On 28/01/16 16:59, Karanbir Singh wrote:
On 28/01/16 15:39, John Trowbridge wrote:
On 01/28/2016 10:25 AM, Karanbir Singh wrote:
On 28/01/16 11:54, John Trowbridge wrote:
That would be awesome. I tried putting one of them on a personal RAX account, but it was not really much of an improvement. I have currently updated the documentation for tripleo-quickstart to include instructions to pre-download the images. This has been a viable workaround for the near term, but it is still probably not ideal to be pointing end users at artifacts.ci.centos.
as a data point - are you able to grab some images currently in place on buldlogs.centos.org hosts and see if its any better speeds ? try these urls http://buildlogs.centos.org/centos/7/atomic-host/x86_64/images/CentOS
-Atomic-Host-7.20160125-GenericCloud.qcow2.gz
After some further troubleshooting with folks, I am not so sure it is a
speed issue so much as nginx dropping the connection. My workaround to pre-download the images uses wget for the built-in retry w/ resume, and I have not had any reported issues using that process.
Ansible uses curl, which is not handling the dropped http connection well, resulting in a broken download.
I tried the above link and it works great for me, but so does the artifacts server. Is the buildlogs server also going through nginx?
the buildlogs services are geo distributed across the US and EU. and run apache httpd natively. We are working with a CDN provider to get upto 40 pop's across asia, europe, america's - but thats a few weeks ( best case) out yet.
just had a very quick chat with John on IRC - the summary of which is that the wget process he's using ( with retry ), seems to work around the problem they had. So we can park this issue for the time being.
Ci-users mailing list Ci-users@centos.org https://lists.centos.org/mailman/listinfo/ci-users