FIRESTORE (8.2.1) INTERNAL ASSERTION FAILED: Unexpected state

[Benny739]

We're seeing this error a lot in sentry atm.
I can not reproduce this or say where it comes from or what it consequences are.
I guess it is related to offline support.

[google-oss-bot]

I couldn't figure out how to label this issue, so I've labeled it for a human to triage. Hang tight.

[wu-hui]

Hi there,

Our SDK strips error messages (for sizes), and the error message you provide is the stripped one. This is not very helpful for debugging unfortunately.

Given that you cannot reproduce this yourself, the only thing we can try to to roll some of your users to a SDK without the stripping, and see if we can get some better error messages.

If I provide you instructions to get a SDK without stripping, are you willing to roll this out to a portion of your users?

Thanks

[Benny739]

Hi,

thanks for your help. Yes we're willing to try it out. What are the risks when using unstripped error messages?

[bigjimhere]

This error started appearing in our Sentry logs some time around version 7.20.0, with offline persistence disabled. The error occurred sporadically when firebase.firestore().disableNetwork() was called seconds after app startup. Removing firebase.firestore().disableNetwork() from our code resolved the issue.

My guess is that our was app was occasionally calling disableNetwork() while Firestore was still initialising. This may have been interrupting Firebase/Firestore authentication in some way, perhaps by cutting off the connection before authentication could complete. Read and write operations consequently seemed to fail, perhaps due to an "unexpected auth state" (see below).

If you are not using disableNetwork() in your app perhaps Firestore initialisation/authentication is being interrupted in some other way? https://stackoverflow.com/questions/63319638/firestore-internal-assertion-failed-unexpected-state-when-unit-testing-with-j

Our app initialises multiple Firebase instances on startup. Each instance immediately starts reading a fairly large amount of data using Firestore onSnapshot listeners. We were seeing the following errors in Sentry, all of which were resolved when we stopped using disableNetwork():

"FIRESTORE (8.2.6) INTERNAL ASSERTION FAILED: Unexpected state"

"FirebaseError: missing stream token" (less frequently)

"TypeError: undefined is not an object (evaluating 'e.batchId')" (very rarely)

"FirebaseError: the first request in a stream cannot contain writes" (very rarely)

The "Unexpected state" error seemed to cause both reads and writes to fail ("hang") without a promise rejection. We were able to reproduce this at the time, but I'd need to do more testing to confirm. Prior to 7.20.0 we had been able to reproduce hanging writes, but in earlier versions the writes failed silently without the "Unexpected state" error. Offline persistence was enabled for those earlier tests. If the "Unexpected state" error was the cause in those earlier cases it is possible that the writes failed silently because the error messages above weren't added to the SDK until some time around 7.20.0. For example I noticed the following in the release notes for 7.23.0- "The SDK now includes more information in the error message for failed IndexedDB transactions."

The "missing stream token" error caused reads to fail but I can't confirm whether or not writes were affected. As a side note we would see onSnapshot listeners failing with the following error message around the time that this error was logged in Sentry: "FirebaseError: [code=already-exists]: Target ID already exists: 6". This error also disappeared when we stopped using disableNetwork().

I'm not sure what effect the last two errors above had on the application.

I've attached a stack trace for "FIRESTORE (8.2.1) INTERNAL ASSERTION FAILED: Unexpected state"

Firestore Unexpected State Error.txt

UPDATE:

It looks like the "Unexpected state" error was thrown when our app called disableNetwork() on multiple Firestore instances, waited a few seconds and then called enableNetwork() on those same instances. This was likely happening while Firestore was initialising.

[wu-hui]

@bigjimhere Are you describing the same project as @Benny739 here, or seeing similar issues with your own project?

Either way, thanks for the information. I will try to call disableNetwork with multiple firebase instance myself.

@Benny739 I will provide a SDK with all error messages enabled, hopefully soon.

[bigjimhere]

@wu-hui I'm referring to issues with my own project.

[mlahp7]

We too got a ton of errors in our Sentry that begin with FIRESTORE (8.2.4) INTERNAL ASSERTION FAILED: Unexpected state. It seems to have started happening for us yesterday.
@bigjimhere @Benny739 is the error coming from just one user or multiple?

[mcarriere]

We're experiencing this issue with FIRESTORE (8.2.4) INTERNAL ASSERTION FAILED but we're not calling disableNetwork anywhere in our code. It happened over 43k times from a single browser today. Luckily it was an internal user so we could tell them to refresh the page but it's not really something we want to tell our clients.