#fdio-csit: CSIT weekly 2019-12-04

Meeting started by vrpolak at 15:00:15 UTC (full logs).

Meeting summary

    1. Peter Mikus (pmikus, 15:01:02)
    2. Vratko Polak (vrpolak, 15:02:00)

  2. Agenda Bashing (vrpolak, 15:02:13)
    1. Juraj Linkeš (jlinkess, 15:02:24)
    2. Tibor Frank (tifrank, 15:03:12)

  3. Physical labs, testbeds (vrpolak, 15:06:25)
    1. Arm build server (1/3) power cycled, hard drive failure on one s5ci node, otherwise looks good. (vrpolak, 15:08:25)
    2. Outage notice for Jenkins.plugins update. (vrpolak, 15:08:45)
    3. Dave Wallace (dwallacelf, 15:10:56)
    4. Arm docker images need to be updated for Python 3 to work correctly, pull requests: 1 ready, 1 in progress. (vrpolak, 15:11:03)

  4. operations (vrpolak, 15:14:13)
    1. There was a breakage after Python3 migration, due to "ply" package not available on virtualenv (even when installed on system). Fixed quicly, but root cause still under investigation. (vrpolak, 15:15:07)

  5. Inputs from projects (vrpolak, 15:15:55)
    1. No news from VPP, TSC. (vrpolak, 15:16:08)

  6. Developments (vrpolak, 15:21:50)
    1. Testbeds: Will need a way to limit jobs with Mellanox tests to run on testbeds that have them (one testbed does not yet). We will work on it after other Mellanox related changes are merged. (vrpolak, 15:22:50)
    2. Performance trending and analytics: New graph organization is merged, results will be visible soon. (vrpolak, 15:25:00)
    3. API functional test coverage: Next patches submitted, still under review. (vrpolak, 15:29:59)
    4. Question about bidirectional traffic. Answer: use the same type as in the preformance test. If functional test fail, we can identify a Bug and fix it. (vrpolak, 15:35:07)
    5. Gcov coverage: More a metric for VPP, but if they want we can create a weekly job for that. (vrpolak, 15:39:57)
    6. Improvements to test teardown: Use gdb "script" to process and print information useful to developers. (vrpolak, 15:43:08)

  7. TRex (vrpolak, 15:49:02)
    1. Vratko thinks there is a bug in TRex that occasionally increases the time Trex takes to send all the packets, effectively leading to MRR resylts bigger than the real performance. But Vratko also has trouble to gather enough data to compare two TRex versions, so the question is still open. (vrpolak, 15:50:36)
    2. Question about triggering a specific performance test. The full answer will be on csit-dev mailing list. (vrpolak, 15:56:21)


Meeting ended at 15:56:58 UTC (full logs).

Action items

  1. (none)


People present (lines said)

  1. vrpolak (22)
  2. collabot` (3)
  3. jlinkess (1)
  4. pmikus (1)
  5. dwallacelf (1)
  6. tifrank (1)


Generated by MeetBot 0.1.4.