v1.2.2-kata-ci

Fix the container image for the workloads using the clusterbuster-workloads image.

The intent was always for them to use the :-kata-ci-v1.2 image, but by oversight this was only fixed for the primary image (clusterbuster), not for the workloads images. This broke when the :-latest tags were deleted as part of planned cleanup.

v1.2.1-kata-ci

Bugfix:

• cpusoaker should run all cases in each runtimeclass sequentially rather than interleaved. The problem with interleaving them is that the Prometheus metrics gathered during a run may partially reflect the previous run, as Prometheus gathers data over a 30 second interval, which may result in incorrect per-pod memory calculation if one runtimeclass consumes much more memory per pod than another.

v1.2.0-kata-ci

Rebase the CI to the current mainline.

• Extensive changes since v1.1.7:
  • Fix pod privilege issues going forward
  • Improved robustness, particularly in case of failure
  • Improved reliability of reporting in the case of run failures
  • More reliable log collection
  • All pod-side workloads converted to Python:
    • cpusoaker will give lower performance numbers due to use of Python vs. perl; results cannot be compared against v1.1
    • memory consumption will likely be lower with cpusoaker test
  • Pod monitoring is more efficient
  • Support for Multus
  • Support ARM
  • Image tag is separate from the mainline image to allow future development without impacting CI.

Except as noted (cpusoaker performance results), this release should be compatible with v1.1.7 for purposes of CI.

v1.2.0-rc.3-kata-ci

• Changes since v1.2.0-rc.2-kata-ci:
  • Fix drop cache pod in fio and files workload to properly tell clusterbuster to set privilege on the container.
  • Fix run-kata-perf-suite to properly create the command line to analyze-clusterbuster-results.
  • Retrieve pod description if sync pod fails to start.
• Changes since v1.2.0-rc.1-kata-ci:
  • Fix trivial typo in image name.
• Changes since v1.2.0-rc.0-kata-ci:
  • Fixed regression: missing --force-cleanup-i-know-this-is-dangerous
  • Created new tags for the container image to separate the branch from mainline development.
• Extensive changes since v1.1.7:
  • Fix pod privilege issues going forward
  • Improved robustness, particularly in case of failure
  • Improved reliability of reporting in the case of run failures
  • More reliable log collection
  • All pod-side workloads converted to Python:
    • cpusoaker will give lower performance numbers due to use of Python vs. perl; results cannot be compared against v1.1
    • memory consumption will likely be lower with cpusoaker test
  • Pod monitoring is more efficient
  • Support for Multus
  • Support ARM

Except as noted (cpusoaker performance results), this release should be compatible with v1.1.7 for purposes of CI.

v1.2.0-rc.2-kata-ci

• Changes since v1.2.0-rc.1-kata-ci:
  • Fix trivial typo in image name.
• Changes since v1.2.0-rc.0-kata-ci:
  • Fixed regression: missing --force-cleanup-i-know-this-is-dangerous
  • Created new tags for the container image to separate the branch from mainline development.
• Extensive changes since v1.1.7:
  • Fix pod privilege issues going forward
  • Improved robustness, particularly in case of failure
  • Improved reliability of reporting in the case of run failures
  • More reliable log collection
  • All pod-side workloads converted to Python:
    • cpusoaker will give lower performance numbers due to use of Python vs. perl; results cannot be compared against v1.1
    • memory consumption will likely be lower with cpusoaker test
  • Pod monitoring is more efficient
  • Support for Multus
  • Support ARM

Except as noted (cpusoaker performance results), this release should be compatible with v1.1.7 for purposes of CI.

v1.2.0-rc.1-kata-ci

• Changes since v1.2.0-rc.0-kata-ci:
  • Fixed regression: missing --force-cleanup-i-know-this-is-dangerous
  • Created new tags for the container image to separate the branch from mainline development.
• Extensive changes since v1.1.7:
  • Fix pod privilege issues going forward
  • Improved robustness, particularly in case of failure
  • Improved reliability of reporting in the case of run failures
  • More reliable log collection
  • All pod-side workloads converted to Python:
    • cpusoaker will give lower performance numbers due to use of Python vs. perl; results cannot be compared against v1.1
    • memory consumption will likely be lower with cpusoaker test
  • Pod monitoring is more efficient
  • Support for Multus
  • Support ARM

Except as noted (cpusoaker performance results), this release should be compatible with v1.1.7 for purposes of CI.

v1.2.0-rc.0-kata-ci

• Extensive changes since v1.1.7:
  • Fix pod privilege issues going forward
  • Improved robustness, particularly in case of failure
  • Improved reliability of reporting in the case of run failures
  • More reliable log collection
  • All pod-side workloads converted to Python:
    • cpusoaker will give lower performance numbers due to use of Python vs. perl; results cannot be compared against v1.1
    • memory consumption will likely be lower with cpusoaker test
  • Pod monitoring is more efficient
  • Support for Multus
  • Support ARM

Except as noted (cpusoaker performance results), this release should be compatible with v1.1.7 for purposes of CI.

v1.1.7-kata-ci

Backport #52

v1.1.6-kata-ci

• Backport #44 (fix race condition in initial timestamp handshake)

v1.1.5-kata-ci

• Enhancements
  • Undocumented feature to enable force delete.