Benjamin Petit

> Is this right? The difference here, is if stream is active, it starts at newest = D instead of B. Correct > It sounds like we can increase MetadataMinTimeInCache...

That means 1 stream per device. If you didn't change `CollectionAge` you might get unwanted cache exception. But you could choose to ignore them if you see that we didn't...

Sorry for the delay. Multiple things here: > . Since Azure SDK API [ReceiveMessagesAsync](https://github.com/dotnet/orleans/blob/5ec98fb485ea89be192947878216a546f02e35d4/src/Azure/Orleans.Streaming.AzureStorage/Storage/AzureQueueDataManager.cs#L226) called inside GetQueueMessage has a retry mechanism when transient network errors (e.g., timeout) happen, a retry...

That's expected. Basically, B and C think that A is still alive, so at startup, they try to open a connection to A. They will try for `ClusterMembershipOptions.MaxJoinAttemptTime` (by default...

`ClusterMembershipOptions.ValidateInitialConnectivity = false;` Please, don't do that. Your silos might be able to start, but the cluster will not be working correctly. I think we should remove this setting in...

I both cases, it seems that the cluster is not yet "stable" and there is some discrepancies in the distribution of the grain directory. Can you set a 30 seconds...

I have a prototype of a working `IAsyncEnumerable`, but it is backed by persistent streams. > Sending individual messages is too slow given the latency between grains Is it the...

> > I have a prototype of a working `IAsyncEnumerable`, but it is backed by persistent streams. > > Is there a chance that this prototype will become production ready...

When a consumer receives an event with `OnNextAsync`, it also gets a `StreamSequenceToken`. This token allows to resume a subscription to a given stream position. Events stay in the cache...

That's correct, you may lose messages in case of failure. I assumed that in your case, the publisher would enqueue a special message to denote the last event of a...