On 2015-05-29 21:01:23 +0000 (+0000), Dmitry Borodaenko wrote: > I think is unnecessarily maximalist. Trust is not an > all-or-nothing boolean flag: why can't you trust that server to do > more work at the same level of trust and run another batch of > user-submitted code?
Because it turns out to be a lot easier not to have to figure out what's safe for node reuse and what isn't. We used to draw the line between jobs which don't have root access and jobs which do, but that was not actually a completely sane distinction for a number of reasons and hampered flexibility of our job definitions as well. Also, if you turn over your workers fairly rapidly, you don't need to worry about care and feeding, package updates, et cetera. Back when we had long-running workers for things like unit test jobs, a substantial chunk of my time was spent hunting down problems with them becoming disconnected from their Jenkins masters, developing subtle issues which would cause them to start rapidly failing any job run on them, and so on. Our least available resource is people, so saving them from needing to make decisions is one of the best ways we can increase the performance of our community (even if it means reducing performance of the systems being used to support them). -- Jeremy Stanley __________________________________________________________________________ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev