Transcode state lives only in the process that started ffmpeg, so a pod restart
drops every in-flight HLS stream with no way for a peer to pick it up.
- add ITranscodeSessionStore plus the TranscodeSession and LiveStreamSession records
- add RedisTranscodeSessionStore with TTL leases and an atomic Lua takeover script
- add NullTranscodeSessionStore for single-instance deployments
- pick the store from Jellyfin:TranscodeStore:RedisConnectionString at startup
Multi-replica deployments cannot share a SQLite file, so the database has to
move to a server engine before the rest of the HA work is usable.
- add Jellyfin.Database.Providers.PostgreSQL with an EF Core Npgsql provider
- register the provider and a pooled NpgsqlDataSource when DatabaseType is Jellyfin-PostgreSQL
- accept postgresql:// URIs and POSTGRES_CONNECTION_STRING alongside CustomProviderOptions
- add container-backed provider, CRUD, concurrency and migration tests
DateTime.TryParse without an IFormatProvider falls back to the current
thread culture, so the same string can parse differently (or fail)
depending on the server's locale. None of these call sites deal with
user-entered text - they parse dates that come from filenames, an
HTTP header, ffprobe metadata and values the app itself wrote to the
auth database - so InvariantCulture is the correct provider everywhere
here.
Fixes the S6580 / CA1305 warnings on these call sites.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When the server finishes starting, show "Jellyfin started successfully" with a
5-second "Redirecting in N…" countdown and a Cancel button instead of reloading
immediately. Cancel stops the countdown and the background refresh so the
startup output can be reviewed, and offers a "Continue to Jellyfin" button to
reload manually. The buttons use the web client's emby-button styling.
Also drop the transitional "Applying migrations" activity: it only showed
briefly while the pending migration set was read, or for the whole step when
nothing was pending, so startup now goes from "Preparing migrations" straight
into "Running migration X of Y".
Order the startup log oldest-to-newest inside a height-bounded panel that
scrolls internally and never extends past the bottom of the window. Refresh it
with a background fetch that swaps the log list in place instead of reloading
the whole page, preserving the user's scroll position and only following to the
bottom when they are already there. A full page reload now happens only on the
final transition to the running server or to the error state.
Restyle the startup/migration holding page to match the Jellyfin dark theme,
with the inline wordmark logo, a gradient spinner and a recolored startup log
tree, and move the Morestachio template rendering into a reusable
StartupUiRenderer.
Add a curated, non-identifying "current activity" line to the always-visible
header (for example "Initializing server" or "Running migration X of Y"),
reported from the startup flow and the migration service so it never leaks
server details to unauthenticated clients. Move the log download into a
"Download logs" link in the log panel header, and show only the header, with
no log hints, to non-local clients.
After resolving duplicates the migration deleted all items in one silent
pass (per-id GetItemById plus a single DeleteItemsUnsafeFast), which looks
hung for minutes on large libraries. Delete in batches of 500 and log
progress per batch, which also avoids one oversized delete transaction.
unkin-agent
Cody Robibero