Migrate lock timeout on PlanetScale

When using Migrate with PlanetScale, sometimes you get an error message like this:

Error querying the database: Server error: `ERROR 70100 (1317): held locks released: vttablet: rpc error: code = Aborted desc = transaction 1621634564893708666: ended at 2021-05-21 23:54:48.118 UTC (exceeded timeout: 20s) (CallerID: unsecure_grpc_client)'

Also see all the related reports here: https://github.com/planetscale/beta/discussions/34

Two possible causes we can think of:

1. Wait to long when CLI prompts your for migration name
2. Migrations take “to long” to run (Reproducible easily with https://stackoverflow.com/questions/4284524/how-and-when-to-use-sleep-correctly-in-mysql)

We should try to reproduce both so we know if we have something that needs fixing. It might also be worth to investigate if something like this could happen on other database providers as well or if this is PlanetScale/Vitess specific.

(Potential solutions: Only lock after migration name is known, do not lock on PlanetScale at all, etc)

[obibring]

Sure, thanks for the quick reply @janpio

[jde]

I used prisma migrate dev as my primary workflow through the bootstrapping of my app and saw a gradual increase in the time it took.

When I reached 64 migrations (most of these were attempts to debug #10758 issues) my migrate flow hit the 20s Planetscale timeout and I got the error above.

This happened even if I pre-typed my migration name as the process started so that it immediately filled and didn’t add any real time to the process.

I’m using prisma db push now, which executes in 1.1s.

I will create fresh dbs locally and on planetscale and run a single initial migration soon. If this jumps me back into the timeout I’ll report back here.