Orka 3.2.1 Release Notes

Improvements and fixes

Image caching deletion fix

The Orka Operator can in rare occurrences fail to update an Orka node status if a cache deletion operation does not fully complete. This can cause numerous unexpected errors including subsequent image cache failures. The Operator can now recover node status properly even after a cache deletion operation fails.

Fix: Issue where image import directory was removed by the cache clearing processs

Previously, when importing an image, the destination directory was created prior to extracting the image to a temporary location. Users were experiencing an issue where the extraction process did not complete prior to a caching process starting. Upon seeing that space is needed, empty directories would then be pruned in the cache. This would then result in the destination directory being removed, and when trying to move files into place, the process would fail due to the destination directory no longer existing. This was fixed in Orka 3.2.1, and the destination directory is now created at the time when the extracted files are moved into place.

Additionally, temporary files are now cleaned up during caching, resolving an issue where users were unable to cache additional images due to temporary files previously not being deleted as expected.

Orka node status logic update

In corner cases during cache operations, an incorrect/outdated node status is returned during the operation resulting in a failure. The logic was improved to prevent this condition.

Improved Orka node architecture detection

To ensure a node type is properly identified in mixed node environments, Intel vs. Apple Silicon (M-series) systems, the system info is checked with the node label to identify architecture type.

GitHub Actions Plugin null VM name issue

A recent upstream change to the ARC project caused previously used variables that Orka runners used to provide unique names to Orka runners return null VM names. The v1.1.5 GitHub Actions Orka plugin was changed to use natively created VM names from Kubernetes. The plugin also now implements jobId to identify jobs to improve support of concurrent/multiple runners.