Firestore: Deleting a couple of documents returned by a limited subcollection query triggers all documents to disappear locally #1548
Comments
aleh
changed the title
|
Thank you for the repro app and video! I'll take a look and get back to you. |
|
Thanks for filing this. It turns out to be the cause of this discussion. We'll treat this as the canonical issue for this bug. The problem exists on all Firestore mobile/web SDKs. For context, when a client is listening to a limit query (and in certain other cases) and a document moves out of the query results the server is allowed to tell the client merely that a document is removed without fully describing why that was. This is an intentional ambiguity that gives us some wiggle room in the server implementation. When this happens the client considers the document to be in limbo and attempts to figure out what happened to that document by requesting it out of band. Without this limbo resolution process, the next time the client ran the query you would see a flicker: the client would serve stale data from the cache which would disappear when the server gave up-to-date results again. This is something we try to avoid. So with that background, the bug is in this limbo resolution process. There's a certain order of events from the server that can cause the client to mistakenly determine that a limbo document doesn't exist and this is causing the client to populate negative entries in its cache. The fix for this is in flight here: firebase/firebase-js-sdk#1014. This alone would not be a problem because the next update from the server would fix the client's cache, but the client is also recording the commit time of the delete as the current time of the limbo resolution read, which is typically later than the last server-side modification. There are corner cases around updates to documents that don't match a query where the server is not obligated to send the absolute latest version of a document so the client will prefer its version of a document if it's later than the server issued one. This compounds the issue and causes it to be persistent. A fix that will allow the client to self-heal its cache is progress here: firebase/firebase-js-sdk#1015. (Please don't be alarmed: we're fixing the js-sdk first because we have a cross-platform test suite written in typescript. It's easier to fix there and then port.) |
|
@mikelehen @wilhuff Thanks for the detailed update! |
This is a force fix for potential existence filter mismatches caused by firebase/firebase-ios-sdk#1548 The essential problem is that when resuming a query, the server is allowed to forget deletes. If the number of incorrectly synthesized deletes matches the number of server-forgotten deletes then the existence filter can give a false positive, preventing the cache from self healing. Dropping the query cache clears any client-side resume token which prevents a false positive existence filter mismatch. Note that the remote document cache and mutation queues are unaffected so any cached documents that do exist will still work while offline.
…tes in it Port of firebase/firebase-js-sdk#1015 Addresses #1548
…tes in it Port of firebase/firebase-js-sdk#1015 Addresses #1548
…tes in it Port of firebase/firebase-js-sdk#1015 Addresses #1548
This is a force fix for potential existence filter mismatches caused by #1548 The essential problem is that when resuming a query, the server is allowed to forget deletes. If the number of incorrectly synthesized deletes matches the number of server-forgotten deletes then the existence filter can give a false positive, preventing the cache from self healing. Dropping the query cache clears any client-side resume token which prevents a false positive existence filter mismatch. Note that the remote document cache and mutation queues are unaffected so any cached documents that do exist will still work while offline firebase/firebase-js-sdk#1019
* Add a schema migration that drops the query cache This is a force fix for potential existence filter mismatches caused by firebase/firebase-ios-sdk#1548 The essential problem is that when resuming a query, the server is allowed to forget deletes. If the number of incorrectly synthesized deletes matches the number of server-forgotten deletes then the existence filter can give a false positive, preventing the cache from self healing. Dropping the query cache clears any client-side resume token which prevents a false positive existence filter mismatch. Note that the remote document cache and mutation queues are unaffected so any cached documents that do exist will still work while offline. * Implement review feedback
…tes in it (#1557) * Update spec tests from the js-sdk * Allow remote updates from watch to heal a cache with synthesized deletes in it Port of firebase/firebase-js-sdk#1015 Addresses #1548
|
I see the js and ios fixes have been merged. Into master. Any ETA on when we can expect new version releases with the fix? |
This is a force fix for potential existence filter mismatches caused by #1548 The essential problem is that when resuming a query, the server is allowed to forget deletes. If the number of incorrectly synthesized deletes matches the number of server-forgotten deletes then the existence filter can give a false positive, preventing the cache from self healing. Dropping the query cache clears any client-side resume token which prevents a false positive existence filter mismatch. Note that the remote document cache and mutation queues are unaffected so any cached documents that do exist will still work while offline firebase/firebase-js-sdk#1019
…tes in it (#1557) * Update spec tests from the js-sdk * Allow remote updates from watch to heal a cache with synthesized deletes in it Port of firebase/firebase-js-sdk#1015 Addresses #1548
This is a force fix for potential existence filter mismatches caused by #1548 The essential problem is that when resuming a query, the server is allowed to forget deletes. If the number of incorrectly synthesized deletes matches the number of server-forgotten deletes then the existence filter can give a false positive, preventing the cache from self healing. Dropping the query cache clears any client-side resume token which prevents a false positive existence filter mismatch. Note that the remote document cache and mutation queues are unaffected so any cached documents that do exist will still work while offline firebase/firebase-js-sdk#1019
|
@scottmas It's now possible to try this out. Add these lines to your |
|
Great! What about Android? |
|
That's coming, but not ready yet. I'm mostly interested in verifying that with this in place your iOS simulator with the broken calendar entries is fixed. |
|
Hey Gil, those pod entries do not work when I try to pod install them. You
were missing a slash "Firebase/Core", but even after I added that it still
doesn't work.
…On Thu, Jul 19, 2018 at 2:15 PM Gil ***@***.***> wrote:
That's coming, but not ready yet. I'm mostly interested in verifying that
with this in place your iOS simulator with the broken calendar entries is
fixed.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#1548 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ACJxubImdTGPlJkwdsaFJpqRsnM1P1h4ks5uIOjvgaJpZM4VSiT2>
.
|
|
Could you elaborate on how it didn't work? Note After pruning unrelated things out I have the following in my Podfile for a test app and I'm able to build successfully: |
|
This has been released in Firebase JavaScript SDK 5.3.0 and Firebase iOS SDK 5.4.1. Android will follow separately. |
|
Any ETA on Android?
…On Fri, Jul 20, 2018 at 11:26 AM Gil ***@***.***> wrote:
This has been released in Firebase JavaScript SDK 5.3.0 and Firebase iOS
SDK 5.4.1. Android will follow separately.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#1548 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ACJxuejR6kyPkEtBwyLZYp9SiH3g7w3Yks5uIhKygaJpZM4VSiT2>
.
|
* Add @davideast as a CODEOWNER (#996) * Embed metadata directly into the RPC call (#979) * Embed metadata directly into the RPC call * [AUTOMATED]: Prettier Code Styling * Use ...args * [AUTOMATED]: Prettier Code Styling * Minimize diff * Add the OAuth assertion back in * Added missing type for optional database url. (#1001) * RxFire Realtime Database (#997) * initial database code * test setup * database tests * auditTrail and database tests * [AUTOMATED]: Prettier Code Styling * [AUTOMATED]: License Headers * Josh's comments. Database docs * [AUTOMATED]: Prettier Code Styling * Firestore docs * auth docs * declaration fixes * switch to peerDeps * [AUTOMATED]: Prettier Code Styling * test config * Expose array transforms and array contains queries. (#1004) Also remove test code that was combining multiple array contains queries since those were disallowed in 04c9c3a. * Move fieldFilter (free function) to Filter.create() (#988) This is a refactoring to unify filter creation across platforms. * Enable firestore sdk to talk to emulator (#1007) * Enable firestore sdk to talk to emulator * [AUTOMATED]: Prettier Code Styling * Revert firestore sdk changes * [AUTOMATED]: Prettier Code Styling * Revert credentials.ts * Cleanup * [AUTOMATED]: Prettier Code Styling * Set webSafe=false * Combine initializeTestApp and initializeFirestoreTestApp * [AUTOMATED]: Prettier Code Styling * Cleanup * [AUTOMATED]: Prettier Code Styling * Update major version since this is a breaking change that will cause the testing sdk to no longer work with old versions of the RTDB emulator * Completely remove admin sdk * Change version back to 0.1.0 * Setting GarbageSource in SyncEngine's constructor (#1010) * b/72533250: Fix issue with limbo resolutions triggering incorrect manufactured deletes. (#1014) This fixes an issue occurring when a limbo target receives a documentUpdate, then a global snapshot, and then a CURRENT. Because there was a global snapshot before the CURRENT, WatchChangeAggregator has no pending document updates and calls SyncEngine.targetContainsDocument() to see if we previously got any document from the backend for the target. See: https://github.com/firebase/firebase-js-sdk/blob/6905339235ad801291edc696dd75a08e80647f5b/packages/firestore/src/remote/watch_change.ts#L422 Prior to this change, targetContainsDocument() returned false because it relies on our Views to track the contents of the target, and we don't have Views for limbo targets. Thus WatchChangeAggregator incorrectly manufactures a NoDocument document update which deletes data from our cache. The fix is to have SyncEngine track the fact that we did indeed get a document for the limbo resolution and return true from targetContainsDocument(). * Updating yarn.lock * Add @firebase/util as a dep of @firebase/testing * Allow remote updates from watch to heal a cache with synthesized deletes in it (#1015) * Write a spec test for the busted cache * Modify spec test to demonstrate deletedDoc issue. (#1017) * Allow updates for targets where the document is modified * Fix getRemoteKeysForTarget() method name in comment. (#1020) While porting I noticed this was slightly wrong. targetContainsDocument() is the method in WatchChangeAggregator. The SyncEngine method I meant to reference is getRemoteKeysForTarget(). * Making sure we don't export 'experimental' (#1023) * Add a schema migration that drops the query cache (#1019) * Add a schema migration that drops the query cache This is a force fix for potential existence filter mismatches caused by firebase/firebase-ios-sdk#1548 The essential problem is that when resuming a query, the server is allowed to forget deletes. If the number of incorrectly synthesized deletes matches the number of server-forgotten deletes then the existence filter can give a false positive, preventing the cache from self healing. Dropping the query cache clears any client-side resume token which prevents a false positive existence filter mismatch. Note that the remote document cache and mutation queues are unaffected so any cached documents that do exist will still work while offline. * Implement review feedback * Add a release note for the fix to #1548 (#1024) * Ensure that we create an empty TargetGlobal row. (#1029) Ensure the v3 migration unconditionally creates the TargetGlobal row. Remove the no-longer-necessary v2 schema migration. * Remove unnecessary `any` (#1030) * Fix an errant any usage * [AUTOMATED]: Prettier Code Styling * Publish [email protected] * Unify local.QueryData with the other platforms (#1027) This makes it line up with it's own docs, and also the other platforms. * Fix to #1027 to allow SnapshotVersion == 0 (#1033) * Add iat to fake access token payload (#1022) * Add iat to fake access token payload * [AUTOMATED]: Prettier Code Styling * Simpler tests * [AUTOMATED]: Prettier Code Styling * Do not clobber iat * catch server error RESET_PASSWORD_EXCEED_LIMIT (#1037) * Merging Master into Multi-Tab (#1038)
[REQUIRED] Step 2: Describe your environment
[REQUIRED] Step 3: Describe the problem
Steps to reproduce:
Set up a query on any subcollection with a limit smaller than the number of elements in the subcollection already, for example, 5.
Set up another query on the same subcollection with a limit set to 1. (Possibly unneeded step.)
Start deleting documents returned by the first query randomly one by one.
Observe that at some points the documents will begin disappearing from the query snapshot on their own. They will remain gone even if the app restarts but will be still visible via Firestore Console.
Relevant Code:
See example app along with more detailed description and video here.
Also, the following code in
Firestore/Source/Remote/FSTRemoteEvent.mminremoteEventAtSnapshotVersion:method seems to have something to do with this:The text was updated successfully, but these errors were encountered: