A lot of our tests on OS X and Linux timed out on Saturday while waiting for the download of the full mar file. Generally it has taken about 6min to download only 60MB on OS X. Given that we serve all those bits via S3 those days, I wonder if there might be something broken. Nick or Ben, does one of you have an idea?
Given that so many of the update tests were failing I wonder if there were some bandwith problems in SCL3 that day. I will wait for the results of the builds from today before digging deeper into this problem.
Created attachment 8663541 [details] smokeping graph (UTC timestamps) The quality of connection from scl3 to us-east-1 was poor on Saturday. The packetloss would have reduced transfer rates significantly. The situation in AWS has been resolved by them now.
(In reply to Nick Thomas [:nthomas] from comment #16) > The quality of connection from scl3 to us-east-1 was poor on Saturday. The > packetloss would have reduced transfer rates significantly. The situation in > AWS has been resolved by them now. Thanks Nick! This would totally explain the massive failures of our update tests then. Given that the situation was fine yesterday I would assume that we went back stable with the tests. I will reopen the bug if it turns out otherwise.