls, volumes, networks and amd64 images all worked, and tests/e2e.sh passed apart from the Finder checks, which I had turned off.

Questions before I write anything
  1. Is ~/.kiri/storage.json the right place, or would you rather have a key in an existing settings file?
  2. Linux: should the agent's data root become configurable, or would you prefer another way, such as a systemd mount?
  3. Windows: is moving the WSL distro's VHDX acceptable, and does it belong in the first version? CONTRIBUTING asks for all three systems in one PR, so I'd plan for it unless you say otherwise.
  4. Keep the old data until the user removes it (my preference), or remove it once the copy is verified?
  5. Is refusing FAT and exFAT targets OK?
Would you like to build it?

Yes, once we've agreed on the approach

Dominant language
Rust
Stars
7.5k
Forks
329
Avg merge
7h 48m
Merged PRs (30d)
105

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from libracoder/kiri

All issues in libracoder/kiri

Similar issues

More Rust issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.