git.net

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Tripleo Standalone Deployment Issues - Reproduceable


On Tue, Sep 8, 2020 at 2:07 PM <yuuta.takanashi at aol.com> wrote:
>
> Hi Experts,
>
> I am facing on below issues while following guide of Tripleo Standalone Deployment:
> https://docs.openstack.org/project-deploy-guide/tripleo-docs/latest/deployment/standalone.html
>
> Issue 1) Using custom domain will cause vm instantiation error after machine reboot
>
>
> Description
> ===========
> After Openstack installed, it's adding centos82.localdomain on /etc/hosts and causing OVN Southbound (ovn-sbctl show) using this centos82.localdomain. But after the machine rebooted, "ovn-sbctl show" is showing the correct fqdn: centos82.domain.tld, and it is causing VM instantiate error (Refusing to bind port <UUID> due to no OVN chassis for host: centos82.localdomain).
>
>
> Steps to reproduce
> ==================
> Base installation: https://docs.openstack.org/project-deploy-guide/tripleo-docs/latest/deployment/standalone.html
> Tested on Centos 8.2 latest, tested several times on Train and Ussuri.
> 100% reproduceable.
>
>
> Set the custom FQDN hostname
>
> hostnamectl set-hostname centos82.domain.tld
> hostnamectl set-hostname centos82.domain.tld --transient    (also tested with or without this line)
>
> cat /etc/hosts
> 127.0.0.1       centos82.domain.tld localhost localhost.localdomain localhost4 localhost4.localdomain4
> ::1             centos82.domain.tld localhost localhost.localdomain localhost6 localhost6.localdomain6
> 192.168.156.82  centos82.domain.tld
>
> also tested with
>
> hostnamectl set-hostname centos82.domain.tld
> hostnamectl set-hostname centos82.domain.tld --transient    (also tested with or without this line)
>
> cat /etc/hosts
> 127.0.0.1       localhost localhost.localdomain localhost4 localhost4.localdomain4
> ::1             localhost localhost.localdomain localhost6 localhost6.localdomain6
> 192.168.156.82  centos82.domain.tld centos82
>
> also tested with
>
> hostnamectl set-hostname centos82.domain.tld
> hostnamectl set-hostname centos82.domain.tld --transient    (also tested with or without this line)
>
> cat /etc/hosts
> 127.0.0.1       localhost localhost.localdomain localhost4 localhost4.localdomain4
> ::1             localhost localhost.localdomain localhost6 localhost6.localdomain6
> 192.168.156.82  centos82.domain.tld
>
>
> Part of standalone_parameters.yaml for domain:
>   # domain name used by the host
>   NeutronDnsDomain: domain.tld
>
>
>
> Follow the guide to install (without ceph)
>
> (Issue 1B) with ceph was having another error, just tested once on Train:
> fatal: [undercloud]: FAILED! => {
>     "ceph_ansible_std_out_err": [
>         "Using /home/stack/standalone-ansible-huva1ccw/ansible.cfg as config file",
>         "ERROR! the playbook: /usr/share/ceph-ansible/site-container.yml.sample could not be found"
>     ],
>     "failed_when_result": true
> }

It looks like ceph-ansible was executed by tripleo but that
ceph-ansible isn't installed. Either install ceph-ansible or ensure
your deployment command does not include:

  -e /usr/share/openstack-tripleo-heat-templates/environments/ceph-ansible/ceph-ansible.yaml
\

  John

>
>
>
>
> All will be installed successfully, and able to instantiate VMs and these VMs are able to ping to each other.
>
> Then do "reboot" on this standalone host/server.
>
> Do instantiate a new VM, and then this new VM status will be ERROR.
>
> On Neutron logs:
> Refusing to bind port <UUID> due to no OVN chassis for host: centos82.localdomain bind_port
>
>
>
> Found the cause on OVN:
>
> Previous after installed (before reboot)
>
> [root at centos82 /]# ovn-sbctl show
> Chassis "<UUID>"
>     hostname: centos82.localdomain
>     Encap geneve
>         ip: "192.168.156.82"
>         options: {csum="true"}
> [root at centos82 /]#
>
>
> After reboot
>
> [root at centos82 /]# ovn-sbctl show
> Chassis "<UUID>"
>     hostname: centos82.domain.tld
>     Encap geneve
>         ip: "192.168.156.82"
>         options: {csum="true"}
> [root at centos82 /]#
>
>
> So that the OVN seems could not bind the port to different hostname (that already changed).
>
>
> Environment
> ===========
> 1. Exact version of OpenStack you are running: Any, tested Train and Ussuri
> 2. Centos 8.2 latest update
> Others (without ceph): https://docs.openstack.org/project-deploy-guide/tripleo-docs/latest/deployment/standalone.html
>
>
>
> Workaround
> ==========
> Don't use custom domain, instead use "localdomain" as on the document
>
> hostnamectl set-hostname centos82.localdomain
>
> [root at centos82 ~]# cat /etc/hosts
> 127.0.0.1   localhost localhost.localdomain localhost4 localhost4.localdomain4
> ::1         localhost localhost.localdomain localhost6 localhost6.localdomain6
> 192.168.156.82 centos82.localdomain centos82
>
> Deploy, test instantiate, reboot, test instantiate, and working perfectly fine.
>
>
> Question: For custom domain, where did I do wrongly? Is it expected or kind of bugs?
>
>
>
>
>
> Issue 2) This is not Openstack issue, perhaps bugs or perhaps my issue, but it is still related with the topic on Tripleo Standalone Deployment.
>
> On the same Standalone Deployment Guide (https://docs.openstack.org/project-deploy-guide/tripleo-docs/latest/deployment/standalone.html)
> with Centos 7.6 or 7.7 with "python-tripleoclient" and this has dependency of Docker 1.13.1, encounters kind of Dns querying issue which I posted here: https://serverfault.com/questions/1032816/centos-7-and-docker-1-13-1-error-timeout-exceeded-while-awaiting-headers-no
> Perhaps if anybody knows how to resolve this.
> Also, just wondering is there any way for "python-tripleoclient" to use newer docker-ce 1.19.x instead of docker 1.13.1?
>
>
>
>
> Thanks
> Best regards,
> TY
>