#fdio-csit: FD.io CSIT prroject meeting

Meeting started by mackonstan at 15:01:55 UTC (full logs).

Meeting summary

    1. Jan Gelety (jgelety, 15:02:00)
    2. Tibor Frank (tifrank, 15:02:03)
    3. Peter Mikus (pmikus, 15:02:07)
    4. Vratko Polak. (vrpolak, 15:02:10)
    5. Dave Wallace (dwallacelf, 15:03:05)

  2. Agenda bashing (mackonstan, 15:03:23)
  3. DST starts in Europe this coming weekend back to normal time difference (mackonstan, 15:04:21)
  4. Physical labs, testbeds (mackonstan, 15:04:34)
    1. two 2n-skx testbeds out of production - tb22, tb23 - for use by Intel NPG guys t/shooting skx poor NDR performance (mackonstan, 15:05:05)

  5. Inputs from projects - VPP (mackonstan, 15:05:30)
    1. DaveW - need to sync up with Andrew when ready to do 1908.2 - will update CSIT (mackonstan, 15:06:11)
    2. Maciek - Benoit has a patch adding mlx dpdk into VPP build, so it will be including in VPP build (mackonstan, 15:07:03)
    3. Vratko - will require enabling vfio-pci driver for mlx NICs should do the trick if it works out of the box, if not then someone/Peter would need to add keyword to handle the specifics (mackonstan, 15:08:45)

  6. Inputs from projects - VSAP (mackonstan, 15:09:03)
    1. DaveW - no word from VSAP (mackonstan, 15:09:12)

  7. Inputs from projects - TSC (mackonstan, 15:09:25)
    1. Some continuing discussion about Scapy/GPLv2 - need to ask individual contributors (mackonstan, 15:09:56)
    2. Ray is driving this (mackonstan, 15:10:10)

  8. CNCF CNF Testbed (mackonstan, 15:12:50)
    1. Maciek still to reach out to Michael Pedersen (mackonstan, 15:13:02)

  9. Releases CSIT-2001 (mackonstan, 15:13:21)
    1. Jan - last two jobs on 3n-skx are running. This will finish the round of 10. (mackonstan, 15:14:13)
    2. rls to rls comparison tables got updated with root cause analysis notes (mackonstan, 15:16:14)
    3. https://docs.fd.io/csit/master/report/_static/vpp/performance-changes-3n-skx-2t1c-pdr.txt (mackonstan, 15:17:32)
    4. daily performance regressions progressions and failures are auto detected and info distributed over email via csit-report mailer (mackonstan, 15:27:12)
    5. and are also presented in graphs on trending pages: https://docs.fd.io/csit/master/trending/trending/ip4-2n-skx-xxv710.html (mackonstan, 15:28:15)
    6. https://docs.fd.io/csit/master/trending/trending/ip4-2n-skx-xxv710.html (mackonstan, 15:28:19)
    7. data available for download in csv format (mackonstan, 15:28:41)
    8. https://docs.fd.io/csit/master/trending/data/index.html (mackonstan, 15:28:54)
    9. currently the history duration is 3mths (mackonstan, 15:30:11)
    10. 1mth worth of history data compressed is ~1 GB; (mackonstan, 15:35:11)
    11. discussion about option to extend trending coverage to cover complete release period (mackonstan, 15:36:08)
    12. skx full tests - wait till skx issues resolved by/with Intel (mackonstan, 15:38:14)

  10. Releases CSIT-1908 (mackonstan, 15:38:42)
    1. Readiness of stable/1908 weekly coverage - Jan - weekly tests are running, new weekly job with mrr is running on oper branch (oper-rls1908-200325) on top of 1908.1, will be moving to the top of vpp stable/1908 tomorrow (mackonstan, 15:43:01)
    2. Jan will send email update (mackonstan, 15:43:10)

  11. Release CSIT-2005 (mackonstan, 15:43:34)
  12. Operations (mackonstan, 15:43:48)
  13. VPP code performance (mackonstan, 15:45:09)
    1. Peter - focusing on fixing vhost and vm issues, Mohsin helping on vpp side. (mackonstan, 15:45:54)
    2. Peter - AVF init failing after rebooting skx, t/shooting. (mackonstan, 15:46:17)
    3. Jan - one vxlan suite is failing, Juraj reported, already fixed. (mackonstan, 15:46:35)
    4. updated vpp native rdma code resulted in huge rate increase - 120 Mpps, currently max set by csit code for trex (mackonstan, 15:49:26)
    5. https://docs.fd.io/csit/master/trending/trending/ip4-2n-clx-cx556a.html (mackonstan, 15:49:29)
    6. Maciek - looks like `show run` papi is broken, stopped showing breakdown per worker (mackonstan, 15:52:48)

  14. Developments (mackonstan, 15:53:04)
    1. Govind - upgrading Taishan to -72 is failing (mackonstan, 15:54:16)
    2. Looking to push the update of CSIT-2001 report with Skx data tomorrow, after RCA for progressions and regressions is completed. (mackonstan, 16:02:11)


Meeting ended at 16:02:17 UTC (full logs).

Action items

  1. (none)


People present (lines said)

  1. mackonstan (46)
  2. collabot` (4)
  3. pmikus (1)
  4. vrpolak (1)
  5. jgelety (1)
  6. dwallacelf (1)
  7. tifrank (1)


Generated by MeetBot 0.1.4.