Hello,
Could someone explain why it was decided to use the auto_delete=true feature on the “results” queue?
If this is enabled, when all the sensu-servers are down all client data is lost and not buffered…
Regards,
Sebastian
Ok I have found some evidence as why it was activated back in 2013 (auto delete the result & keepalive queues to prevent the stampede by portertech · Pull Request #614 · sensu/sensu · GitHub)
I think we should find a better compromise and let the user decide to activate this option or not, many people are now using Sensu as a metric collection / transport solution and losing data is not an option.
We now also have the RabbitMQ TTL (Time-To-Live and Expiration — RabbitMQ) functionality that could be managed through the policy definition…
I will open a PR for making this functionality user configurable.
Best regards,
Sebastian
···
On Thursday, July 7, 2016 at 5:08:56 PM UTC+2, Sebastian YEPES wrote:
Hello,
Could someone explain why it was decided to use the auto_delete=true feature on the “results” queue?
If this is enabled, when all the sensu-servers are down all client data is lost and not buffered…
Regards,
Sebastian