#fdio-meeting: FD.io CSIT project meeting

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

Meeting summary

    1. Tibor Frank (tifrank, 14:00:13)
    2. Peter Mikus (pmikus, 14:00:23)
    3. Vratko Polak. (vrpolak, 14:00:33)
    4. Fan Zhang (fan_zhang, 14:00:37)
    5. mackonstan (mackonstan, 14:00:38)

  1. Agenda bashing (mackonstan, 14:01:26)
  2. Inputs from projects - VPP (mackonstan, 14:02:12)
    1. Andrew: vpp v21.01.1 maintenance is almost out (mackonstan, 14:03:09)
    2. Maciek: Need to agree what and when should be CSIT tested in v21.01.1. (mackonstan, 14:04:02)

  3. Inputs from projects - VSAP (mackonstan, 14:04:25)
    1. No updates. (mackonstan, 14:04:54)

  4. Inputs from projects - TSC (mackonstan, 14:05:05)
    1. Vratko: re TSC update, there will be election for committer-at-large. Ed is creating a mailer for FD.io committers only to facilitate the process. (mackonstan, 14:06:07)
    2. This will be followed by TSC appointing the members. (mackonstan, 14:06:44)

  5. New performance tests - GTPU tests (mackonstan, 14:07:16)
  6. New functional Tests (mackonstan, 14:07:31)
    1. Dave Wallace (dwallacelf, 14:08:22)
    2. Fan: re flow "offload" patch - would like to know the status of adding E810 NICs to vpp_device 1n-clx testbeds. (mackonstan, 14:08:33)
    3. Peter: still working on it. (mackonstan, 14:08:52)
    4. Juraj LinkeŇ° (jlinkes, 14:10:21)
    5. Maciek: aside, will open the ticket tracking all new HW shipment status (one of the QSFP28 cables to be used for E810 in 1n-clx testbeds). (mackonstan, 14:12:31)
    6. Juraj: re IPsec SPD entry tests, separate patches for inbound and outbound, 1 patch merged, 2 outstanding. (mackonstan, 14:14:30)

  7. Releases CSIT-2106 (mackonstan, 14:15:49)
    1. Tibor: things looking quite well, almost all tests are finished. Only coverage 3n-skx, 2n-zn2, 3n-tsh are finishing, less than 50% remaining. (mackonstan, 14:17:11)
    2. Tibor: report status, some methodologies sections are missing, added TODO placeholders for methodology sections. (mackonstan, 14:18:30)
    3. Fan: content for gtpu test methodology and ipsec async crypto methodology is available, will push patches (mackonstan, 14:20:23)
    4. Maciek: here are the pointers to the right places in the report .rst files (mackonstan, 14:22:02)
    5. Maciek: gtpusw pointer (mackonstan, 14:22:16)
    6. https://git.fd.io/csit/tree/docs/report/introduction/methodology_tunnel_encapsulations.rst#n40 (mackonstan, 14:22:21)
    7. Maciek: ipsec async pointer (mackonstan, 14:22:36)
    8. https://git.fd.io/csit/tree/docs/report/introduction/methodology_ipsec.rst#n50 (mackonstan, 14:22:58)
    9. Tibor: everything else got merged, CSIT-2106 report generation job started 2hrs ago in rls2106 branch. Should be published within the next 2hrs. (mackonstan, 14:25:52)
    10. Vratko: two main regression RCAs under investigation: a) ipsec, b) nat44ed cps low scale (PDR results unstable). RCA markup would need to be added in the next maintenance version of csit-2106 report. (mackonstan, 14:31:38)
    11. Vratko: should mention performance issues with IPsec crypto async tests. (mackonstan, 14:34:28)
    12. https://gerrit.fd.io/r/c/vpp/+/30386 (fan_zhang, 14:34:32)
    13. Fan: IPsec crypto async performance tests failing is related to this patch not being merged (mackonstan, 14:36:28)
    14. https://gerrit.fd.io/r/c/vpp/+/30386 (mackonstan, 14:36:36)
    15. Maciek: looks like IPsec async crypto VPP code is broken. Need to add this to CSIT-2106 rls notes. (mackonstan, 14:39:15)
    16. Vratko: IPsec async crypto VPP, some tests pass but with low performance, likely to be related to the same issue. (mackonstan, 14:40:35)
    17. Vratko: will test the aforementioned patch, to see if issue is gone. (mackonstan, 14:40:55)
    18. Discussion about how to catch important feature-enabling and bug-fix VPP patches that are neglected, and making the feature broken. (mackonstan, 14:45:05)

  8. Release - CSIT-2101.1 (mackonstan, 14:46:07)
    1. VPP v21.01.1 is almost out (mackonstan, 14:46:26)
    2. Need to decide about CSIT coverage ... (mackonstan, 14:46:42)
    3. Dave: last time we picked up a subset of tests, and run lower number of iterations for them. (mackonstan, 14:47:40)
    4. Dave: it was more a smoke-test than anything else. (mackonstan, 14:48:11)
    5. Vratko: proposal - use existing iterative job_specs, just run a single iteration, and see if results fall within the avg+/-stdev of the previous full report run. Would need to add a new type of comparison table. (mackonstan, 14:50:25)
    6. Maciek, Vratko: no coverage tests. (mackonstan, 14:50:44)
    7. Dave: if results fall outside the avg+/-stdev, then do RCAs. (mackonstan, 14:51:51)
    8. Tibor, Maciek: will take a closer look at scheduling CSIT-2101.1 tests. (mackonstan, 14:53:02)
    9. Tibor: MRR trending tests run at reduced cadence of twice a week at the moment. Need to analyse when we are returning to daily cadence. (mackonstan, 14:54:20)

  9. Physical Infrastructure (mackonstan, 14:56:31)
    1. Maciek, Peter: hsw servers are to be fully decommissioned. (mackonstan, 14:57:32)

  10. Virtual Infrastructure, CI/CD (mackonstan, 14:57:46)
    1. Dave, virtual infra and CI infra are stable. Vanessa is testing the Jenkins upgrade in the sandbox, assuming it's stable, the production maintenance is scheduled for 19-Jul. (mackonstan, 14:58:32)

  11. Trending (mackonstan, 14:59:57)
    1. https://gerrit.fd.io/r/c/vpp/+/32878/7#message-51b619758963db4106d55d3da07300ddbe6928b6 <-- scale tests fail in master (vrpolak, 15:00:55)
    2. Vratko, CSIT ip4 and ip6 scale test are failing as VAT functionality that is used has been removed from VPP code as part of refactor. (mackonstan, 15:00:58)

Meeting ended at 15:01:57 UTC (full logs).

Action items

  1. (none)

People present (lines said)

  1. mackonstan (58)
  2. collab-meetbot` (6)
  3. vrpolak (2)
  4. fan_zhang (2)
  5. tifrank (1)
  6. pmikus (1)
  7. dwallacelf (1)
  8. jlinkes (1)

Generated by MeetBot 0.1.4.