We noticed about a week ago that Docker image pushes to Quay.io have gotten a lot slower. Pushes to Quay are still quick from our own EC2 machines. Did something change on Circle’s side?
What kind of data transfer speed are you getting now? It would also be great to know what was the bandwidth you were seeing previously.
docker push doesn’t report bandwidth, so it’s hard to tell! We didn’t have any formal benchmarks setup before this issue, and haven’t made any since. From builds before and after the slowdown, I can see that
docker push used to take ~1m and now takes ~5m. In both cases we’re only pushing ~15MB to Quay.
I am afraid this might be a question for Quay.io support. We haven’t observed any network issues recently, and so we don’t think this is directly related to any of our actions. Thank you for your understanding.
Late answer but I noticed in debugging CircleCI is using Docker v1.8. In Docker 1.9 there is a fix for the buffering to disk that greatly speeds up pushes. I filed a ticket pointing this out and hoping they default upgrade.
In the mean time I’m trying to figure out how to script a forced upgrade to docker 1.9