On 15-03-27 05:26 PM, Anthony Di Franco wrote: > Are we hitting some policy limit that can be increased, or is it > load-dependent?
You are. It's possible to increase the limit but that doesn't scale well (as it requires adding exceptions in the rules, which is not well exposed and difficult to manage). A solution could be to combine some of the workload into fewer, larger jobs or to explicitly serialize them. > We have been considering moving to a dedicated instance > for various reasons; is this a sign that it's a good time to do so? That's one of the "normal" paths along which the more mature tools tend to progress - especially when they have unusual requirements or very large workloads. The upside is that you get to manage your resources according to your needs, of course. The downsides are that doing this in a way that is maintainable and reliable for the long term requires more work and that system administration of the instances becomes part of the maintainers' workload. In general, this is accomplished by writing a puppet manifest that properly sets up your software and its configuration. There are many people who can help you get started with this - including me and the other Labs admins. You can keep your tool running in Tool Labs during the period when you work on this, however, so you can take your time. -- Marc _______________________________________________ Labs-l mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/labs-l
