`Sorry, something went wrong` when opening task in 1.5.0
Hey, i upgraded our twenty from version 1.3.1 > 1.4.1 > 1.5.0 in docker. When i open task or any task filters they show this error, my console shows
TypeError: c is undefined. I also noticed that you guys didn't push the latest minor versions (1.5.1 and 1.5.2) to the docker hub, maybe the issue fix is in there but i can't use it.
53 Replies
Hey @Caspersonn, could you please share any relevant twenty-server logs ?
Unable to reproduce on my side 🤔
What's your workspace version in database ? ( core.workspace.version )
Sure, these are some of my latest logs. We also encounter
Uncompiled message detected! since version 1.4.1.Still at
1.3.1 which is weird...
@Prastoin There is also some extra background information. We plan our upgrades for our production environment, to prepare for that i migrate the database to our non-production environment to test the upgrade. This is a different server with the same environment variables. Could this be a issue?Still at 1.3.1 which is weird...High chances this is the root cause Mhmm I don't think that should be an issue, it's even more a good practice from my point of view What I would recommend here is, either restore a backup from 1.3.1 if you have one not that old, or directly try downgrading to 1.3.1 with the instance you've attempted to migrate to 1.5.1 with ( it's more risky but do-able ) From 1.3.1 re-attempt 1.4.1 upgrade, it should be failing regarding your workspace version. From this point please share whole twenty-server logs by dm or obfuscated if necessary. And we should be able to investigate further from here
Okay. I think i will choose to start over with a new backup from our production environment, hopefully it will work this time. Thanks!
Ok sounds good ! Feel free to reach out !
@Prastoin. I already encounter an error when i run v1.3.1 on non-production with production database image.
Is this a known issue?
The upgrade went smooth without any errors in the console, i still encounter the error i mentioned but nothing in the console or any logs about that. but i noticed if i create a separate view with no filter just like all tasks, then it does work.
The other weird thing is that
Assigned to Me changed to All Tasks as well. The image i attached is from our production environment running v1.3.1.
Mhm you might be facing a view corruption
You've completly refactored the view lately to be store as core engine entities and not workspace schema level resources
Could you please confirm your workspace verison in database has been upgraded too ?
If yes manually fixing corrupted will do the trick
The workspace version is updated and thus
1.5.0. Is this manaul fix gonna work for future updates? Or could we encounter corruption again?I don't think you will be prone to same issue in the future
From my understanding you might had an issue with the existing views migration to the new core table
The command handling the view migration might override existing ones if already migrated and mutated since the migration
Are all your views broken ?
What's not working exactly ? Lets investigate a little bit further before going for the easy manual fix
Yeah. Our old database was very corrupted and i migrated manaully to a new database, that was about 3 months ago. All the views in tasks that where created before the migrations are broken, if i create a new one in tasks it does work. All the other views seem to be working fine.
Mhmm
Please proceed to create a backup of current state and try running:
in your twenty-server
Please not that previous view deletion will be done later in the future, we prefer proceeding step by step before introducing any potential data loss
Okay
In a nutshell you will still have previous views in workspace schema, they won't get hydrated anymore tho
If you re-run the command you might override any latest update made to colliding view
Alright. Thanks for explaining. The backup is created and i ran the command.
Deleted your message just in case
It contained your view names
Looks like your views has been migrated
Any errors now ?
Sadly yes, the same as before.
Fair enough
mhm what's not working exactly ?
Any errors or stacktrace ?
though the command has been introduced before 1.5, manual fixing seems like a bad idea now
Nope. The frontend states
Sorry, something went wrong Please refresh the page.. and the console TypeError: c is undefinedCould you please run a quick
yarn command:prod cache:flush ?Sure
Nothing server side ?
No nothing
So this would be frontend side 🤔
Yeah, seems like it. I tried chromium same issue slightly different error though
And after a flush the error still persists
The error on chromium is
TypeError: Cannot read properties of undefined (reading 'type')
Could there have been a migration at another version that failed and caused this?If your workspace version is up to date it should not
Could you please share a find many view response with a brand new view and a legacy failing one ?
if i may ask, how would i do that?
nevermind I've been able to reproduce 👍
Okay. Are you able to figure out the issue?
Investigating !
I've raised an issue @charles is taking it will let you know once we have a fix !
https://github.com/twentyhq/twenty/issues/14369
Thanks for reporting that !
Alright. Thanks @Prastoin!
You're welcome !
Fix deploy in 1.5.3 ! Unfortunately 1.5.0 had a corrupted command, we've removed it from docker hub
You will need to either restore a bakcup of a previous version or manually set to 1.4 your workspace version for the 1.5.3 upgrade to run the upgrade commands
Please let me know
Hey @prastoin. The issue is still there... i have upgraded from our last production version (1.3.1). So I have downloaded the sql database that upgraded to 1.5.3 run twenty with that locally now, i run in to the same error but i get more detail. Do you know anything about this?
Hey @Caspersonn, is your workspace version at 1.5.3 ?
Jep
The does not seem to be minified
Which is a new problem
but this gives more hint about the issue you're facing tho
Yeah exactly
Is it failing at the very same place ?
Same view ?
Are new views still working ?
Yes at
All Tasks only
NEw views work, old ones don'tMhm
Could you please inspect all your objectMetadataItems searching for an object having no fieldMetadataItems ?
you can either search at your bootstrap app a graphql request ( or use https://chromewebstore.google.com/detail/graphql-network-inspector/ndlbedplllcgconngcnfmkadhokfaaln )
query ObjectMetadataItems {
Or directly use the api playground of your twenty instance accessible from the settings in GRAPHQL COREGraphQL Network Inspector - Chrome Web Store
Simple and clean network inspector for GraphQL
You guys did not activated the morph relation feature flag by any chance ?
No, not that i know of
will do
Okay @prastoin. I found the query in the GraphQL Chrome plugin. and there is a lot of data in the response. Could i send to you?
Yes of course be aware that it might contain some private info
I will send it to you in a private dm
Nothing wrong here, I suspect one of migrated view to make a reference to a deleted field
I was thinking that ass wel, but how can i easily check that
TSConfig Reference - Docs on every TSConfig option
From allowJs to useDefineForClassFields the TSConfig reference includes information about all of the active compiler flags setting up a TypeScript project.
@Prastoin. I have an update on this mater. I upgraded to the latest release which is now 1.6.7, but the error still occurs. There is now a more descriptive error. Could we take another look what is going on? This id seem to be a
fieldMetadataId in table core."viewField" and thus links back to the table core."fieldMetadata".
Hey @Caspersonn, we had no identical reports nor experienced same issue on cloud prod
As discussed I think you got a corrupted view somehow
I would recommend manually removing it a re-creating a new one