Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
fix 7260 headroom pool watermark test failure (#15536)
What is the motivation for this PR? observed consistent headroom wartermark test failure on 7260 and it's known issue of test script, as below RCA: RCA: summarize test step first: PTF send lots of pkt to multiple src ports to fill multiple PG's share buffer PTF send one or a few pkts to multiple src ports to trigger pfc on multiple PG check watermark before test headroom's watermark PTF send pkt to multiple src port to consum headroom pool, and test if watermark changes as expected after step2, already send 20 pkts into headroom to trigger PFC on 10 src ports (20 PG) but, so far, "upper_bound" value is static hardcode "2 * margin + 1", didn't consider headroom pool consumption in step2. since we use dynamically threshold calculating, it can get accurate threshold value, we pretty sure the headroom pool consumption equal "pgs_num" in step2. so I change "upper_bound" value to "2 * margin + self.pgs_num", and it pass the tests. How did you do it? change "upper_bound" value to "2 * margin + self.pgs_num" How did you verify/test it? this change already verified in MSFT nightly for 202305 and 202311 branch, just commit to github. Any platform specific information? this change is dedicated to below platform and topology: if (hwsku == 'Arista-7260CX3-D108C8' and self.testbed_type in ('t0-116', 'dualtor-120')) \ or (hwsku == 'Arista-7260CX3-C64' and self.testbed_type in ('dualtor-aa-56', 't1-64-lag')): upper_bound = 2 * margin + self.pgs_num if other platform and topology have hit similar issue, can add affected platform and topo to above condition checking. Note: for generic fix, qos refactor project will covert.
- Loading branch information