retry the quick connect startup probe, and exit non-zero when a start fails
This commit is contained in:
@@ -13,6 +13,11 @@ namespace Emby.Server.Implementations.ScheduledTasks;
|
||||
/// TTL lease on a shared key. The lease value is this instance's pod identity; a leader that keeps
|
||||
/// renewing retains the lease, and any instance can claim it once the previous leader's lease expires.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// This fails open on an unreachable Redis while quick connect's startup read fails closed on the same
|
||||
/// connection. They do not compete: the startup read decides whether the instance runs at all, and this
|
||||
/// only decides what a running instance does about a store that went away afterwards.
|
||||
/// </remarks>
|
||||
public sealed class RedisScanLeaderLease : IScanLeaderLease
|
||||
{
|
||||
private const string LeaderKey = "jellyfin:scanleader";
|
||||
|
||||
Reference in New Issue
Block a user