#fdio-csit: 04-Oct FD.io CSIT Project Meeting

Meeting started by mackonstan1 at 14:01:30 UTC (full logs).

Meeting summary

    1. Jan Gelety (jgelety, 14:01:54)
    2. Peter Mikus (pmikus1, 14:01:55)
    3. Samuel Elias (selias, 14:01:58)
    4. Tibor Frank (tifrank, 14:02:15)

  2. Agenda bashing (mackonstan1, 14:08:10)
  3. Inputs from LF and FD.io projects: VPP (mackonstan1, 14:08:33)
    1. Peter: additional logging being added to catch cases when VPP doesn't take VAT calls and/or crashes (mackonstan1, 14:09:17)
    2. dwallace_: make test added API postmortem files and core dump files (mackonstan1, 14:10:46)
    3. mackonstan1: it makes sense to do add same/similar to CSIT - func-virl and perf (mackonstan1, 14:12:25)
    4. Dave Wallace (dwallace_, 14:13:20)
    5. ACTION: Peter: will explore the size of core dumps and report back. (mackonstan1, 14:16:38)
    6. ACTION: Maciek, Dave, Peter: drive implementation into CSIT. (mackonstan1, 14:17:03)

  4. Honeycomb update (mackonstan1, 14:17:53)
    1. Samuel: packer scripts do not seem to work anymore for building CSIT VM ubuntu images (mackonstan1, 14:19:09)
    2. need to investigate the problem (mackonstan1, 14:19:51)

  5. VPP project: tracking VPP gerrit rechecks (mackonstan1, 14:20:58)
    1. no updates (mackonstan1, 14:21:02)
    2. make test seem to have been the main problem - culprit tests removed (mackonstan1, 14:21:54)

  6. Operational status - LF infra stability and performance - all good (mackonstan1, 14:22:31)
  7. rls1707 experience (mackonstan1, 14:24:51)
    1. VPP Performance regression for l2bdscale tests fixed in vpp 17.07.1 - fixed and verified (mackonstan1, 14:25:10)
    2. ACTION: mackonstan1 to send email with detailed breakdown of feature that was responsible for this regression, and slight perf degradation in 17.07 vs. earlier. (mackonstan1, 14:25:47)

  8. rls1710 plan (mackonstan1, 14:25:59)
    1. https://wiki.fd.io/view/CSIT/csit1710_plan (mackonstan1, 14:27:02)
    2. Peter: GoVPP and vnf-agent integration - code is complete, latest issue resolved, but code still to be merged and exercised in LF CSIT testbed, need to complete jjb review for that. (mackonstan1, 14:28:19)
    3. still waiting for open-source version of sfc_controller, vs. docker image, from ligato project.. (mackonstan1, 14:28:49)
    4. SRv6 performance - list of L1 and L2 KWs and traffic profiles - need a review.. (mackonstan1, 14:29:40)
    5. Tibor: CSIT Data analytics and presentation enhancements - PAL - code complete. finishing LLD, completing code cleanup and then ready for review. will announce on the list. (mackonstan1, 14:32:10)
    6. CSIT AutoConfig - no update (mackonstan1, 14:33:01)
    7. Centos7 VIRL tests - no udpate (mackonstan1, 14:33:15)
    8. so are on target to pull the branch on 2017-10-11 (mackonstan1, 14:36:08)
    9. next week creating csit branch for rls1710, but also oper branch for vpp stable/1710 (mackonstan1, 14:37:02)
    10. Tibor: expecting PAL to get merged early next week.. and from there on work on static content of the report can start in earnest (mackonstan1, 14:41:07)

  9. Performance tests: SW cryptodev tests (mackonstan1, 14:42:33)
    1. doing this in VIRL is an easy change: just update startup.conf (mackonstan1, 14:43:01)
    2. doing this in performance testbed is also an easy change: just update startup.conf (mackonstan1, 14:44:05)
    3. ACTION: Maciek, Sergio, Jan - work SW cryptodev tests into VIRL tests. (mackonstan1, 14:46:24)
    4. ACTION: Maciek, Sergio, Peter - work SW cryptodev tests into perf tests. (mackonstan1, 14:46:37)
    5. HW l2fwd-cryptodev - delayed - still delayed (mackonstan1, 14:48:22)
    6. Jan: security-groups, mac-ip - issues with NDR for mac-ip in multi-core (mackonstan1, 14:49:54)
    7. VPP functional tests - virl (mackonstan1, 14:50:15)

  10. VPP functional tests - virl (mackonstan1, 14:50:19)
    1. CSIT VIRL vs. VPP make-test - functional test coverage delta: dpdk-virtio, native vhost-user, crypto-dev (to be added), .. (mackonstan1, 14:55:55)
    2. continued: honeycomb, … (mackonstan1, 14:57:02)
    3. ACTION: mackonstan1: email to csit-dev to collect all points on CSIT VIRL vs. VPP make-test, incl. current func test delta (as of july…), drive consensus within csit-dev, and then readout to vpp-dev (mackonstan1, 14:57:53)

  11. Test environment updates, adding test equipment: Netgate - CPE test cases (mackonstan1, 14:59:54)
  12. NestedVM refresh - csit-dev @ lists (mackonstan1, 15:02:51)
    1. continue on the list (mackonstan1, 15:04:27)

  13. VIRL hosts upgrade - tb4-virl3 - status - no progress so far; (mackonstan1, 15:05:08)
  14. ARM additions to CSIT lab (mackonstan1, 15:05:18)
    1. mackonstan1: proposal is to follow the recommendation from vpp-dev call yesterday, and start with running vpp make-test functional test on ARM platform(s) that will be installed in LF labs (mackonstan1, 15:06:27)
    2. mackonstan1: once vpp make test is running in LF, the natural next step will be to consider CSIT functional tests in virtual/containerized? environment and performance tests (mackonstan1, 15:07:39)
    3. adding ARM related points as standing CSIT call agenda item (mackonstan1, 15:08:02)


Meeting ended at 15:08:05 UTC (full logs).

Action items

  1. Peter: will explore the size of core dumps and report back.
  2. Maciek, Dave, Peter: drive implementation into CSIT.
  3. mackonstan1 to send email with detailed breakdown of feature that was responsible for this regression, and slight perf degradation in 17.07 vs. earlier.
  4. Maciek, Sergio, Jan - work SW cryptodev tests into VIRL tests.
  5. Maciek, Sergio, Peter - work SW cryptodev tests into perf tests.
  6. mackonstan1: email to csit-dev to collect all points on CSIT VIRL vs. VPP make-test, incl. current func test delta (as of july…), drive consensus within csit-dev, and then readout to vpp-dev


Action items, by person

  1. mackonstan1
    1. mackonstan1 to send email with detailed breakdown of feature that was responsible for this regression, and slight perf degradation in 17.07 vs. earlier.
    2. mackonstan1: email to csit-dev to collect all points on CSIT VIRL vs. VPP make-test, incl. current func test delta (as of july…), drive consensus within csit-dev, and then readout to vpp-dev
  2. UNASSIGNED
    1. Peter: will explore the size of core dumps and report back.
    2. Maciek, Dave, Peter: drive implementation into CSIT.
    3. Maciek, Sergio, Jan - work SW cryptodev tests into VIRL tests.
    4. Maciek, Sergio, Peter - work SW cryptodev tests into perf tests.


People present (lines said)

  1. mackonstan1 (51)
  2. collabot (4)
  3. jgelety (1)
  4. dwallace_ (1)
  5. selias (1)
  6. pmikus1 (1)
  7. tifrank (1)


Generated by MeetBot 0.1.4.