Umami update 1.37.0 (package version 1.6.0) won't start, errors in logs
-
Hello,
Updating Umami to the latest version released a bit ago shows Umami constantly trying to start but fails, and the logs show these errors:
Aug 09 23:00:31 Gerror Command failed with exit code 1. Aug 09 23:00:31 Ginfo Visit https://yarnpkg.com/en/docs/cli/run for documentation about this command. Aug 09 23:00:38 ==> Changing ownership Aug 09 23:00:38 => Building database Aug 09 23:00:38 => Existing prisma managed, running migrations Aug 09 23:00:39 Gyarn run v1.22.17 Aug 09 23:00:39 Gwarning Skipping preferred cache folder "/home/cloudron/.cache/yarn" because it is not writable. Aug 09 23:00:39 Gwarning Selected the next writable cache folder in the list, will be "/tmp/.yarn-cache-1000". Aug 09 23:00:39 G$ /app/code/node_modules/.bin/prisma migrate resolve --applied 01_init Aug 09 23:00:39 Gwarning Cannot find a suitable global folder. Tried these: "/usr/local, /home/cloudron/.yarn" Aug 09 23:00:39 Environment variables loaded from .env Aug 09 23:00:39 Prisma schema loaded from prisma/schema.prisma Aug 09 23:00:39 Datasource "db": PostgreSQL database "dbc437d27f129b45cca6a7b0ce4bcb6fa9", schema "public" at "postgresql:5432" Aug 09 23:00:39 Error: P3008 Aug 09 23:00:39 Aug 09 23:00:39 The migration `01_init` is already recorded as applied in the database. Aug 09 23:00:39 Aug 09 23:00:40 Gerror Command failed with exit code 1. Aug 09 23:00:40 Ginfo Visit https://yarnpkg.com/en/docs/cli/run for documentation about this command. Aug 09 23:00:40 Gyarn run v1.22.17 Aug 09 23:00:40 Gwarning Skipping preferred cache folder "/home/cloudron/.cache/yarn" because it is not writable. Aug 09 23:00:40 Gwarning Selected the next writable cache folder in the list, will be "/tmp/.yarn-cache-1000". Aug 09 23:00:40 G$ prisma migrate deploy Aug 09 23:00:40 Gwarning Cannot find a suitable global folder. Tried these: "/usr/local, /home/cloudron/.yarn" Aug 09 23:00:41 Environment variables loaded from .env Aug 09 23:00:41 Prisma schema loaded from prisma/schema.prisma Aug 09 23:00:41 Datasource "db": PostgreSQL database "dbc437d27f129b45cca6a7b0ce4bcb6fa9", schema "public" at "postgresql:5432" Aug 09 23:00:41 Aug 09 23:00:41 2 migrations found in prisma/migrations Aug 09 23:00:41 Aug 09 23:00:41 Error: P3009 Aug 09 23:00:41 Aug 09 23:00:41 migrate found failed migrations in the target database, new migrations will not be applied. Read more about how to resolve migration issues in a production database: https://pris.ly/d/migrate-resolve Aug 09 23:00:41 The `02_add_event_data` migration started at 2022-08-10 06:00:10.447969 UTC failed Aug 09 23:00:41 Aug 09 23:00:41 Aug 09 23:00:41 Gerror Command failed with exit code 1. Aug 09 23:00:41 Ginfo Visit https://yarnpkg.com/en/docs/cli/run for documentation about this command.Not sure if this is a packaging issue or a Umami issue. Rolled back for now.
-
Looks like maybe related to https://github.com/umami-software/umami/issues/1229
-
@girish yeah, see also the note here https://github.com/umami-software/umami/releases
Important
This release includes a new database migration. In this update we are making some changes to Events. Now instead of event_type and event_value columns, there is going to be a single renamed column, event_name. Having the two separate columns caused a lot of confusion regarding what to store in each field. There is now a new table called event_data which is used to store event related data. In the future you will be able to run reports against your events and event data.
Now to track events, you simply choose an event name, like button-click and optionally send some JSON data along with it.
umami.trackEvent('button-click', { userId: 123, buttonName: 'Newsletter signup' });
When you start the app for the first time, the database changes will be automatically applied. Your data however will not be automatically migrated because we can't determine what event_name you will want to use. Your old event table has been renamed to _event_old, which you can use to manually move data to the new table. If you've never used events, feel free to delete the _event_old table.
-
The root cause is an upstream issue - https://github.com/umami-software/umami/issues/1393 and https://github.com/umami-software/umami/issues/1406 . Umami changed their db migration to prisma some releases ago. However, the index names created by prisma and old migration logic is different.
I don't know if this is easy to automation for migration, let's see what they come up with upstream.
-
@girish I know you revoked the previous update, but there is now another Umami 1.37.0 update available - is that one safe to update to?
It's a bit unclear, even though there is this warning
This update is a pre-release and not considered stable yet. Please update at your own risk.
-
@jdaviescoates yeah, did that because we didn't want new installations to have the old schema. It seems people in old installs have to re-install - https://github.com/umami-software/umami/issues/1406#issuecomment-1213355769 . The schema migration between old umami installs and the new one is not easily automatable it seems if you follow that thread.
-
OK, I have gone ahead and re-published as package version 2.0.0. This means that the app won't auto upgrade.
I put this in the changelog:
- IMPORTANT: The database upgrade fails if you have been running umami for a while. There is no easy work around for this, you can manually try the database modifications in this upstream bug after you upgrade and place the app in Repair mode.
- If possible, just start with a fresh installation of Umami. It can be quite a time sink to fix the database schema and you might still hit some issue later in future versions.
-
Thanks @girish but re:
The database upgrade fails if you have been running umami for a while.
How long is "a while"?
Seems a bit vague.
Also, it's hard for me to grok what difference how long one has been running the app would/ could make any difference?
Surely the db schema has changed from one version to the next even if I've only been running it for a millisecond? (I've actually been running it since May - is that "a while"?)
-
@jdaviescoates Best I can make out, it all started with Umami release 1.33 - https://github.com/umami-software/umami/releases/tag/v1.33 . This release was in Cloudron package 1.2.0 (around Jun 27 2022). From upstream release notes, "This release will use Prisma migrations for the first time".
If you installed umami before that, I think upgrades will fail.
Best way to check really is to clone current app and try to upgrade the cloned instance and see if it upgrades or not
-
@girish This should be unacceptable to Cloudron IMO. If Umami can't provide reliable updates or at least release a repaired update after it's known to be broken, it shouldn't be in Cloudron. I use Cloudron so I don't have to deal with these kinds of things and certainly won't reinstall and start from scratch due to a bug in an update - that would include re-tagging 7 sites currently tracked in the software.
-
@echokos I understand your sentiment, but disagree.
It's impossible to know the future and any app could potentially change resulting in something like this happening (although it's obviously more likely with less mature apps).
But regardless of this existing issue I'd hate for Umami not to be on Cloudron because it's great! So easy to get up and running and works really nicely.
-
I agree, such situations where upstream leaves us with no good path to migrate existing instances are pretty bad. We usually spend much time on finding ways to migrate, as it is indeed part of what we see as value we provide. In this case however @girish was next to me for quite some time trying to find a way, but after explaining the introduced mismatch of table foreign key constraint ids, it simply wasn't possible.
Depending on the case, we sometimes create a new appstore id for the app package to avoid existing installations to break on update, but it also means there won't be any more updates and users are still stuck
-
This is a difficult one. I feel like to some degree I've lost my use of Umami now, because I can't upgrade properly and it doesn't seem easy enough to deploy a copy of the latest version and migrate the data to it.
That's not Cloudron's fault at all, I'm just very surprised this issue wasn't taken more seriously by the upstream developers. Guess I'll have to stick to Matomo for a while longer.I had been using Umami (in addition to continuing with Matomo) for the past 2 or 3 months, hoping to cut over to it fully for my clients in another month or so and ditch Matomo after they had about 90 days of data, but that seems unlikely with the breaking changes they made to Umami and being unable to proceed without losing data. Unless I've misunderstood something about the situation (fingers crossed I may have and there's maybe a way I've overlooked to migrate / upgrade), but I'm guessing that's not the case and may need more time to review.
-
@d19dotca Rough indeed.
You could think about using two instances and a cut-over by installing a new Umami, copying over the config file, then cutting over traffic to the new instance, keeping the old one around a bit, in case you need the data, which will become less useful over time and ready for cleanup in a month or two.
-
N Neiluj referenced this topic on
-
-
