On 11/11/15 14:40, David Moreau Simard wrote:

Hi,

I've heard that network connectivity for accessing resources outside of the ci.centos.org infrastructure can sometimes be unreliable. I haven't yet experienced this first hand as the jobs I'm constantly monitoring and am responsible for have not yet been migrated there.

A lot of things can be put within the ci.centos.org infrastructure if need be but it's unrealistic to have *every* resource there.

Is anyone aware of ongoing issues ? Should we report jobs that fail due to network errors ?

We share a gbit uplink between cbs.centos.org and ci.centos.org, and there are a few other services that would see occasional traffic spurts from the same DC. But by and large the most regular traffic would be from cbs.c.o and ci.c.o; Once traffic leaves out infra, we have no control on the network side of things - so if you were to let us know whenever there is an issue, it would allow us to go back to the upstream folks and work it that way. Ideally, let us know what time/date you had the issue and details like end point etc where you had this problem to/from.

A recent example is this job that failed due to DNS resolution errors [1] "Could not resolve host: trunk.rdoproject.org; Name or service not known".

Was this via the default nameservers setup in /etc/resolv.conf on the machine ?

Thanks,