Background jobs

Jobs run on Hangfire with the storage of the database provider: Postgres, SQL Server, or in memory with SQLite. A job runs as the user who enqueued it, so permissions, tenant filters and audit work as in a request.

public sealed record RenderPreview(Guid AssetId);

internal sealed class RenderPreviewJob(IDispatcher dispatcher) : IBackgroundJob<RenderPreview>
{
    public Task ExecuteAsync(RenderPreview args, CancellationToken cancellationToken) =>
        dispatcher.SendAsync(new CreatePreviewCommand(args.AssetId), cancellationToken);
}

jobs.Enqueue<RenderPreviewJob, RenderPreview>(new RenderPreview(asset.Id), queue: "processing");
jobs.Schedule<RenderPreviewJob, RenderPreview>(new RenderPreview(asset.Id), TimeSpan.FromMinutes(5));

Recurring jobs run outside of a user and tenant:

internal sealed class CleanupJob : IRecurringJob
{
    public Task ExecuteAsync(CancellationToken cancellationToken) => Task.CompletedTask;
}

services.AddRecurringJob<CleanupJob>("cleanup", "0 3 * * *");

Work in every organisation goes through TenantJobRunner: once per active organisation, as the system of that organisation (tenant filters, audit) and in a DI scope of its own. A failing organisation is logged and returned, the others still run:

internal sealed class ReminderJob(TenantJobRunner tenants) : IRecurringJob
{
    public async Task ExecuteAsync(CancellationToken cancellationToken) =>
        await tenants.ForEachTenantAsync((services, tenantId, ct) =>
            services.GetRequiredService<IDispatcher>().SendAsync(new SendReminders(), ct), cancellationToken);
}

ForTenantsAsync(tenantIds, ...) does the same for the organisations a query found (due rows grouped by TenantId), so organisations without work cost nothing. ForEachTenantAsync lists the organisations through the Identity module. The run is sequential; with many organisations keep the work per organisation short.

Coworkee:Jobs
RunServer	process jobs in this host (the app host turns it off for the auth server)
Queues	queues this host works on, default default, mail
WorkerCount	parallel workers (5)
Attempts, RetryDelaysInSeconds	retries

The Hangfire dashboard is under /admin/jobs for users with the jobs permission; the admin page Services shows it as a tile. Never block a job waiting for another one; schedule a follow-up instead.