-
Notifications
You must be signed in to change notification settings - Fork 926
Firestore indexDB persistence corrupted after user "Clear site data" #8593
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
I couldn't figure out how to label this issue, so I've labeled it for a human to triage. Hang tight. |
Hi @entrenadorhispano, thank you for sharing a very detailed reproduction steps, video recording and the code snippet. I was able to easily reproduce same the behavior. I tried both the SDK version that you used and the latest 11.0.1 version. I also verified that adding I will inform our engineering team of this issue. Please keep an eye out for any additional information they may provide. |
Another Part of the BugI found another part of this bug. Once the document cache is corrupted, even getDocFromServer() will fetch the cached, non-existent document. Severity of the BugI would also advise that this bug is severe and makes Firestore offline persistence unusable in production web apps. For example: In JS Frameworks, It May Be Even WorseI also tried the same process in frameworks like Svelte. There, you don't even have to modify data in the console. Simply clearing site data is enough to trigger this bug and start serving cached, non-existent documents. Thank you very much for your support and time in dealing with this. |
I would like to share my experiences with persistence on this issue:
I want to thank @entrenadorhispano for the very detailed description of the problem. This is a very serious issue and must be addressed urgently. It's incredibly difficult to debug unless you have a deep understanding of the inner workings of Firestore. Problems like this can drive developers away from Firebase. Personally, I am very close to abandoning Firebase entirely because of issues like this. I didn't create an issue on my own earlier because I just couldn't pinpoint the problem and wasn't able to make a good description without that. Also, there were many cases where I thought it was my fault and probably fixed it somehow, but it the problem came back again. Please prioritize a fix for this. |
Hi guys, any news on this issue? |
Apologies for the delay. I have not had time to dig into it yet. I will ASAP though. |
After investigation, what's happening is that when the IndexedDB database is wiped the database connection gets closed and Firestore's internals recover by transparently re-opening the database connection as if nothing had happened. But, the snapshot listener causes a bunch of state to be loaded from the IndexedDB database into memory, which unbeknownst to Firestore, becomes stale once the IndexedDB database is cleared. When an updated snapshot is received from the backend then Firestore tries to save it into the IndexedDB database, which had since been wiped, and the data it writes is not consistent because it assumes its in-memory cached data is still in the database. So when the website is refreshed in the browser it tries to load the partial data from the IndexedDB database, which is basically corrupted. Hopefully that explanation made some sense. As for a "fix", the best thing I can think of is to fail loudly if the IndexedDB database connection is unexpectedly closed, and disallow any writes to the IndexedDB database until a refresh. That way, no partial and/or inconsistent data can be written. Then, upon refreshing the web page, everything will work as if it were the first page load and the IndexedDB database were empty. After clearing site data, I presume that a page refresh would be the "natural" thing to do anyways. Do people generally expect that a web site continues to function after clearing site data? I know I wouldn't but I'm also a technical person and not an average user. |
I'm happy to report that I have a fix in progress: #8871. I'll update here once it's merged. |
Hey, these are excellent news! Thank you for your time and effort! 1. The problem isn't just local to our app.Sometimes people clear their entire browsing history... So, it's not just that they delete our app cache or site data. (In my tests, clearing the entire browsing history from another page also creates the bad Firestore state... 2. Is there any way or method to detect when Firestore is broken?I've tried to fetch documents explicitly from the server using getDocFromServer(reference)... and it fails when Firestore is already broken. (This may be another related bug.) 3. Simply refreshing the app may not solve the problem.The only way I could reset the state is to use clearIndexedDbPersistence(db). 4. Regarding the current fix implementation... is it a patched version, or do we have to wait?I saw the fix is merged. THANKS A LOT! |
The fix has now been merged. Unfortunately, it didn't get merged in time for next week's release, so it will get included in the release after that, which is scheduled for end of April. If you'd like to test out the fix prior to the release just let me know and I'll put instructions here for building the Firebase JS SDK from source. |
Operating System
Windows 11, Chrome, Edge, Android, iOS
Environment (if applicable)
Chrome all major versions, Edge, Safari iOS tested on 16
Firebase SDK Version
10.14
Firebase SDK Product(s)
Firestore
Project Tooling
Vanilla JS.
Also tried on other web frameworks.
Detailed Problem Description
If Firestore persistence is active and a doc is cached. Then the user clears the site data, and and doc is updated on backend.
Cache gets corrupted and now the firestore listener returns NULL document FROM CACHE
I, leave some concrete steps for reproduction below.
(You have to put your own firebase_api_key for security)
Firebase.err.mp4
Steps and code to reproduce issue
The text was updated successfully, but these errors were encountered: