Martin Sebor wrote: > Farid Zaripov wrote: [snip] >> But anyway maybe it makes sense to schedule the builds on the same >> machine to not run at the same time?
The scheduling system doesn't allow for the concept of mutually exclusive requests/runs, meaning only one build of X can be running at a given time. > > Most of our machines are set up to run N or N + 1 builds in parallel > (where N is the number of CPUs). We might be able to throttle one > down now but eventually it'll get cranked up again otherwise we > wouldn't get the kind of throughput we need. Se we might as well > figure out what's causing it now. If you can't reproduce it maybe > it's an installation problem again. > > Btw., another interesting thing to note is that the links to the > same build logs on the result page are unavailable: > http://people.apache.org/~sebor/stdcxx/results/ > > Andrew, any idea if that's relevant and what it might mean? Looking at the date of the runs (nearly a month old), I think what happened is the build log attachment aged out of the backend database, so the exporter script didn't generate a link to them - even if they might exist on disk at Apache. Given the backlog of requests on the operating system, I suspect the workers may have glitched, so I'll be taking a look at them. > > Martin
