[CentOS] Problem with cloud-init on re-boot

Mon Jul 18 15:05:59 UTC 2016
Warren Young <wyml at etr-usa.com>

On Jul 16, 2016, at 4:12 AM, Timothy Murphy <gayleard at eircom.net> wrote:
> When I re-boot my CentOS 7 machine I get lots of error/warning messages
> like the following:
> ------------------------------
> Jan  1 14:39:04 alfred cloud-init: 2016-01-01 14:39:04,351 - 
> url_helper.py[WARNING]:
>  Calling '' failed 
> [2/120s]:
>  unexpected error ['NoneType' object has no attribute 'status_code']
> ------------------------------

That appears to be an OpenStack problem, so you may have better luck talking to those folks.  Here is a thread on their mailing list showing a similar problem:


> Re-booting with the current kernel, 3.10.0-327.22.2.el7.x86_64,
> sometimes fails, I think because of this problem,
> although the previous kernel seems to avoid the issue.

I don’t see how this could be a kernel-specific issue.

As a rule, kernel problems are driver problems, so unless you’ve got some kind of strange initialization problem with your network IC driver or similar, I don’t see how it can affect things.

That other mailing list thread suggests a useful test: if OpenStack fails to init on boot, can you always bring it up by hand after the system is fully booted?  If so, then that suggests you’ve got OpenStack loading too early in the boot process.

A pre-systemd init setup converted to systemd without thorough testing could give such problems by not declaring its dependencies explicitly enough, so that it only *happens* to work depending on how fast the rest of the system boots.

> I don't remember installing cloud-init;

If you don’t use it, remove it.  I know next to nothing about OpenStack, but I do know it’s pretty heavy infrastructure.  No point paying for it if you aren’t using it.

> maybe it was brought in by some other package?

You don’t have to guess.  RPM will tell you.

> is it a necessary adjunct to libvirtd?

More likely the other way around: OpenStack is probably built on top of libvirtd.