I forgot to mention about clean up of durable function orchestration history in my previous post, so I thought i’ll include that here.
As we saw in the last post, the entire Orchestration history is maintained in our TestHubNameHistory table instance and over time these records will keep growing, so a good practice would be to clean those up on periodic basis.
The DurableClient provides a helpful method for that, called PurgeInstanceHistoryAsync. This method provides couple overloads and as you will see, I have used one of them in a TimerTrigger function for performing cleanup in the SyncOverAsyncApi.cs class, shown below:
[FunctionName(nameof(CleanUpOldWeatherResponsesAsync))]
public async Task CleanUpOldWeatherResponsesAsync([TimerTrigger("0 0 0 * * *")] /* execute every day */ TimerInfo myTimer, [DurableClient] IDurableOrchestrationClient client)
{
var result = await client.PurgeInstanceHistoryAsync(DateTime.UtcNow.AddDays(-15), DateTime.UtcNow.AddDays(-1), new List<OrchestrationStatus> { OrchestrationStatus.Completed, OrchestrationStatus.Terminated });
this.Logger.LogInformation("Cleaned up records: ", result?.InstancesDeleted);
}
The above method CleanUpOldWeatherResponsesAsync is straightforward, it executes daily and accepts DurableClient as one of the parameter provided by the runtime. The overload of PurgeInstanceHistoryAsync, takes the date range and the orchestration statuses to be purged.
NOTE: If you test this locally, you might run into BadRequest issue depending on your setup.
Finally, there are some performance considerations to take into account when using these APIs, if you need an alternative / faster way to clean up your taskhub state, refer to this article by Chris Gillum.