Skip to content

refactor: enhance scaling configuration and resource management in tests - #141

Open
karimra wants to merge 5 commits into
mainfrom
test-suite-cont
Open

refactor: enhance scaling configuration and resource management in tests#141
karimra wants to merge 5 commits into
mainfrom
test-suite-cont

Conversation

@karimra

@karimra karimra commented Aug 11, 2026

Copy link
Copy Markdown
Collaborator
  • Updated test.mk to include new environment variables for scaling configuration: SCALE_REPLICAS, SCALE_CPU_REQUEST, SCALE_CPU_LIMIT, SCALE_MEMORY_REQUEST, and SCALE_MEMORY_LIMIT.
  • Refactored integration tests to utilize new helper functions for managing subscription paths and outputs, improving test reliability and clarity.
  • Adjusted baseline YAML to dynamically reference CPU and memory resource requests and limits, enhancing flexibility in scaling tests.

- Updated test.mk to include new environment variables for scaling configuration: SCALE_REPLICAS, SCALE_CPU_REQUEST, SCALE_CPU_LIMIT, SCALE_MEMORY_REQUEST, and SCALE_MEMORY_LIMIT.
- Refactored integration tests to utilize new helper functions for managing subscription paths and outputs, improving test reliability and clarity.
- Adjusted baseline YAML to dynamically reference CPU and memory resource requests and limits, enhancing flexibility in scaling tests.
@cloudflare-workers-and-pages

cloudflare-workers-and-pages Bot commented Aug 11, 2026

Copy link
Copy Markdown

Deploying gnmic-operator with  Cloudflare Pages  Cloudflare Pages

Latest commit: f3192b2
Status: ✅  Deploy successful!
Preview URL: https://927df36b.gnmic-operator2.pages.dev
Branch Preview URL: https://test-suite-cont.gnmic-operator2.pages.dev

View logs

karimra and others added 4 commits August 11, 2026 18:30
gNMIc can ACK config/apply before every target re-subscribes; ApplyCache then
short-circuited the same plan for 5m and left paths=[] long enough to flake
Sub005/Sub010. Verify each hash with a second POST, and wait on fleet state
atomically in those tests.

Co-authored-by: Cursor <cursoragent@cursor.com>
TestOut006, TestTarget003, and TestOut003 sampled their invariant right
after triggering a config apply or waiting on Cluster status, both of
which can return before the actual gNMIc streams/Service exist. Wait for
the settled state first so the assertion window doesn't start mid-reconnect.

Co-authored-by: Cursor <cursoragent@cursor.com>
gNMIc's config/apply can ACK before every Subscribe stream has
re-established, and rarely one never comes back on its own, leaving a
target stuck with paths=[] indefinitely. Track whether an applied hash
has been confirmed by a later reconcile; if it's still unverified after
a 30s grace period, requeue and re-POST the same config once to kick
the stream loose.

The delay matters: firing the repost immediately (an earlier attempt)
doubled every apply back-to-back and made pods work harder to
reconnect right when they were already struggling, making CI flakes
worse instead of better. 30s clears both unrelated reconcile churn and
tests' own "stays stable" assertion windows.

Co-authored-by: Cursor <cursoragent@cursor.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant