#fdio-csit: FD.io CSIT project meeting

Meeting started by pmikus at 14:02:09 UTC (full logs).

Meeting summary

    1. Peter Mikus (pmikus, 14:02:18)
    2. Tibor Frank (tifrank, 14:02:28)

  2. Agenda bashing (pmikus, 14:03:48)
    1. Vratko Polak. (vrpolak, 14:05:00)

  3. Physical labs, testbed (pmikus, 14:05:07)
    1. Agreed with Maciek that we should have at least 10x SSD spares available in stock for quick replacement (pmikus, 14:06:57)
    2. Mail to be sent to TSC/Ray Kinsella/Trishan with current status and plan for spares (pmikus, 14:09:27)
    3. Functional and driver testing: Xeon Skx, Arm. - no issues. (pmikus, 14:10:16)

  4. Inputs from projects (pmikus, 14:10:56)
    1. https://wiki.fd.io/view/TSC/Relicensing_Procedure (vrpolak, 14:11:34)
    2. TCS: Vratko reported communication about SSD failures, Licensing issue - procedure established with lawyers link above from Vratko (pmikus, 14:12:01)
    3. Not clear the order of steps due to running out of time on call. Will follow later. (pmikus, 14:13:16)
    4. Dave Wallace (dwallacelf, 14:13:21)
    5. VPP: Nothing new (pmikus, 14:13:54)
    6. VSAP: Nothing new (pmikus, 14:14:04)
    7. CNCF CNF Testbed - Nothing new (pmikus, 14:14:25)

  5. Releases (pmikus, 14:14:40)
    1. CSIT-2001 Release, 2n-skx, 3n-skx testing status - Jobs finished, data collected, Tibor processed in PAL to new comparison tables (pmikus, 14:15:43)
    2. CSIT-1908.x Currently 2n-skx are finished, 3n-skx few jobs finished and rest is blocked due to SSD failures on both 3n-skx, waiting for RMA's (see above). (pmikus, 14:17:30)
    3. Vratko: 1908.x weekly mrr working, l2patch failing. (pmikus, 14:19:26)
    4. Currently maintenance releases are released prior testing (unlike stable releases with RCx approach) (pmikus, 14:20:20)
    5. Plan is to use weekly testing (pmikus, 14:20:55)
    6. CSIT-2005 Release (pmikus, 14:21:09)
    7. CSIT-2005 Release - new Virtio VPP native tests, increase coverage of functional tests (ipsec, srv6...), vswtich-less tests (pmikus, 15:14:03)
    8. VPP-2001.1 release prior RC of stable 2005, Dave asking for capacity. Peter reply that should be possible based on 1908.1 experience and limited amount of tests (pmikus, 15:15:38)
    9. ACTION: Finish all work on 1908.1 test lists. Create 2001.1 test list from 1908.1 analogues (in gerrit), add few interesting tests (Load Balancing), review and merge (pmikus, 15:17:27)

  6. Operations (pmikus, 15:17:38)
    1. LFN FD.io CI/CD infra: Jenkins, no outages spotted, one hiccup spotted by Jan when job gets stuck in init phase, closer monitoring not occured anymore yet (pmikus, 15:18:46)
    2. Jenkins jobs: VPP, CSIT, trending, one failure daily jobs for DNV (did not repeat), vpp-verify-centos one failure with compiling (timestamps issue, NTP related?). (pmikus, 15:19:03)
    3. Centos CI - currrently not supported in CSIT (pmikus, 15:19:11)
    4. CSIT committers - PTL will follow and update the list (pmikus, 15:19:21)

  7. VPP code performance (pmikus, 15:19:27)
    1. VHOST test fixed, AVF init - VPP tickets opened VPP-1858, no progress, IPSEC HW failure - investigation in progress, few sporadic failure (ip4, ip6) (pmikus, 15:20:26)
    2. Regression on IP6 scale tests, tests are passing but performance is low. Issue related to VAT (PAPI SCALE is not affected). Discussion about support of VAT or merge PAPI SCALE. (pmikus, 15:21:08)

  8. Developments (pmikus, 15:21:14)
    1. Framework: PAPI scale in CSIT - CSIT is deeply interacting with VPP backend (PAPI threads), Vratko wants to keep modyfing running VPP code via CSIT. Peter: Would be good to implement the patch/fix also into VPP to benefit for everyone outside CSIT. Vratko: Benefits are that having CSIT code allows us to do bisecting, on the other side implementing fixes in VPP would mean to support multiple code (pmikus, 15:23:41)
    2. Dave proposed to discuss with Ole and to have all the fixes in VPP as well. Peter: Having temporary solution is fine if it will be followed with proper patch. (pmikus, 15:24:26)
    3. Vratko sees an action plan to split change into 4 parts including current patch, vpp patch and two other to have permanents fix. No other objections. (pmikus, 15:24:58)

  9. TRex (pmikus, 15:25:05)
    1. Peter working on calibrating the mellanox NIC with T-rex. Peter to follow up findings and consult with T-Rex-dev. Peter to implement DPDK testpmd baseline mellanox tests as a start point (pmikus, 15:26:41)
    2. Vratko reported that tests with randomizing flows are misbehaving - fix proposed with not "fixed" seed (pmikus, 15:26:55)


Meeting ended at 15:27:02 UTC (full logs).

Action items

  1. Finish all work on 1908.1 test lists. Create 2001.1 test list from 1908.1 analogues (in gerrit), add few interesting tests (Load Balancing), review and merge


People present (lines said)

  1. pmikus (42)
  2. collabot` (4)
  3. vrpolak (2)
  4. tifrank (1)
  5. dwallacelf (1)


Generated by MeetBot 0.1.4.