Twenty•4mo ago

Upgrade from 1.0 > 1.3 broke twenty

After a recent upgrade i v1.3 I get "an error occured" when logging in. And container logs for twenty-server says this:

Computing new Datasource for cacheKey: 10f069ce-793a-4ada-b841-0bb11359b874-17 out of 0 │
metadata query time: 1.6534839999949327 ms                                              │
Uncompiled message detected! Message:                                                   │
                                                                                        │
> sVtYE5                                                                                │
                                                                                        │
That means you use raw catalog or your catalog doesn't have a translation for the message and fallback was. │
ICU features such as interpolation and plurals will not work properly for that message.                     │
                                                                                                            │
Please compile your catalog first.                                                                          │

Computing new Datasource for cacheKey: 10f069ce-793a-4ada-b841-0bb11359b874-17 out of 0 │
metadata query time: 1.6534839999949327 ms                                              │
Uncompiled message detected! Message:                                                   │
                                                                                        │
> sVtYE5                                                                                │
                                                                                        │
That means you use raw catalog or your catalog doesn't have a translation for the message and fallback was. │
ICU features such as interpolation and plurals will not work properly for that message.                     │
                                                                                                            │
Please compile your catalog first.                                                                          │

I upgraded from 1.0 to 1.1 to 1.2 to 1.3. I made a db backup in between all upgrades. Not sure if these translation errors have anything to do with the actual problem preventing me from logging in. Are there other logs I should check? Is this a known issue? Did I miss some post-upgrade steps?

42 Replies

flomero•4mo ago

seams to be the same issue thats discussed here: https://discord.com/channels/1130383047699738754/1406698879575130254

EbenezerIbizaOP•4mo ago

I just upgraded to 1.4.0 now. After the upgrade, I still get the blank screen....

EbenezerIbizaOP•4mo ago

If I change to a new browser which has not logged into twenty, I can log in and then I get this briefly:

charles•4mo ago

@EbenezerIbiza let's work from there could you share your network tab and specifically the faily requests errors?

EbenezerIbizaOP•4mo ago

Btw. this error disappears WAAAY too fast. Please consider having errors stay around. Yeah, no failed requests. All 200 OK

EbenezerIbizaOP•4mo ago

I hope this was what you meant?

EbenezerIbizaOP•4mo ago

I have this also. Not sure if relevant

charles•4mo ago

filter on XHR requests I think the metadata queries are actually returning errors, could you check their response it's 200 but that's because graphql standard requires it to always be 200 and errors are embed in the response

EbenezerIbizaOP•4mo ago

I have tried filtering on XHR. Not sure what I am looking for though

EbenezerIbizaOP•4mo ago

Huzef•4mo ago

Click on the file column it will expand the view then look for Response tab

EbenezerIbizaOP•4mo ago

Huzef•4mo ago

this one seems fine, could you repeat it for the 3 other above it?

client-config
metadata
metadata

client-config
metadata
metadata

EbenezerIbizaOP•4mo ago

Is this what you're looking for? I dont see any errors in this

Huzef•4mo ago

Yes, but your backend seems fine

EbenezerIbizaOP•4mo ago

Thats good, I guess 🙂

Huzef•4mo ago

Could you DM me this full console log

EbenezerIbizaOP•4mo ago

Sure From firefox or from containers?

Huzef•4mo ago

Firefox

EbenezerIbizaOP•4mo ago

Tried to DM you

Huzef•4mo ago

Sent you a friend request

subscribe https://REDACTED/assets/index-DVBRJWXR.js:97
index-DVBRJWXR.js:4117:74

subscribe https://REDACTED/assets/index-DVBRJWXR.js:97
index-DVBRJWXR.js:4117:74

can't convert null to object

can't convert null to object

,yke=({objectPermissions:e})=>{const t=e.restrictedFields;return Object.entries(t)

,yke=({objectPermissions:e})=>{const t=e.restrictedFields;return Object.entries(t)

seems to be throwing the error here

EbenezerIbizaOP•4mo ago

So is this unrelated to the stuff that was fixed in 1.4?

pmckeown•4mo ago

I'm getting the same issue after upgrading from 1.2 to 1.3. Going then up to 1.4 didn't fix it. Happy to test a patch when it comes out but for now I've downgraded to 1.2 and restored the DB. Stack: Self-hosted in GKE on GCP.

EbenezerIbizaOP•4mo ago

Interesting, so I guess it is not just me then.... What steps did you try? @charles : did you have the chance to look at the errors from @NEO ?

charles•4mo ago

Looks good! What timezones are you guys on? Let's schedule some time to check together

EbenezerIbizaOP•4mo ago

CET/CEST Now is good for me. But I am very flexible What TZ are you in? When would be a good time for you?

charles•4mo ago

I'll do an Office Hour tomorrow at 11am (Paris time), would it work for you?

EbenezerIbizaOP•4mo ago

Hi. Yes! I have a slot from 1100-1130 CEST Do you want me to DM you a Google Meet link?

Prastoin•4mo ago

Hey @EbenezerIbiza if Charles in the end isn't available I will be able to take the call

EbenezerIbizaOP•4mo ago

Excellent!

pmckeown•3mo ago

Did you guys come up with a solution? I've just tried to upgrade from 1.2 -> 1.3 -> 1.4 -> 1.5 and I'm still getting the same issue. Deployment works, database is migrated to correct version every time. But after authenticating against our IDP, the redirection halts when the browser gets the page https://<host>/verify?loginToken=<token> Same behaviour happens on Safari, Chrome and Arc. The Authentication against the IDP is fine, redirects back to /verify where lots of subsequent requests for /client-config and /metadata etc are successful with 200 responses from the server, but I am seeing this error in one of the calls to /metadata:

{
    "errors": [
        {
            "message": "column ObjectMetadataEntity.isUIReadOnly does not exist",
            "extensions": {
                "code": "INTERNAL_SERVER_ERROR",
                "userFriendlyMessage": "An error occurred."
            }
        }
    ],
    "data": null
}

{
    "errors": [
        {
            "message": "column ObjectMetadataEntity.isUIReadOnly does not exist",
            "extensions": {
                "code": "INTERNAL_SERVER_ERROR",
                "userFriendlyMessage": "An error occurred."
            }
        }
    ],
    "data": null
}

I'm self-hosting, on GCP Kubernetes. I've taken a backup again at 1.2, and have rolled back so we can use the system. @Prastoin Any ideas?

Prastoin•3mo ago

Hello @pmckeown you're missing a migrate from previous version Please review your workspace version in database, core.workspace.version

pmckeown•3mo ago

Ahh that makes sense. My workspace version is back to 1.2.1. Any clue as to which might have been missed? Looks like this one: https://github.com/twentyhq/twenty/blob/d308f5adf905c889d773c10ad0466dd08a227f1e/packages/twenty-server/src/database/typeorm/core/migrations/common/1755000000000-addIsUIReadOnlyToFieldMetadata.ts Doesn't seem to be an easy way to determine which set of ORM migrations relate to a particular release?

EbenezerIbizaOP•3mo ago

I recently upgraded from 1.5 to 1.6. Since then, I can log in and when I do, it opens Settings. However, I cannot seem to exist Settings. If I click the X in the top, it just goes to settings again. If I go to /objects/companies directly by entereing the URL, it will show me the companies list. But if I go to the "main page" I am sent to Settings I have verified both app version and db version: v1.6.7 version --------- 1.6.7 (1 row) Output from

echo 'echo $APP_VERSION; exit' | podman exec -i twenty-server /bin/sh
echo 'SELECT version FROM core.workspace;' | podman exec -i twenty-db psql -U twenty

echo 'echo $APP_VERSION; exit' | podman exec -i twenty-server /bin/sh
echo 'SELECT version FROM core.workspace;' | podman exec -i twenty-db psql -U twenty

Prastoin•3mo ago

Hello there please create a dedicated help thread 🙏 You can view the file and change its versioning using all the tag until it disapears

pmckeown•3mo ago

OK so this was a tricky one to pin down. For others I will describe the issue and fix. Upgrading from 1.2.1 to 1.3.1 has a breaking change in that if a user's email account is not marked as verified, when they are redirected back from the AuthN provider (Microsoft in our case) the authentication process stops at the browser having loaded the /verify?loginToken=eyJhbG... URL.
Some requests are made from the browser here, and this one to /metadata is the failing with the attached error. Once the user.isEmailVerified record is marked as true, the user can log in again.
Please add this to the release notes or something for 1.3.1 to avoid this hassle for others.

kmbuthia•3mo ago

Thank you for sharing this! I keep coming across a lot of people facing challenges upgrading and that makes me very nervous about continuing to self-host Twenty. I hope the Twenty team are taking note of all these migration issues.

Gaming

Programming

Upgrade from 1.0 > 1.3 broke twenty

Did you find this page helpful?