My application has a .net core Microservice for handling notifications and It has been deployed on Kubernetes. In there NotificationRequestConsumer as follows, (please be noted this is just a code snippet to elaborate my question)
public class NotificationRequestConsumer : IConsumer<INotificationRequest>
{
public NotificationRequestConsumer()
{
}
public Task Consume(ConsumeContext<INotificationRequest> context)
{
// notification request logic goes here
return Task.CompletedTask;
}
}
how configured the Masstransit in the startup.
public static IServiceCollection AddMassTransitConnection(this IServiceCollection services, IConfiguration configuration)
{
services.AddMassTransit(x =>
{
x.AddBus(context => Bus.Factory.CreateUsingRabbitMq(c =>
{
c.Host(configuration["RabbitMQ:HostUrl"]);
c.ConfigureEndpoints(context);
}));
x.AddConsumer<NotificationRequestConsumer>(c => c.UseMessageRetry(r => r.Interval(1,500)));
});
services.AddMassTransitHostedService();
return services;
}
As per the above code, I have set interval few milliseconds to retry if any error occurs while alert processing. If there is a problem, I use the fault consumer to store the data of the relevant request in the DB, for future use (to send the relevant notification manually in the future).
public class NotificationRequestFaultConsumer : IConsumer<Fault<INotificationRequest>>
{
public Task Consume(ConsumeContext<Fault<INotificationRequest>> context)
{
//For future use, I store the relevant data here
return Task.CompletedTask;
}
}
Even if I did this, the relevant exception would be added to the RabbitMQ error queue. As far as I know, That is part of how transport works.
My concerns are as follows,
- Does the continuous growth of the error queue cause the cluster to crash?
- Is it a good approach to log only to the ELK Stack without throwing exceptions and not adding them to the RabbitMQ error queue?
- Is it possible to give specific expiration criteria to delete the error queue automatically and is it a good thought?
