CSHARP-3757: Redirect read/write retries to other mongos if possible #1304
Conversation
|
|
||
| var deprioritizedServers = new List<ServerDescription> { connected1 }; | ||
|
|
||
| for (int i = 0; i < 15; i++) |
There was a problem hiding this comment.
Running multiple times to account for server selection randomness and guarantee we get the expected result for each run.
| /// <param name="deprioritizedServers">The deprioritized servers.</param> | ||
| /// <param name="cancellationToken">The cancellation token.</param> | ||
| /// <returns>A channel source.</returns> | ||
| IChannelSourceHandle GetReadChannelSource(IReadOnlyCollection<ServerDescription> deprioritizedServers, CancellationToken cancellationToken); |
There was a problem hiding this comment.
This interface is pseudo public (going to be hidden in 3.0), and we are introducing a breaking change anyway.
So should we consider just adding a parameter to the existing method?
src/MongoDB.Driver.Core/Core/Operations/RetryableReadOperationExecutor.cs
Outdated
Show resolved
Hide resolved
|
|
||
| subject.Initialize(); | ||
|
|
||
| var endPoint1 = new DnsEndPoint("localhost", 27017); |
There was a problem hiding this comment.
minor: add simple test to validate deprioritizedServers == null case
| IServer result; | ||
| if (async) | ||
| { | ||
| result = subject.SelectServerAsync(selector, deprioritizedServers, CancellationToken.None).GetAwaiter().GetResult(); |
There was a problem hiding this comment.
Make the test return Task, and just await here.
| @@ -77,14 +76,14 @@ public void Dispose() | |||
| } | |||
|
|
|||
| /// <inheritdoc/> | |||
| public IChannelSourceHandle GetReadChannelSource(CancellationToken cancellationToken) | |||
| public IChannelSourceHandle GetReadChannelSource(CancellationToken cancellationToken, IReadOnlyCollection<ServerDescription> deprioritizedServers = null) | |||
There was a problem hiding this comment.
There is a convention that cancellationToken has to be a last parameter.
| List<ServerDescription> filteredServers = new(); | ||
| foreach (var description in _description.Servers) | ||
| { | ||
| if (!deprioritizedServers.Contains(description, new ServerDescriptionComparer())) |
There was a problem hiding this comment.
micro-optimization: could use single instance of comparer.
Or maybe the following would be slightly simpler:
if (deprioritizedServers.All(d => d.EndPoint != description.Endpoint))
| /// <returns>The selected server.</returns> | ||
| IServer SelectServer(IServerSelector selector, CancellationToken cancellationToken); | ||
| IServer SelectServer(IServerSelector selector, CancellationToken cancellationToken, IReadOnlyCollection<ServerDescription> deprioritizedServers = null); |
There was a problem hiding this comment.
Consider having augmenting IServerSelector to account for deprioritizedServers.
Description
On a sharded cluster, retrying reads and writes will attempt to prioritize other mongos for selection if available.
What is changing?