Sorry for the trouble. Performance issue discussed in that ticket was mitigated. I think that your issues are related to the recent image update. I see that you are using Previous Visual Studio 2017 image, probably also because of performance issues happened after this update. When image update happens, current image becomes "previous" so you are right now using the image you probably tried to avoid. Latest Visual Studio 2017 update contains a lot of fixes so, please switch to Visual Studio 2017 and let us know how it goes.
Please let us know how it goes. And ping us as soon as you see something wrong, no need to reach the point when it becomes a big problem.
Regarding cache, we do not actually suggest it as a way to speed up the build (only in rare special cases when dependencies need to be compiled and then cached). It is mostly used as a safety net in case nuget/npm/chocolatey are not available.
Me again. I checked your builds and see that though you used Visual Studio 2017 image couple of times, it was still 13-14 minutes. Can you please try to open some old 9-minute build in the build history and run RE-BUILD COMMIT?
Thanks for looking into this. There is a bunch of churn happening in our
dependencies that is also causing some increase in time. But I just did as
you suggested, and reran an old "fast" job. Here's the before and after:
I also switched that branch to use the latest image (i.e. from Previous Visual Studio 2017 -> Visual Studio 2017).
So, assuming that a single image update has occurred between today and 15 days ago, and we call the current image N, here are the results:
Executed Today, N: 1 failure at 9m44s (which I've seen before when AppVeyor is overloaded/slow, Azure storage emulator fails to start, if it had succeeded there are still several minutes to run after that), 1 build complete at 13m58s!
Not a good trend. N seems even worse than N-1 was. N-2 is no longer available as far as I know, so there's no way to try that, to control for AppVeyor system load changes between now and 15 days ago.
Let us try to isolate environment issues by running your builds on Google cloud VMs. For that please set APPVEYOR_BUILD_WORKER_CLOUD to GCE. Note that build start will be longer -- about 3 minutes (time needed to provision GCE VM). Let us know what you observe.