NVIDIA / nodewright / 41553710654

84%

Build Type

push

github

Committed by web-flow

Commit Message

fix(operator): hold a merged interrupt until every package sharing it reaches its interrupt stage (#949)

When several packages with interrupts run on a node, their interrupts merge
into one owned by the first package in name order with the highest-ranked
interrupt. The owner fired it as soon as no package Job was running on the
node, which includes the gap between another package's apply and config Jobs.
That package's config then ran after the reboot, and it was either skipped
against an interrupt that had already run or rebooted the node again.

The owner now also waits until every other package with an interrupt has
reached its interrupt stage, where it is marked skipped. A package that is
erroring, or mid-uninstall, does not hold it.

On a serial node the pass stops at a package that is waiting, so it would
stop at the held owner and never reach the packages it waits for. The held
owner runs nothing meanwhile, so the pass moves past it, as it already does
past a skipped package. Waiting for the skip rather than for the end of a
package's config also means the skip is written before the interrupt fires:
a serial pass stops at the owner once it fires and would never write it.

Closes #934

Signed-off-by: Alex Yuskauskas <ayuskauskas@nvidia.com>

Coverage Stats

25 of 25 new or added lines in 1 file covered. (100.0%)

3 existing lines in 1 file now uncovered.

10472 of 12469 relevant lines covered (83.98%)

7.68 hits per line