One thing that could be done is to keep a hash or version number on the server of the current combination of repos that are responsible for puppet runs. Whenever one of them changes, the version changes too. The clients can check in with the server to see if they've already run the version on the server. If not, just don't do a run. That would likely cut back on the vast majority of all runs.
On Sat, Nov 1, 2014 at 11:01 AM, Marc A. Pelletier <[email protected]> wrote: > On 10/31/2014 01:04 PM, Andrew Bogott wrote: > > delays in the client run shouldn't really make a difference to anyone > > who isn't actively debugging puppet and running things over and over by > hand > > No, but a quick look at every graph of every instance shows significant > spikes in resource usage every 20 minutes on the dot; it may not be > currently harmful or problematic, but it's certainly a waste. > > -- Marc > > > _______________________________________________ > Labs-l mailing list > [email protected] > https://lists.wikimedia.org/mailman/listinfo/labs-l >
_______________________________________________ Labs-l mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/labs-l
