We run a small number of t1-micro instances on Amazon Web Services, because the tasks have consistent very low CPU requirements (1-2% average). We monitor these instances via Amazon CloudWatch with alerting on things like extended 100% CPU usage.
For those who don't know, t1.micro instances have have very limited CPU power but have the ability to peak for a short time, after which the CPU will be throttled to maintain a certain average. The concept is explained in http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/concepts_micro_instances.html The instances are managed via Puppet Agent version 3.4.2, talking to our Puppet master, and have 267 resources applied to them. The Puppet agent is the only process on the server that manages to use enough CPU for long enough to trip the throttling threshold, affecting every process on the instance. This in turn triggers CloudWatch after a little while. I would like to reduce the Agent's impact on the host, while keeping the host managed, perhaps by limiting the CPU usage of the process so we avoid triggering the throttle. A Puppet run taking longer would be fine in this case. I've looked at tools like cpulimit and cgroups, but am unsure how to apply those via Puppet. How can I avoid this throttling and the resulting alerts? Reducing the number of Puppet resources further, to reduce the load, appears difficult and would likely not be enough. Any other ideas? Regards, Martijn -- You received this message because you are subscribed to the Google Groups "Puppet Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to puppet-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/puppet-users/ab28b8fa-fd7d-4cc7-9900-406b200225bc%40googlegroups.com. For more options, visit https://groups.google.com/groups/opt_out.
