Open Stack

On Fri, May 4, 2018, at 2:37 AM, Sean McGinnis wrote:
> On Fri, May 04, 2018 at 04:13:41PM +0800, linghucongsong wrote:
> > 
> > Hi all!
> > 
> > Recently we meet a strange problem in our ci. look this link: https://review.openstack.org/#/c/532097/
> > 
> > we can pass the ci in the first time, but when we begin to start the gate job, it will always failed in the second time.
> > 
> > we have rebased several times, it alway pass the ci in the first time and failed in the second time.
> > 
> > This have not happen before  and make me to guess is it really we start the ci from the new fresh vms each time?
> 
> A new VM is spun up for each test run, so I don't believe this is an issue with
> stale artifacts on the host. I would guess this is more likely some sort of
> race condition, and you just happen to be hitting it 50% of the time.

Additionally you can check the job logs to see while these two jobs did run against the same cloud provider they did so in different regions on hosts with completely different IP addresses. The inventory files [0][1] are where I would start if you suspect oddness of this sort. Reading them I don't see anything to indicate the nodes were reused.

[0] http://logs.openstack.org/97/532097/16/check/legacy-tricircle-dsvm-multiregion/c9b3d29/zuul-info/inventory.yaml
[1] http://logs.openstack.org/97/532097/16/gate/legacy-tricircle-dsvm-multiregion/ad547d5/zuul-info/inventory.yaml

Clark