sriovdp generates many log files when not configured properly

[Lucas Ratusznei Fonseca]

*+Brief Description+*

When the label "sriovdp=enabled" is assigned to the host and no PCI-SRIOV interfaces are enabled, the sriov plugin pod remains in a crash loop and generates a large number of log files in /var/log/sriovdp.

*+Severity+*

Minor: No negative system impact other than too many logs being generated which then causes collect to timeout.

*+Steps to Reproduce+*

Step1:
Add host label "sriovdp" to the host, but not enable any of PCI-SRIOV interface.
  ->  Then sriov-device-plugin was failed with "CrashLoopBackOff".  This will create a lot of files under /var/log/sriovdp/ on the target node.

kube-system                   kube-sriov-device-plugin-amd64-97tw8              0/1     CrashLoopBackOff   6983       24d

*+Expected Behavior+*

It should not be possible to unlock the host if the label "sriovdp=enabled" is assigned and no PCI-SRIOV interfaces are enabled.
The sriov plugin pod should not crash, even if the system is not configured properly.

*+Actual Behavior+*

Excessive log files are generated by sriovdp

*+Reproducibility+*

Reproducible

*+System Configuration+*

Any

*+Last Pass+*

N/A

*+Timestamp/Logs+*

Not Required

*+Alarms+*

No alarm

*+Test Activity+*

N/A

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/starlingx/config/+/869774

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/starlingx/integ/+/869777

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/starlingx/tools/+/869778

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/starlingx/stx-puppet/+/869852

[OpenStack Infra (hudson-openstack)]

Change abandoned by "Lucas Ratusznei Fonseca <email address hidden>" on branch: master
Review: https://review.opendev.org/c/starlingx/integ/+/869777
Reason: The solution is not effective, a better one was proposed here: https://review.opendev.org/c/starlingx/stx-puppet/+/869852

[OpenStack Infra (hudson-openstack)]

Change abandoned by "Lucas Ratusznei Fonseca <email address hidden>" on branch: master
Review: https://review.opendev.org/c/starlingx/tools/+/869778
Reason: The solution is not effective, a better one was proposed here: https://review.opendev.org/c/starlingx/stx-puppet/+/869852

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/c/starlingx/config/+/869774
Committed: https://opendev.org/starlingx/config/commit/16e800bcf9a9fc2bd97cc05ba0213091352c6130
Submitter: "Zuul (22348)"
Branch: master

commit 16e800bcf9a9fc2bd97cc05ba0213091352c6130
Author: Lucas Ratusznei Fonseca <email address hidden>
Date: Tue Jan 10 19:19:57 2023 -0300

    Semantic check for proper sriov plugin configuration

    This commit adds a semantic check to host-unlock to prevent unlocking
    if the label "sriovdp=enabled" is present and there are no PCI-SRIOV
    interfaces enabled.

    Test plan

    PASS Unlock system with sriovdp label absent and 0 PCI-SRIOV
         interfaces enabled
    PASS Unlock system with sriovdp label absent and 1 PCI-SRIOV
         interface enabled
    PASS Unlock system with sriovdp label present and 0 PCI-SRIOV
         interfaces enabled
    PASS Try to unlock system with sriovdp label present and 1 PCI-SRIOV
         interface enabled (command fails)

    Closes-Bug: #2002447

    Signed-off-by: Lucas Ratusznei Fonseca
                   <email address hidden>
    Change-Id: I8d9cba7354f066ad334af425fe26f57a93f58d3d

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/c/starlingx/stx-puppet/+/869852
Committed: https://opendev.org/starlingx/stx-puppet/commit/c39047149a5ab9e5b26be32280f336f360ef6fd1
Submitter: "Zuul (22348)"
Branch: master

commit c39047149a5ab9e5b26be32280f336f360ef6fd1
Author: Lucas Ratusznei Fonseca <email address hidden>
Date: Wed Jan 11 12:12:46 2023 -0300

    Add minimal config for the sriov device plugin

    This commit adds a minimal configuration for the sriov device plugin.
    If the config file (/etc/pcidp/config.json) has invalid contents or
    is empty, the plugin remains in a crash loop, generating a large
    number of log files. This change ensures that the config file has
    valid content, thus preventing such an issue.

    Test plan

    PASS Run Puppet in host with 0 PCI-SRIOV interfaces configured
    PASS Run Puppet in host with 1 PCI-SRIOV interface configured

    Closes-Bug: #2002447

    Signed-off-by: Lucas Ratusznei Fonseca
                   <email address hidden>
    Change-Id: I0ec6e9ebaad7b8ed7981f2f1af7b15195a3bfd43

[Ghada Khalil (gkhalil)]

Re-opening to revert the semantic check introduced in https://review.opendev.org/c/starlingx/config/+/869774 as it's causing unlocks to fail for some configurations where the sriovdp=enabled label is defined, but the interfaces are not configured yet.

See review: https://review.opendev.org/c/starlingx/config/+/902888 for more details

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/c/starlingx/config/+/902888
Committed: https://opendev.org/starlingx/config/commit/ce0959e71c76d2e8e3d96a293f855f85efad4e14
Submitter: "Zuul (22348)"
Branch: master

commit ce0959e71c76d2e8e3d96a293f855f85efad4e14
Author: Andre Kantek <email address hidden>
Date: Thu Dec 7 16:12:32 2023 -0300

    Revert "Semantic check for proper sriov plugin configuration"

    This reverts commit 16e800bcf9a9fc2bd97cc05ba0213091352c6130.

    Reason for revert:
    This validation is preventing a 2 step approach to the configuration,
    first configuring labels and later adding the interface configuration.
    With the change:
    https://review.opendev.org/c/starlingx/stx-puppet/+/869852
    there is no reason to keep this validation as the pod sriov device
    plugin will no longer generate log spam since a dummy config file
    will be created, avoiding the state of CrashBackLoop.

    Closes-Bug: 2002447

    Change-Id: Ice4def2a23c5fe5fc0311bafff1a8ea5051e1609
    Signed-off-by: Andre Kantek <email address hidden>