Uploaded image for project: 'Red Hat OpenStack Services on OpenShift'
  1. Red Hat OpenStack Services on OpenShift
  2. OSPRH-2301

As a cloud operator, i want Adoption without having to SSH into compute nodes.

    • As a cloud operator, i want Adoption without having to SSH into compute nodes.
    • 2
    • False
    • Committed
    • Committed
    • Done
    • RHOSSTRAT-204 - Red Hat OpenStack 18.0 Data Plane Adoption
    • openstack-ansible-ee-container-1.0.0-20
    • Committed
    • Committed
    • 0% To Do, 0% In Progress, 100% Done
    • Release Note Not Required
    • Hide

      RHOSO18Beta waived:Upgrades: Adoption

      Show
      RHOSO18Beta waived: Upgrades : Adoption
    • Rejected
    • 2024Q2
    • Approved

      Currently we ssh into compute nodes in at least two places in Adoption:

      • "Stop OpenStack services" step, currently puts together controller and compute services as it was developed with the Standalone use case. (For supporting separate compute we'd have to split the compute services out and ssh to computes.)
      • OVN adoption step, when redirecting compute node ovn-controller services from the old southbound DB to the new podified southbound DB.

      Since there can be many compute nodes in a deployment, it would be very beneficial for Adoption UX to do the necessary compute node actions via EDPM rather than via SSH (using NodeSets and Deployments, regardless if some Adoption-specific Deployment or the "main" data plane Deployment). It would be good if we could push all data plane operations after the control plane has been fully adopted because it would help with rolling back the control plane (OSPRH-1890). Touching the data plane config would be considered a point of no return for the rollback procedure.

      Epic definition of done:

      • We can still Adopt all services as before but there are no steps which SSH into compute nodes.

      Proposed release blocker due to being a hard dependency of OSPRH-1890.

            [OSPRH-2301] As a cloud operator, i want Adoption without having to SSH into compute nodes.

            Errata Tool added a comment -

            Since the problem described in this issue should be resolved in a recent advisory, it has been closed.

            For information on the advisory (Data plane Operators for RHOSO 18.0), and where to find the updated files, follow the link below.

            If the solution does not work for you, open a new bug report.
            https://access.redhat.com/errata/RHEA-2024:5247

            Errata Tool added a comment - Since the problem described in this issue should be resolved in a recent advisory, it has been closed. For information on the advisory (Data plane Operators for RHOSO 18.0), and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHEA-2024:5247

            dataplane adoption does not include any compute ssh functionality as it is now , any traces were removed and implemented in edpm

            pini komarov added a comment - dataplane adoption does not include any compute ssh functionality as it is now , any traces were removed and implemented in edpm

            should be done once 17.1.3 is out :
            17.1.3: Release Window May 4 - May 17
            according to https://docs.google.com/document/d/1-BaMrV5lKNmThnIoKU0gjOkNBOBERaobaKnpCQcNeds/edit

            pini komarov added a comment - should be done once 17.1.3 is out : 17.1.3: Release Window May 4 - May 17 according to https://docs.google.com/document/d/1-BaMrV5lKNmThnIoKU0gjOkNBOBERaobaKnpCQcNeds/edit

            pkomarovI have no idea

            Balazs Gibizer added a comment - pkomarov I have no idea

            jstransk@redhat.com rh-ee-bgibizer how can I signal Jira dependency when "17.1.3 available on the CDN" , here , any release-delivery Jira I can point to ?

            pini komarov added a comment - jstransk@redhat.com rh-ee-bgibizer how can I signal Jira dependency when "17.1.3 available on the CDN" , here , any release-delivery Jira I can point to ?

            jstransk@redhat.comCorrect. We need to have 17.1.3 available on the CDN so that the deployed 17.1 node in CI has the backported feature from 17.1.3. that does the prepopulation of compute-id in 17 side.

            Balazs Gibizer added a comment - jstransk@redhat.com Correct. We need to have 17.1.3 available on the CDN so that the deployed 17.1 node in CI has the backported feature from 17.1.3. that does the prepopulation of compute-id in 17 side.

            Hi jstransk@redhat.com I see that according to the remove adoption stop comp services PR
            That the maintanance for this epic's topic is on the edpm team's (and repo) side.
            So testing and verification is also moving to them I believe.

            So to fulfill this epic's testing requirement we would need to show D/S adoption testing without the use of ssh to copmute to stop services on the doption repo's side. As I see that requirement is fulfilled with the PRs merged, and jobs running without that code.
            If it's not the case please let me know

            pini komarov added a comment - Hi jstransk@redhat.com I see that according to the remove adoption stop comp services PR That the maintanance for this epic's topic is on the edpm team's (and repo) side. So testing and verification is also moving to them I believe. So to fulfill this epic's testing requirement we would need to show D/S adoption testing without the use of ssh to copmute to stop services on the doption repo's side. As I see that requirement is fulfilled with the PRs merged, and jobs running without that code. If it's not the case please let me know

            Yes that's fine, we only wanted to remove compute node ssh.

            Jiri Stransky added a comment - Yes that's fine, we only wanted to remove compute node ssh.

            FTR we cannot remove remaining CONTROLLER_SSH commands as those are expected to be executed on the source cloud controllers, not computes. If we wish to remove it, we need to use tripleo inventories instead

            Bohdan Dobrelia added a comment - FTR we cannot remove remaining CONTROLLER_SSH commands as those are expected to be executed on the source cloud controllers, not computes. If we wish to remove it, we need to use tripleo inventories instead

            Status: Both stories (OSPRH-2302, OSPRH-2303) are still targeted for Q1.

            Jiri Stransky added a comment - Status: Both stories ( OSPRH-2302 , OSPRH-2303 ) are still targeted for Q1.

              jstransk@redhat.com Jiri Stransky
              jstransk@redhat.com Jiri Stransky
              pini komarov pini komarov
              rhos-dfg-upgrades
              Votes:
              0 Vote for this issue
              Watchers:
              13 Start watching this issue

                Created:
                Updated:
                Resolved: