Wallaby install via kayobe onto ubuntu 20 all in one host

Tony Pearce tonyppe at gmail.com
Wed Jun 16 08:48:24 UTC 2021


Possibly fact caching could have been the issue. The Victoria guide
describes to configure fact caching so it was being used by my ACH. I did
not think of this at the time of the issue, sorry about that.

I have a new question about SSL cert, so will send a new mail for the
topic.

Thank you again,

Kind regards,

Tony Pearce



On Wed, 16 Jun 2021 at 16:09, Mark Goddard <mark at stackhpc.com> wrote:

> On Wed, 16 Jun 2021 at 07:09, Tony Pearce <tonyppe at gmail.com> wrote:
> >
> > Hi Mark, thanks again for your reply and direction on this.
> >
> > Regarding the need to do a pip install within the venv, I re-installed
> the host and re-installed kayobe and this issue is no longer present as of
> yesterday afternoon.
> >
> > Regarding the "Waiting for nova-compute services to register themselves"
> - I tailed *.log in the kolla/nova log dir and to be honest there wasnt
> much being logged at that time and no errors, either.
> >
> > Regarding Wallaby hasnt yet been released, I went back to Victoria on
> Centos 8.4 this morning, using the "pip install ." (installation from
> source). I will use this install from source going forward to make sure I
> am getting the latest kayobe patches. I had two blockers during "host
> configure":
> >
> > First was that python was not found on the host:
> >
> > TASK [Verify that a command can be executed]
> *****************************************************************************************************************************
> > fatal: [juc-ucsb-5-p]: FAILED! => {"changed": false, "module_stderr":
> "/bin/sh: /usr/bin/python3: No such file or directory\n", "module_stdout":
> "", "msg": "The module failed to execute correctly, you probably need to
> set the interpreter.\nSee stdout/stderr for the exact error", "rc": 127}
>
> In stable/victoria there is a task called 'Ensure python is installed'
> which should ensure that Python 3 is installed. Did that task run?
>
> >
> > So I manually installed python to see what would happen. I was able to
> get past that point but then it fails again:
> >
> > TASK [Ensure the Python virtualenv package is installed]
> *****************************************************************************************************************
> > [WARNING]: Updating cache and auto-installing missing dependency:
> python3-apt
> > fatal: [juc-ucsb-5-p]: FAILED! => {"changed": false, "cmd": "apt-get
> update", "msg": "[Errno 2] No such file or directory: b'apt-get':
> b'apt-get'", "rc": 2}
> >
> > PLAY RECAP
> ***************************************************************************************************************************************************************
> > juc-ucsb-5-p               : ok=7    changed=0    unreachable=0
> failed=1    skipped=10   rescued=0    ignored=0
> >
> > The above log suggests that the CentOS 8.4 host is being matched as
> ubuntu. As an FYI, what I mean when I say "went back to Victoria" is that I
> followed the initial kayobe install steps again with new top directory
> structure and then manually transferred over the kayobe config to the new
> empty config which is downloaded during the git clone. "manually
> transferred" means I opened each yml config file individually and copied
> over only the variables that have been configured within the file, making
> sure to leave defaults as a prefernce.
>
> Something odd is clearly going on here. We use the package module,
> which uses ansible's ansible_pkg_mgr fact to dispatch the correct
> packaging module. Perhaps you have fact caching enabled, and have
> stale facts from when the host was running Ubuntu?
>
> >
> > After the above issue I decided to go back to Train and have been able
> to successfully deploy Openstack using kayobe Train today - no blocking
> issues, first time.
> >
> > Summary
> > I really appreciate your time to help me with this, so thanks again. I
> Still think kayobe is the best way to deploy and manage openstack and I
> have been "telling my friends" also :) Now I am back to this point I just
> have some minor things to work through before I can start using it.
> >
> > Kind regards,
> >
> > Tony Pearce
> >
> >
> > On Tue, 15 Jun 2021 at 16:02, Mark Goddard <mark at stackhpc.com> wrote:
> >>
> >> On Tue, 15 Jun 2021 at 06:51, Tony Pearce <tonyppe at gmail.com> wrote:
> >> >
> >> > Hi Mark,
> >> >
> >> > I had never used the "pip install ." method. Maybe a miscomprehension
> on my side, from the documentation [1] there are three ways to install
> kayobe. I had opted for the first way which is "pip install kayobe" since
> January 2020. The understanding was as conveyed in the doc "Installing from
> PyPI ensures the use of well used and tested software".
> >>
> >> That is true, but since Wallaby has not been released for Kayobe yet,
> >> it is not on PyPI. If you do install from PyPI, I would advise using a
> >> version constraint to ensure you get the release series you need.
> >>
> >> >
> >> > I have since followed your steps in your mail which is the
> installation from source. I had new problems:
> >> >
> >> > During ansible bootstrap:
> >> > During ansible host bootstrap it errors out and says the
> kolla_ansible is not found and needs to be installed in the same virtual
> environment. In all previous times, I had understood that kolla ansible is
> installed by kayobe at this point. I eventually done "pip install
> kolla-ansible" and it seemed to take care of that and allowed me to move on
> to "host configure"
> >>
> >> Kolla Ansible should be installed automatically during 'kayobe control
> >> host bootstrap', in a separate virtualenv from Kayobe. You should not
> >> need to install it manually, and I would again advise against doing so
> >> without version constraints.
> >>
> >> >
> >> > During host configure:
> >> > I was able to get past the previous python issue but then it failed
> on the network due to a "duplicate bond name", though this config was
> deployed successfully in Train. I dont think I really need a bond at this
> point so I deleted the bond and the host configure is now successful. (fyi
> this is an all-in-one host.)
> >> >
> >> > During kayobe service deploy:
> >> > This then fails with "no module named docker" on the host. To
> troubleshoot this I logged into the host and activated the kayobe virtual
> env (/opt/kayobe/venvs/kayobe/bin/activate) and then "pip install docker".
> It was already installed. Eventually, I issued "pip install
> --ignore-installed docker" within these three (environment) locations which
> resolved this and allowed the kayobe command to complete successfully and
> progress further:
> >> > - /opt/kayobe/venvs/kayobe/
> >> > - /opt/kayobe/venvs/kolla-ansible/
> >> > - native on the host after deactivating the venv.
> >> >
> >> > Now the blocker is the following failure;
> >> >
> >> > TASK [nova-cell : Waiting for nova-compute services to register
> themselves]
> **********************************************************************************************
> >> > FAILED - RETRYING: Waiting for nova-compute services to register
> themselves (20 retries left).
> >> > FAILED - RETRYING: Waiting for nova-compute services to register
> themselves (19 retries left).
> >> >
> >> > I haven't seen this one before but previously I had seen something
> similar with mariadb because the API dns was not available. What I have
> been using here is a /etc/hosts entry for this. I checked that this entry
> is available on the host and in the nova containers. I decided to reboot
> the host anyway (previously resolved similar mariadb issue) to restart the
> containers just in case the dns was not available in one of them and I
> missed it.
> >>
> >> I'd check the nova compute logs here, to find why they are not
> >> registering themselves.
> >> >
> >> > Unfortunately I now have two additional issues which are hard
> blockers:
> >> > 1. The network is no longer working on the host after reboot, so I am
> unable to ssh
> >> > 2. The user password has been changed by kayobe, so I am unable to
> login using the console
> >> >
> >> > Due to the above, I am unable to login to the host to investigate or
> remediate. Previously when this happened with centos I could use the root
> user to log in. This time around as it's ubuntu I do not have a root user.
> >> > The user I am using for both "kolla_ansible_user" and
> "kayobe_ansible_user" is the same - is this causing a problem with Victoria
> and Wallaby? I had this user password change issue beginning with Victoria.
> >> >
> >> > So at this point I need to re-install the host and go back to the
> host configure before service deploy.
> >> >
> >> > Summary
> >> > Any guidance is well appreciated as I'm at a loss at this point. Last
> week I had a working Openstack Train deployment in a single host. "Kayobe"
> stopped working (maybe because I had previously always used pip install
> kayobe).
> >> >
> >> > I would like to deploy Wallaby, should I be able to successfully do
> this today or should I be using Victoria at the moment (or even, Train)?
> >>
> >> We are very close to release of Wallaby, and I expect that it should
> >> generally work, but Ubuntu is a new distro for Kayobe, and Wallaby is
> >> a new release. There may be teething problems, so if you're looking
> >> for something more stable then I'd suggest CentOS & Victoria.
> >>
> >> >
> >> > [1] OpenStack Docs: Installation
> >> >
> >> > Regards,
> >> >
> >> > Tony Pearce
> >> >
> >> >
> >> > On Mon, 14 Jun 2021 at 18:36, Mark Goddard <mark at stackhpc.com> wrote:
> >> >>
> >> >> On Mon, 14 Jun 2021 at 09:40, Tony Pearce <tonyppe at gmail.com> wrote:
> >> >> >
> >> >> > Hi Mark,
> >> >> >
> >> >> > I followed this guide to do a "git clone" specifying the branch
> "-b" to "stable/wallaby" [1]. What additional steps do I need to do to get
> the latest commits?
> >> >>
> >> >> That should be sufficient. When you install it via pip, note that
> 'pip
> >> >> install kayobe' will still pull from PyPI, even if there is a local
> >> >> kayobe directory. Use ./kayobe, or 'pip install .'  if in the same
> >> >> directory.
> >> >>
> >> >> Mark
> >> >> >
> >> >> > [1] OpenStack Docs: Overcloud
> >> >> >
> >> >> > Kind regards,
> >> >> >
> >> >> > Tony Pearce
> >> >> >
> >> >> >
> >> >> > On Mon, 14 Jun 2021 at 16:10, Mark Goddard <mark at stackhpc.com>
> wrote:
> >> >> >>
> >> >> >> On Mon, 14 Jun 2021 at 07:21, Tony Pearce <tonyppe at gmail.com>
> wrote:
> >> >> >> >
> >> >> >> > Hi Pierre, thanks for replying to my message.
> >> >> >> >
> >> >> >> > To install kayobe I followed the documentation which summarise:
> installing a few system packages and setting up the kayobe virtual
> environment and then pulling the correct kayobe git version for the
> openstack to be installed. After configuring the yaml files I have run
> these commands:
> >> >> >> >
> >> >> >> > - kayobe control host bootstrap
> >> >> >> > - kayobe overcloud host configure -> this one is failing with
> /usr/libexec/platform-python: not found
> >> >> >> >
> >> >> >> > After reading your message on the weekend I concluded that
> maybe I had done something wrong. Today, I re-pulled the kayobe wallaby git
> and manually transferred the configuration over to the new directory
> structure on the ansible host and set up again as per the guide but the
> same issue is seen.
> >> >> >> >
> >> >> >> > What I ended up doing to try and resolve was finding where this
> "platform-python" is coming from. It is coming from the virtual environment
> which is being set up during the kayobe ansible host bootstrap. Initially,
> I found the base.yml and it looks like it tries to match what the host is.
> I noticed that there is no ubuntu 20 listed there so I created it however
> it did not resolve the issue.
> >> >> >> >
> >> >> >> > So then I tried systematically replacing this reference in the
> other files found in the same location "venvs\kayobe\share\kayobe\ansible".
> The file I changed which allowed it to progress is "kayobe-target-venv.yml"
> >> >> >> >
> >> >> >> > But unfortunately it fails a bit further on, failing to find an
> selinux package [1]
> >> >> >> >
> >> >> >> > Seeing as the error is mentioning selinux (a RedHat security
> feature not installed on ubuntu) could the root cause issue be that kayobe
> is not matching the host as ubuntu? I did already set in kayobe that I am
> using ubuntu OS distribution within globals.yml [2].
> >> >> >> >
> >> >> >> > Are there any extra steps that I need to complete that maybe
> are not listed in the documentation / guide?
> >> >> >> >
> >> >> >> > [1] TASK [MichaelRigart.interfaces : Debian | install
> current/latest network package - Pastebin.com
> >> >> >> > [2] ---# Kayobe global
> configuration.######################################### - Pastebin.com
> >> >> >>
> >> >> >> Hi Tony,
> >> >> >>
> >> >> >> That's definitely not a recent Wallaby checkout you're using.
> Ubuntu
> >> >> >> no longer uses that MichaelRigart.interfaces role. Check that you
> have
> >> >> >> recent commits. Here is the most recent on stable/wallaby:
> >> >> >> 13169077aaec0f7a28ae1f15b419dafc2456faf7.
> >> >> >>
> >> >> >> Mark
> >> >> >>
> >> >> >> >
> >> >> >> > Regards,
> >> >> >> >
> >> >> >> > Tony Pearce
> >> >> >> >
> >> >> >> >
> >> >> >> >
> >> >> >> > On Fri, 11 Jun 2021 at 21:05, Pierre Riteau <
> pierre at stackhpc.com> wrote:
> >> >> >> >>
> >> >> >> >> Hi Tony,
> >> >> >> >>
> >> >> >> >> Kayobe doesn't use platform-python anymore, on both
> stable/wallaby and
> >> >> >> >> stable/victoria:
> >> >> >> >>
> https://review.opendev.org/q/I0d477325e0edd13d1aba211c13dc2e8b7a9b4c98
> >> >> >> >>
> >> >> >> >> Can you double-check what version you are using, and share how
> you
> >> >> >> >> installed it? Note that only stable/wallaby supports Ubuntu 20
> hosts.
> >> >> >> >>
> >> >> >> >> Best wishes,
> >> >> >> >> Pierre
> >> >> >> >>
> >> >> >> >> On Fri, 11 Jun 2021 at 13:20, Tony Pearce <tonyppe at gmail.com>
> wrote:
> >> >> >> >> >
> >> >> >> >> > I'm trying to run "kayobe overcloud host configure" against
> an ubuntu 20 machine to deploy Wallaby. I'm getting an error that python is
> not found during the host configure part.
> >> >> >> >> >
> >> >> >> >> > PLAY [Verify that the Kayobe Ansible user account is
> accessible]
> >> >> >> >> > TASK [Verify that a command can be executed]
> >> >> >> >> >
> >> >> >> >> > fatal: [juc-ucsb-5-p]: FAILED! => {"changed": false,
> "module_stderr": "/bin/sh: 1: /usr/libexec/platform-python: not found\n",
> "module_stdout": "", "msg": "The module failed to execute correctly, you
> probably need to set the interpreter.\nSee stdout/stderr for the exact
> error", "rc": 127}
> >> >> >> >> >
> >> >> >> >> > Python3 is installed on the host. When searching where this
> platform-python is coming from it returns the kolla-ansible virtual envs:
> >> >> >> >> >
> >> >> >> >> > $ grep -rni -e "platform-python"
> >> >> >> >> >
> venvs/kolla-ansible/lib/python3.8/site-packages/ansible/config/base.yml:1450:
>     '8': /usr/libexec/platform-python
> >> >> >> >> >
> venvs/kolla-ansible/lib/python3.8/site-packages/ansible/config/base.yml:1470:
> - /usr/libexec/platform-python
> >> >> >> >> >
> >> >> >> >> > I had a look through the deployment guide for Kayobe Wallaby
> and didnt see a note about changing this.
> >> >> >> >> >
> >> >> >> >> > Do I need to do further steps to support the ubuntu
> overcloud host? I have already set (as per the doc):
> >> >> >> >> >
> >> >> >> >> > os_distribution: ubuntu
> >> >> >> >> > os_release: focal
> >> >> >> >> >
> >> >> >> >> > Regards,
> >> >> >> >> >
> >> >> >> >> > Tony Pearce
> >> >> >> >> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-discuss/attachments/20210616/12c4a51b/attachment-0001.html>


More information about the openstack-discuss mailing list