Feed shows five-days old videos - but the channel itself works fine.

[IngwiePhoenix]

Official Instance

• The bug is reproducible on the official hosted instance, or is API-related.

Describe the bug

I am running the Helm Chart on Version 7.0.0 to which I updated after a friend let me know that the instance had gone dark. Turns out that the PostgresQL password rotated, and something in the auto reconfiguration chain didn't do it's job. Oh well - so I just restarted it after manually updating the ConfigMap

Now however, the videos shown in my feed are five days old (and then some):

Checking the logs on the backend, everything seems perfectly fine:

Jan 03, 2025 11:21:26 PM liquibase.database
INFO: Set default schema name to public
Jan 03, 2025 11:21:27 PM liquibase.changelog
INFO: Reading from public.databasechangelog
Database is up to date, no changesets to execute
Jan 03, 2025 11:21:27 PM liquibase.changelog
INFO: Reading from public.databasechangelog

UPDATE SUMMARY
Run:                          0
Previously run:               3
Filtered out:                 1
-------------------------------
Total change sets:            4


FILTERED CHANGE SETS SUMMARY
DBMS mismatch:                1

Jan 03, 2025 11:21:27 PM liquibase.util
INFO: UPDATE SUMMARY
Jan 03, 2025 11:21:27 PM liquibase.util
INFO: Run:                          0
Jan 03, 2025 11:21:27 PM liquibase.util
INFO: Previously run:               3
Jan 03, 2025 11:21:27 PM liquibase.util
INFO: Filtered out:                 1
Jan 03, 2025 11:21:27 PM liquibase.util
INFO: -------------------------------
Jan 03, 2025 11:21:27 PM liquibase.util
INFO: Total change sets:            4
Jan 03, 2025 11:21:27 PM liquibase.util
INFO: FILTERED CHANGE SETS SUMMARY
Jan 03, 2025 11:21:27 PM liquibase.util
INFO: DBMS mismatch:                1
Jan 03, 2025 11:21:27 PM liquibase.util
INFO: Update summary generated
Jan 03, 2025 11:21:27 PM liquibase.lockservice
INFO: Successfully released change log lock
Jan 03, 2025 11:21:27 PM liquibase.command
INFO: Command execution complete
ThrottlingCache: 0 entries
SLF4J: No SLF4J providers were found.
SLF4J: Defaulting to no-operation (NOP) logger implementation
SLF4J: See https://www.slf4j.org/codes.html#noProviders for further details.
Jan 03, 2025 11:21:27 PM org.hibernate.Version logVersion
INFO: HHH000412: Hibernate ORM core version [WORKING]
Jan 03, 2025 11:21:28 PM org.hibernate.cache.internal.RegionFactoryInitiator initiateService
INFO: HHH000026: Second-level cache disabled
Jan 03, 2025 11:21:28 PM org.hibernate.engine.jdbc.connections.internal.ConnectionProviderInitiator initiateService
INFO: HHH000130: Instantiating explicit connection provider: org.hibernate.hikaricp.internal.HikariCPConnectionProvider
Logged in as user: null
Room ID: !CfXSiQMnWTEPfnBNuK:matrix.org
Filter ID: null
Jan 03, 2025 11:21:28 PM org.hibernate.engine.jdbc.dialect.internal.DialectFactoryImpl constructDialect
WARN: HHH90000025: PostgreSQLDialect does not need to be specified explicitly using 'hibernate.dialect' (remove the property setting and it will be selected by default)
Jan 03, 2025 11:21:29 PM org.hibernate.engine.transaction.jta.platform.internal.JtaPlatformInitiator initiateService
INFO: HHH000489: No JTA platform available (set 'hibernate.transaction.jta.platform' to enable JTA platform integration)
Database connection is ready!
Cleanup: Removed 1 old videos
PubSub: queue size - 0 channels

And watching videos also works. But if I visit one of the channels, say WrestleTalk, I can totally see their recent videos:

I could not find any endpoint to force a channel refresh, and none of the logs were any indication. Hence why I decided to post it here.

To Reproduce

1. Let your instance go offline for 2-4 days.
2. Restart your instance
3. Observe "Feed" versus channel.

Expected behavior

After it starts, I expected the backend to go out and sync the subscribed channels - I only subscribed to like three or four, so querying them shouldn't take too long, I'd assume.

Logs/Errors

None that stick out. The blurbs above are practically everything - aside fron the frontend's access logs, but those are...well, access logs.

Browser, and OS with Version.

Windows 10, amd64, Brave 1.73.97 Chromium: 131.0.6778.108 (Offizieller Build) (64-Bit)

Additional context

This instance is hosted from home, so it has a non-blocked IP and has been working quite fine :)

[Bnyro]

Duplicate of #1130