I recently realized we could effectively use a spot instance from AWS EC2 at an 80% discount to their on-demand cost and that we have $15k in AWS credits, so I’ve been working on adding a new Jenkins agent node from there. It has 72 cores, a fast NVM SSD, and lots of RAM. Surprisingly I’m not seeing huge speedups on distribution tests (something like 90 minutes vs 110 minutes on a 16 core mac pro), so there is probably some work to be done figuring out what is up there at some point (would love help with that).
You may have seen some jobs fail while I added the node and tried to recreate what the dependencies are for our jobs (they aren’t written down anywhere). Here is the list I’ve come up with so far for a non-GPU Linux box just for records sake:
apt-get install build-essential clang-6.0 clang-format-6.0 latexmk texlive-latex-recommended texlive-latex-extra r-base pandoc libxml2-dev libcurl4-openssl-dev libssl-dev pandoc-citeproc openmpi-bin doxygen python-dev
If you see a weird failure today, please go ahead and restart or ask me to. I think things should be approaching stable again.
I will also take this time to show a new feature recently enabled. From the Blue Ocean view you should now have the option to restart a specific test stage of a test (especially useful for a failed test at a later stage):