It seems like at present if something went wrong during npm install, the site would be broken until we reverted/cleaned up the problem (since it tears down the old install before starting the now-broken one and thus it fails in a broken state).
We should make this more resilient / smarter (e.g. work in a copy of the image to ensure it builds and then swap the copy in?)
It seems like at present if something went wrong during npm install, the site would be broken until we reverted/cleaned up the problem (since it tears down the old install before starting the now-broken one and thus it fails in a broken state).
We should make this more resilient / smarter (e.g. work in a copy of the image to ensure it builds and then swap the copy in?)