The snap migrates itself to FerretDB

WeKan runs on FerretDB. Every platform, every install: the snap, Docker, the bundles, the Helm chart. MongoDB is still in the amd64 and arm64 snaps, but only to be read — it is the source a migration copies from, and it is stopped when that is done.

There is nothing to choose and nothing to type. The database snap setting that used to select between the two is gone; so is snap run wekan.database. If your snap still has the setting from an older release, the configure hook says so once and unsets it.

What the migration moves

Text data ends up in SQLite, binaries end up on the filesystem, and nothing is deleted from MongoDB: its files stay in $SNAP_COMMON exactly as they were.

Every collection is copied and then counted against the source, and any that came up short is copied again — a resume must not skip a collection it never finished (wekan/wekan#6585).

Which MongoDB it can read

A MongoDB server starts only on data whose featureCompatibilityVersion is at most one major version behind it, so no single mongod can open everything the WeKan snap has shipped since 2019. The snap therefore carries four readers and tries them newest-first:

All of them are read-only here: the data is copied out through the bundled Node.js driver (or mongoexport for 3.2) and the reader is stopped again. On x86_64, mongod 5.0 and 7 need AVX, so a CPU without it runs them under qemu-user emulation through bin/cpu-exec — slow, and it happens once.

The other architectures (armhf, s390x, ppc64el, riscv64, …) never had a MongoDB: MongoDB publishes no server for them, and those snaps have been FerretDB from their first boot.

When it runs, and what happens if it fails

It runs by itself, on the first start after the snap finds MongoDB data that has not been migrated. WeKan keeps serving from MongoDB while it works, and the progress dashboard is on the normal WeKan address.

A failure does not end it. The attempt is recorded — how many, when, and which snap revision — and retried:

• immediately after the next snap refresh, because the next release is the most likely thing to have fixed it (this is how the mongod 5.0 reader reaches an instance that was stuck before it existed);
• otherwise after a wait that doubles from an hour up to a day, so a migration that cannot succeed yet costs one attempt a day rather than one per start.

snap run wekan.migrate retries now, from scratch. snap set wekan migrate=off stops it completely — an admin saying "not now" is a decision, not a failure.

Two copies, and how they are reconciled

$SNAP_COMMON is shared across snap revisions, and snap revert does not roll it back — so an instance can end up with a MongoDB and a FerretDB that have both been written to since they were copies of each other.

Both copies can be read, so they are read: each database is started on a temporary port, its documents are counted and the newest timestamp in them found, and the documents that exist in only one of them are inserted into FerretDB. Nothing is overwritten and nothing is deleted.

This is safe because of how WeKan stores history: activities, comments and the change-history rows are append-only (History.md), so merging them can only add to what a card's Activity feed shows. Work done on the copy that is not being served becomes readable in the History of the one that is, instead of sitting in a database nobody opens.

Earlier releases answered this differently — they switched the snap to database=mongodb, which is where "WeKan changed to old MongoDB data" came from, and when the staleness detector guessed wrong (#6583) it put a site on a copy that was weeks behind. Merging is the repair; MongoDB is served only for the one start where the merge could not run, and it is tried again on the next.

Which database is running, and why

Nothing stores that decision — it is read from the data itself, by bin/database-role, and every part of the snap asks the same helper:

sudo snap run wekan.problems         # among other things, says which database is serving

• ferretdb — the migration is done and the FerretDB database has data in it (or there is nothing to migrate at all, or the MongoDB files cannot be read by this snap and the FerretDB copy can).
• mongodb — a migration is still owed, and MongoDB holds the data. That includes the case where a migration is marked done but its FerretDB is missing: an empty FerretDB beside a MongoDB full of boards must never be served, because it looks exactly like total data loss.

An interrupted migration is told apart from a finished one by the importer's checkpoint (migration-progress.json): with a checkpoint it resumes, without one a FerretDB that has data is a migration that finished — even if its marker went missing.

The commands that still exist

sudo snap run wekan.migrate     # re-run the whole migration now, from scratch
sudo snap run wekan.problems    # status: which database, what has run, what failed
sudo snap set wekan migrate=off # stop the automatic retries (and on to resume)