Skip to content

Enhance link event damping HLD#1488

Open
Ashish1805 wants to merge 1 commit intosonic-net:masterfrom
Ashish1805:aksingh_patch
Open

Enhance link event damping HLD#1488
Ashish1805 wants to merge 1 commit intosonic-net:masterfrom
Ashish1805:aksingh_patch

Conversation

@Ashish1805
Copy link
Copy Markdown
Contributor

  • Update the unit of config parameter.
  • Update the CLI format.

- Update the unit of config parameter.
- Update the CLI format.
@Ashish1805
Copy link
Copy Markdown
Contributor Author

@Ashish1805
Copy link
Copy Markdown
Contributor Author

Adding @Junchao-Mellanox for review.

@DavidZagury
Copy link
Copy Markdown

@Ashish1805 - Is there a PR for sonic-mgmt with tests that will covers this feature? Is there a plan to add such tests?

@Ashish1805
Copy link
Copy Markdown
Contributor Author

@Ashish1805 - Is there a PR for sonic-mgmt with tests that will covers this feature? Is there a plan to add such tests?

We dont have any sonic-mgmt test planned for the feature. Can you point to any sonic-mgmt tetss for any existing feature for reference? Thanks.

@DendroLabs
Copy link
Copy Markdown

Hi @Ashish1805 and @DavidZagury,

This PR has been inactive for over 2 years. I'm picking up the link event damping feature and driving it to completion.

@DavidZagury -- you asked about sonic-mgmt test coverage. I've drafted a comprehensive test plan with 38 test cases covering configuration, functional dampening behavior, CLI (show/clear), counters, monitor-only mode, syslog verification, edge cases, and scale testing on the vs platform. I'll include it when I open the implementation PRs.

I also plan to update the HLD with:

  • Industry-standard recommended defaults (Cisco/Juniper: half-life=5s, reuse=1000, suppress=2000, max-suppress=20s)
  • show interfaces dampening command specification
  • sonic-clear interfaces dampening command specification
  • YANG model additions for sonic-port.yang

Additionally, based on a review of operator complaints across Cisco, Juniper, and Arista forums, I'm proposing two enhancements that address the top industry pain points:

  1. Syslog on suppress/unsuppress -- The most common complaint across all vendors is that dampening operates silently. Syncd will emit WARNING on suppress and NOTICE on unsuppress.

  2. Monitor-only mode (algorithm="aied-monitor") -- RFC 7196 recommends a "Calculate But Do Not Damp" capability so operators can tune parameters without risking production outages. No vendor has implemented this. In monitor mode, penalties are calculated, STATE_DB is updated, and syslog messages are emitted, but events are NOT suppressed.

Will credit @Ashish1805 as original author of the HLD.

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.

3 participants