NVIDIA / nodewright / 36905879430
83%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 59
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 82.027% (+0.06%) from 81.967%
36905879430

push

github

web-flow 
test(agent): assert what gracefulShutdown guarantees in sigterm_grace (#713)

sigterm_grace asserted that progress held exactly one "start" and one "end"
after the package pod was deleted mid-step. That is a transient state.
Deleting the pod makes the Job count the attempt as failed whatever the agent
exits with, so a replacement pod runs the apply stage again, and the
shellscript steps declare idempotence: true, which both agents read as "run
regardless of the completion flag". apply.sh runs twice by design, and the
count passed in CI only by being taken within seconds of the first "end",
before the replacement pod appended its own pair. On a real cluster the
settled file had two pairs.

The test now lets the package converge, then checks the property
gracefulShutdown actually guarantees: the attempt that received SIGTERM
finished. apply.sh tags each progress line with an attempt id, and one awk on
the node checks that the id on the first "start" also appears on an "end" and
that every start has an end, printing the counts when it fails so
check_node.sh shows why. Counts or timing alone cannot catch the #672
regression: a killed first attempt leaves "start A, start B, end B", whose
only end still comes a full sleep after the first start.

Two docs record what running both agents on one node during release testing
showed. agent/RELEASE_NOTES.md gains the v7 log-file differences the
[out]/[err] entry did not cover: stderr is merged into the step's .log at
mode 0600, where the Python agent wrote a 0644 .log and a .log.err that
neither agent reaps, and interrupt logs are one <type>-<timestamp>.log per
interrupt run, reaped to five, where the Python agent wrote one per operation
and never reaped them; the directory is the same. The Breaking entry now
claims unchanged log directories rather than unchanged log paths.
docs/architecture/lifecycle.md said deletion cleans the state annotations off
nodes and removes the runtime-required taint; the finalizer keep... (continued)

9516 of 11601 relevant lines covered (82.03%)

7.62 hits per line

Jobs
ID Job ID Ran Files Coverage
1 36905879430.1 59
82.03
GitHub Action Run
Source Files on build 36905879430