I know this is probably not the best place for an issue like this, but I have been unable to get any response from email@example.com, firstname.lastname@example.org, or email@example.com despite repeated emails asking for help with this issue.
I work for a nation-wide SD-WAN network provider. We have been seeing numerous issues for customers in the Northwest connecting to sites using Fastly’s services. Specifically, we are seeing an increasing problem TCP resets causing connection failures to sites like stackoverflow.com, cnn.com, rubygem.org, pypi.python.org, github.com, and others. It seems that making the initial connection to the sites often fails with a TCP reset, and then the page doesn’t load.
We’ve been testing doing something like
curl -I cnn.com, which we see fail roughly 50% of the time with an error like “curl: (56) Recv failure: Connection reset by peer”. We see cnn.com resolving to 126.96.36.199, and when we do a packet capture of these requests we see a TCP reset as one of the first packets back from 188.8.131.52, which is what kills the connection.
We’ve done packet captures all the way to the edge of our network, and can confirm those TCP resets are already present in the traffic stream when they hit our network, so it is not some device within our network sending the TCP resets. It really does look like a problem on Fastly’s end, and we need their help to resolve it.
I wouldn’t normally post an issue like this publicly, instead preferring to work through the problems with the provider directly. However, this problem is causing impact for our customers, and we need to resolve it as soon as possible. Therefore, I’m posting here in case Fastly might monitor this forum better than their emails, or in case someone can provide a better means of contacting an engineer at Fastly that can help with this matter.
Please sign in to leave a comment.