fix: make transcode leases ownership-checked and cleanup-aware

Cleanup never matched a live session because the controller registered empty
manifest and segment paths, renewal was a read-modify-write that could revert a
takeover, and the takeover script compared an ISO date to a number, so it errored.

- populate the session record's manifest and segment paths from the playlist path
- renew the lease via a Lua compare-and-set on the owning pod
- store the lease expiry as unix milliseconds so the scripts can compare it
- retain the session record past its lease so an orphan can still be taken over
- test the Redis store against a real Redis, including the renew-vs-takeover race
- drop the live stream record nothing ever read back
This commit is contained in:
2026-09-12 10:19:29 +10:00
parent d825f8ac81
commit baa16b6586
22 changed files with 644 additions and 1221 deletions
@@ -0,0 +1,186 @@
using System;
using System.Threading.Tasks;
using Jellyfin.MediaEncoding.Tests.Fakes;
using MediaBrowser.Controller.MediaEncoding;
using Xunit;
namespace Jellyfin.MediaEncoding.Tests.Transcoding;
/// <summary>
/// Unit tests for the <see cref="ITranscodeSessionStore"/> contract — lease expiry, double-claim
/// prevention, ownership-checked renewal and stale-session cleanup — against
/// <see cref="InMemoryTranscodeSessionStore"/>, the reference implementation. The Redis-backed
/// implementation is covered by <c>RedisTranscodeSessionStoreTests</c> against a real Redis.
/// </summary>
public class InMemoryTranscodeSessionStoreTests
{
private static TranscodeSession CreateSession(
string id,
string pod,
DateTime leaseExpiry,
int lastSegmentIndex = 0,
long lastOffset = 0L)
=> new TranscodeSession
{
PlaySessionId = id,
OwnerPod = pod,
LeaseExpiresUtc = leaseExpiry,
ManifestPath = $"/transcode/{id}/manifest.m3u8",
SegmentPathPrefix = $"/transcode/{id}/segment",
MediaSourceId = $"media-source-{id}",
LastCompletedSegmentIndex = lastSegmentIndex,
LastDurablePlaybackOffset = lastOffset,
};
/// <summary>
/// Lease expiry: <see cref="ITranscodeSessionStore.TryGetAsync"/> returns <c>null</c>
/// once <see cref="TranscodeSession.LeaseExpiresUtc"/> has passed.
/// </summary>
[Fact]
[Trait("Category", "UnitTest")]
public async Task TryGetAsync_AfterLeaseExpires_ReturnsNull()
{
var store = new InMemoryTranscodeSessionStore();
var session = CreateSession("session-expired", "pod-a", DateTime.UtcNow.AddMilliseconds(-1));
await store.SetAsync(session, TestContext.Current.CancellationToken);
var result = await store.TryGetAsync("session-expired", TestContext.Current.CancellationToken);
Assert.Null(result);
}
/// <summary>
/// A session whose lease has not yet expired is returned correctly.
/// </summary>
[Fact]
[Trait("Category", "UnitTest")]
public async Task TryGetAsync_WithinLease_ReturnsSession()
{
var store = new InMemoryTranscodeSessionStore();
var session = CreateSession("session-live", "pod-a", DateTime.UtcNow.AddMinutes(5), lastSegmentIndex: 3, lastOffset: 18_000_000L);
await store.SetAsync(session, TestContext.Current.CancellationToken);
var result = await store.TryGetAsync("session-live", TestContext.Current.CancellationToken);
Assert.NotNull(result);
Assert.Equal("session-live", result.PlaySessionId);
Assert.Equal("pod-a", result.OwnerPod);
Assert.Equal(3, result.LastCompletedSegmentIndex);
Assert.Equal(18_000_000L, result.LastDurablePlaybackOffset);
}
/// <summary>
/// Double-claim prevention: <see cref="ITranscodeSessionStore.TryTakeoverAsync"/> returns
/// <c>false</c> while the first pod's lease is still valid.
/// </summary>
[Fact]
[Trait("Category", "UnitTest")]
public async Task TryTakeoverAsync_WhileLeaseValid_ReturnsFalse()
{
var store = new InMemoryTranscodeSessionStore();
var session = CreateSession("session-valid", "pod-a", DateTime.UtcNow.AddMinutes(5));
await store.SetAsync(session, TestContext.Current.CancellationToken);
var firstAttempt = await store.TryTakeoverAsync("session-valid", "pod-b", TestContext.Current.CancellationToken);
var secondAttempt = await store.TryTakeoverAsync("session-valid", "pod-c", TestContext.Current.CancellationToken);
Assert.False(firstAttempt);
Assert.False(secondAttempt);
}
/// <summary>
/// After a lease expires, the first concurrent caller that invokes
/// <see cref="ITranscodeSessionStore.TryTakeoverAsync"/> wins; the second caller returns <c>false</c>.
/// </summary>
[Fact]
[Trait("Category", "UnitTest")]
public async Task TryTakeoverAsync_AfterLeaseExpires_OnlyFirstClaimerSucceeds()
{
var store = new InMemoryTranscodeSessionStore();
var session = CreateSession("session-stale", "pod-a", DateTime.UtcNow.AddMilliseconds(-1));
await store.SetAsync(session, TestContext.Current.CancellationToken);
// First pod wins; its takeover renews the lease atomically.
var firstTakeover = await store.TryTakeoverAsync("session-stale", "pod-b", TestContext.Current.CancellationToken);
// Second pod is too late pod-b already holds a fresh lease.
var secondTakeover = await store.TryTakeoverAsync("session-stale", "pod-c", TestContext.Current.CancellationToken);
Assert.True(firstTakeover);
Assert.False(secondTakeover);
}
/// <summary>
/// Heartbeat renewal: <see cref="ITranscodeSessionStore.RenewLeaseAsync"/> extends
/// <see cref="TranscodeSession.LeaseExpiresUtc"/> beyond its original value.
/// </summary>
[Fact]
[Trait("Category", "UnitTest")]
public async Task RenewLeaseAsync_ExtendsLeaseExpiry()
{
var store = new InMemoryTranscodeSessionStore();
var originalExpiry = DateTime.UtcNow.AddSeconds(5);
var session = CreateSession("session-renew", "pod-a", originalExpiry, lastSegmentIndex: 2, lastOffset: 10_000_000L);
await store.SetAsync(session, TestContext.Current.CancellationToken);
Assert.True(await store.RenewLeaseAsync("session-renew", "pod-a", TestContext.Current.CancellationToken));
var renewed = await store.TryGetAsync("session-renew", TestContext.Current.CancellationToken);
Assert.NotNull(renewed);
Assert.True(
renewed.LeaseExpiresUtc > originalExpiry,
"Renewed lease expiry should be later than the original expiry.");
}
/// <summary>
/// A renewal from a pod that no longer owns the lease must fail and must not revert ownership.
/// </summary>
[Fact]
[Trait("Category", "UnitTest")]
public async Task RenewLeaseAsync_ByNonOwner_ReturnsFalse()
{
var store = new InMemoryTranscodeSessionStore();
var session = CreateSession("session-renew-other", "pod-a", DateTime.UtcNow.AddMilliseconds(-1));
await store.SetAsync(session, TestContext.Current.CancellationToken);
Assert.True(await store.TryTakeoverAsync("session-renew-other", "pod-b", TestContext.Current.CancellationToken));
Assert.False(await store.RenewLeaseAsync("session-renew-other", "pod-a", TestContext.Current.CancellationToken));
var current = await store.TryGetAsync("session-renew-other", TestContext.Current.CancellationToken);
Assert.NotNull(current);
Assert.Equal("pod-b", current.OwnerPod);
}
/// <summary>
/// Stale-session cleanup: an expired session can be deleted without error, and a
/// subsequent <see cref="ITranscodeSessionStore.TryGetAsync"/> returns <c>null</c>.
/// </summary>
[Fact]
[Trait("Category", "UnitTest")]
public async Task DeleteAsync_ExpiredSession_CompletesWithoutError()
{
var store = new InMemoryTranscodeSessionStore();
var session = CreateSession("session-delete", "pod-a", DateTime.UtcNow.AddMilliseconds(-1));
await store.SetAsync(session, TestContext.Current.CancellationToken);
var ex = await Record.ExceptionAsync(() => store.DeleteAsync("session-delete", TestContext.Current.CancellationToken));
Assert.Null(ex);
var result = await store.TryGetAsync("session-delete", TestContext.Current.CancellationToken);
Assert.Null(result);
}
/// <summary>
/// Deleting a session that was never stored must complete without error.
/// </summary>
[Fact]
[Trait("Category", "UnitTest")]
public async Task DeleteAsync_NonExistentSession_CompletesWithoutError()
{
var store = new InMemoryTranscodeSessionStore();
var ex = await Record.ExceptionAsync(() => store.DeleteAsync("nonexistent-session", TestContext.Current.CancellationToken));
Assert.Null(ex);
}
}