[Starlingx-discuss] [Containers] LP 1812519: config_controller --kubernetes fails at step 06

Qi, Mingyuan mingyuan.qi at intel.com
Thu Jan 24 06:28:35 UTC 2019


Erich,

They are separate issues.
1. For system application-apply, I've finished the apply and openstack are running good. No need to set proxy to the system env. All images are pulled by docker through docker proxy setting. So just watch out for the golang net lib limitation I've mentioned.

2. For the system proxy config, centos supports CIDR notation in no_proxy. But which step do you need to set system proxy config?
I do encountered an issue during sysinv interaction because httplib2(which get system env proxy settings) used by cgts-client doesn't support CIDR notation :( , which means you still have to set a specific ip address in system no_proxy.
https://github.com/httplib2/httplib2/blob/master/python2/httplib2/__init__.py#L1046-L1059
Fortunately, it doesn't block config_controller and application-apply.

Mingyuan


-----Original Message-----
From: Cordoba Malibran, Erich 
Sent: Thursday, January 24, 2019 1:30
To: Frank.Miller at windriver.com; Qi, Mingyuan <mingyuan.qi at intel.com>
Cc: Al.Bailey at windriver.com; starlingx-discuss at lists.starlingx.io
Subject: Re: [Containers] LP 1812519: config_controller --kubernetes fails at step 06

Hi Mingyuan, 

My issue is solved now, I didn't use a NAT network and that caused the problem.

Regarding the no_proxy, did you were able to pass the 'system application-apply stx-openstack' step?

I'm seeing a problem on tiller trying to get install osh-openstack- mariadb, which I think is related to not having a proper no_proxy value.

Is there a way to get all the IP needed to configure no_proxy, so the gRPC (and other mechanisms) can connect each other in the k8s cluster?

Thanks 

-Erich

On Wed, 2019-01-23 at 09:30 +0800, Qi, Mingyuan wrote:
> Erich,
> 
> Current docker version 18.03.1-ce in starlingx doesn't support 
> wildcard or CIDR notation subnet in no_proxy.
> Go-lang net lib used by docker supports CIDR notation in this 
> commit[0] which is after docker 18.03.1-ce.
> Have you tried with a particular ip address in no_proxy?
> 
> [0] https://github.com/golang/net/blob/c21de06aaf072cea07f3a65d6970e5
> c7d8b6cd6d/http/httpproxy/proxy.go#L39-L52
> 
> Mingyuan
> 
> 
> -----Original Message-----
> From: Cordoba Malibran, Erich
> Sent: Tuesday, January 22, 2019 23:53
> To: Frank.Miller at windriver.com; Qi, Mingyuan <mingyuan.qi at intel.com>
> Cc: Al.Bailey at windriver.com; starlingx-discuss at lists.starlingx.io
> Subject: Re: [Containers] LP 1812519: config_controller --kubernetes 
> fails at step 06
> 
> On Tue, 2019-01-22 at 09:20 +0800, Qi, Mingyuan wrote:
> > My latest try is built from last Wednesday’s code. I’ll try the 
> > latest code today along with the proxy patch.
> >  
> > Mingyuan
> >  
> > From: Miller, Frank [mailto:Frank.Miller at windriver.com]
> > Sent: Tuesday, January 22, 2019 3:39
> > To: Cordoba Malibran, Erich <erich.cordoba.malibran at intel.com>;
> > Qi,
> > Mingyuan <mingyuan.qi at intel.com>
> > Cc: Bailey, Henry Albert (Al) <Al.Bailey at windriver.com>; 
> > 'starlingx-d iscuss at lists.starlingx.io' 
> > <starlingx-discuss at lists.starlingx.io>
> > Subject: [Containers] LP 1812519: config_controller --kubernetes 
> > fails at step 06
> >  
> > Erich:
> >  
> > You indicated you saw this failure when using an ISO from Jan 17:
> > htt
> > ps://bugs.launchpad.net/starlingx/+bug/1812519
> >  
> > Al cannot reproduce this.  The one difference between Al’s 
> > environment and yours is Al does not use a proxy.  Mingyuan used a 
> > proxy a few weeks ago and this worked for him and he added steps to 
> > use a proxy to the wiki.  I have 2 questions:
> > 1.      Erich can you explain any changes you had to make in your
> > environment that are not listed on the current wiki:
> > https://wiki.openstack.org/wiki/StarlingX/Containers/Installation
> 
> I'm using a libvirt/qemu setup. I wasn't unable to set a NAT network 
> but I define some iptables rules to get internet access from the VM.
> 
> I set the proxy settings on /etc/environment and 
> /etc/systemd/system/docker.service.d/
> 
> This was my initial no_proxy
> no_proxy=localhost,127.0.0.1,192.168.206.2,172.16.0.0/16,10.96.0.0/12
> 
> I started adding some networks there as I saw errors on the 
> puppet.log.
> 
> At this point I can curl any host on internet, so external networking 
> is working.
> 
> With this setup I run the config_controller --kubernetes command and I 
> see the behavior described in the initial bug report.
> 
> Then I tried a different no_proxy value, Jose told me that he as able 
> to pass config_controller on virtualbox using only no_proxy=127.0.0.1.
> 
> With no_proxy=127.0.0.1 the config_controller --kubernetes succeed but 
> I can't do a 'source /etc/platform/openrc', I get the following
> errors:
> 
> controller-0:~$ source /etc/platform/openrc Openstack Admin 
> credentials can only be loaded from the active controller.
> 
> I tried to force execution of the controller_config script but I
> get: 
> 
> *****************************************************
> *****************************************************
> Unable to get IP from host: controller-0
> *****************************************************
> *****************************************************
> Pausing for 5 seconds...
> 
> and checking /etc/hosts the file is incomplete: 
> 
> # HEADER: This file was autogenerated at 2019-01-21 18:59:58 +0000 #
> HEADER: by puppet.  While it can still be managed manually, it #
> HEADER: is definitely not recommended.
> 127.0.0.1	localhost	localhost.localdomain localhost4
> localhost4.localdomain4
> 127.0.0.1	localhost	localhost.localdomain controller
> ::1	controller
> 
> 
> Anyway, something bad happened even with the config_controller running 
> without errors.
> 
> Also, I tried the VirtualBox in a non-proxy environment and everything 
> worked. Today I'll try VirtualBox behind a proxy.
> 
> As my libvirt/qemu setup has internet access I can't think on reason 
> on why it works with virtualbox, but no with my setup. I'll confirm 
> this.
> 
> 
> > 2.      Mingyuan can you tell us if you are able to get
> > config_controller –kubernetes to succeed with a load from Jan 17th 
> > or later?
> >  
> > Hopefully answers to the above 2 questions will point to why you are 
> > seeing failures while others are not.
> >  
> > Frank
> >  


More information about the Starlingx-discuss mailing list