[CSIT-1974] cx7: Persistent losses in testpmd 4c imix test

[vvalderrv]

Description

Other NICs appear unaffected, including cx6dx on 3n-alt, but cx7 is affected in each run. The losses of around 10% seem constant within test run, at least at lower loads. For example here [0] the 9001pps trials have the same loss of 1100 packets, but run #2 has 1135 packets and #3 has 1130. Maybe one of 8 RX (or TX) queues does not work? We are not running 4c 1518B test, but 4c 64B is not affected.

[0] https://s3-logs.fd.io/vex-yul-rot-jenkins-1/csit-dpdk-perf-report-iterative-2410-2n-icx/1/log.html.gz#s1-s1-s1-s3-t9-k2-k5-k14

Assignee

Unassigned

Reporter

Vratko Polak

Comments

• vrpolak (Wed, 20 Nov 2024 12:34:41 +0000): As similar CSIT-1943 got fixed, probably by DPDK version bump, it is possible the next release this will be fixed also for CSIT DPDK tests.
• vrpolak (Wed, 6 Nov 2024 12:55:00 +0000): Rls2406 was already affected: [1].

[1] https://s3-logs.fd.io/vex-yul-rot-jenkins-1/csit-dpdk-perf-report-iterative-2406-2n-icx/1/log.html.gz#s1-s1-s1-s3-t9-k2-k5-k14

Original issue: https://jira.fd.io/browse/CSIT-1974

[vvalderrv]

As similar CSIT-1943 got fixed, probably by DPDK version bump, it is possible the next release this will be fixed also for CSIT DPDK tests.

[vvalderrv]

Rls2406 was already affected: [1].

[1] https://s3-logs.fd.io/vex-yul-rot-jenkins-1/csit-dpdk-perf-report-iterative-2406-2n-icx/1/log.html.gz#s1-s1-s1-s3-t9-k2-k5-k14