#fdio-meeting: FD.io CSIT Project Meeting

Meeting started by mackonstan at 13:59:11 UTC (full logs).

Meeting summary

    1. mackonstan (mackonstan, 13:59:16)
    2. Vratko Polak. (vrpolak, 14:02:05)

  1. Agenda bashing (mackonstan, 14:02:29)
    1. Peter Mikus (pmikus, 14:02:32)

  2. Inputs from projects - VPP (mackonstan, 14:04:29)
    1. Tibor Frank (tifrank, 14:04:53)
    2. 36352: ipsec: change wildcard value for any protocol of spd policy | https://gerrit.fd.io/r/c/vpp/+/36352 (vrpolak, 14:06:36)
    3. Fan, still working on getting above patch merged, but currently targeting next rls 2210 (mackonstan, 14:07:33)

  3. Inputs from projects - TSC (mackonstan, 14:08:29)
    1. ICX-D Wireguard VPP Perf whitepaper: https://builders.intel.com/docs/networkbuilders/intel-avx-512-and-intel-qat-accelerate-wireguard-processing-with-intel-xeon-d-2700-processor-technology-guide-1647024663.pdf (fan_zhang, 14:12:43)
    2. Maciek, on a hook to provide updated HW lab CAPEX spend proposal to TSC. Currently mainly looking at MLX 100 GbE cards (mackonstan, 14:13:04)
    3. Fan, proposed we look at ICX-D with QAT1.8 that offers VPP running with up to 50 Gbps per core for wireguard (mackonstan, 14:13:53)
    4. Fan, Maciek, evaluate the merit of adding ICX-D into the FD.io CSIT labs (mackonstan, 14:14:12)

  4. New or being refactored tests (mackonstan, 14:14:30)
    1. Fan, E810 flow "offload" tests - device tests, Marvin is taking over, targeting 2210 rls cycle for completing this (mackonstan, 14:16:10)
    2. Fan, E810 flow "offload" tests - perf tests, all merged now, if possible would like to have them added to trending tests (mackonstan, 14:19:39)
    3. Tibor, should have capacity with two of 3n-icx to run those tests, need to know the quantity of tests and their approx. duration (mackonstan, 14:20:36)
    4. Fan, will push a patch with updated job_specs, add Tibor and Maciek as reviewers (mackonstan, 14:21:29)

  5. Report CSIT-2206 (mackonstan, 14:22:45)
    1. Tibor, jenkins was down for 3 days Sat - Mon, now it's back, but introduced delay to RC2 testing schedule (mackonstan, 14:23:18)
    2. Tibor, despite above, still on target to complete rc2 testing phase by eow, report with 1st data eta tomorrow (mackonstan, 14:24:06)
    3. Maciek, nobody knows the reason for jenkins down, waiting for Dave Wallace and Vanessa to provide post mortem readout (mackonstan, 14:25:05)
    4. Vratko, on regressions, already checked rc1, nothing major spotted, awaiting rc2 results (mackonstan, 14:26:18)
    5. Vratko, so far only checked Xeons (mackonstan, 14:26:33)
    6. Maciek, Viliam, will work on updating known issues in master report, capturing the backlog from 22.02 and the new ones (mackonstan, 14:28:51)
    7. Maciek, target to point it to vpp community before the next call on Tue 28-jun (mackonstan, 14:29:20)
    8. Tibor, report to do list, see email to csit-dev (mackonstan, 14:30:17)
    9. https://lists.fd.io/g/csit-dev/message/4522 (mackonstan, 14:30:19)
    10. Maciek, 100 GbE tests, let's make sure CVL flow tests are covered on them, and phy testbeds static content is update (mackonstan, 14:33:15)
    11. Maciek, Peter, Vratko, evaluate the multi-core speedup and verify that we see close to linear speedup (mackonstan, 14:37:27)
    12. Maciek, re telemetry, hit an issue with VPP cli `wait` command, that makes some of the csit telemetry results incorrect. Peter is working on the longer-term resolution. (mackonstan, 14:40:29)
    13. Peter, new ICX testbeds are coming online, all 5 racks are now completely full. (mackonstan, 14:42:41)
    14. Peter, new SNR testbed, awaiting some further assistance from the Intel team to make use of IPMI for console for remote operation (mackonstan, 14:44:28)

  6. Physical Infrastructure (mackonstan, 14:45:03)
  7. Virtual Infrastructure, CI/CD (mackonstan, 14:45:39)
    1. Juraj, aside, working with Arm team to work out the large discrepancy ( 5x ) between PDR and NDR rates (mackonstan, 14:47:49)

  8. Environment, Framework (mackonstan, 14:48:47)
    1. Vratko, telemetry issue is another example of CSIT using CLI vs relying on APIs (mackonstan, 14:51:02)
    2. Peter, the point is that vpp community needs to find resources to provide a complete coverage of the telemetry that CSIT is using, for now we only have access to CLI (mackonstan, 14:54:27)

  9. Trending (mackonstan, 14:55:04)
    1. Tibor, added e810 NIC 100 GbE to trending (mackonstan, 14:56:00)
    2. Viliam, see for trending failing tests in email: CSIT failing perf tests for week 25 (06/13 – 06/19) (mackonstan, 14:58:06)
    3. https://lists.fd.io/g/csit-report/message/5784 (mackonstan, 14:58:19)

Meeting ended at 14:58:57 UTC (full logs).

Action items

  1. (none)

People present (lines said)

  1. mackonstan (40)
  2. collab-meetbot` (3)
  3. vrpolak (3)
  4. pmikus (1)
  5. tifrank (1)
  6. fan_zhang (1)

Generated by MeetBot 0.1.4.