#fdio-csit: FD.io CSIT project meeting

Meeting started by mackonstan at 14:03:06 UTC (full logs).

Meeting summary

    1. Dave Wallace (dwallacelf, 14:03:16)

  2. Agenda bashing (mackonstan, 14:03:19)
    1. Juraj Linkeš (jlinkes, 14:03:26)
    2. Vratko Polak. (vrpolak, 14:03:29)
    3. Peter Mikus (pmikus, 14:04:16)

  3. FD.io CSIT lab infrastructure (up to 15min) (mackonstan, 14:04:27)
    1. Maciek: new HW should have arrived, at least partially (mackonstan, 14:06:14)
    2. Mohammed: in fact all HW arrived, servers with motherboards, Processors, NICs (mackonstan, 14:06:37)
    3. Mohammed: build servers already racked and installed (mackonstan, 14:07:08)
    4. Mohammed: awaiting low level design for configuration the testbeds (mackonstan, 14:07:30)
    5. Mohammed: 14 x clx-8280, 6 x 6252N (mackonstan, 14:10:41)
    6. Maciek: proposal to use 6252N on SUTs for 2n-clx, for consistency, vs. mixing 6252N and 8280 on SUTs. (mackonstan, 14:16:40)
    7. Vratko: +1 (mackonstan, 14:16:45)
    8. Mohammed: number of questions re Linux and disk partitioning/RAID. Will send email to the list. (mackonstan, 14:22:52)
    9. Mohammed: failed processors, skx, need to confirm RMA procedure. Will send email to Trishan, Maciek, Ray Kinsella. (mackonstan, 14:23:39)
    10. Maciek: will push patch with proposed clx testbed setup to gerrit and announce on csit-dev (mackonstan, 14:24:48)

  4. Inputs from LFN and FD.io projects (mackonstan, 14:24:53)
  5. VPP (mackonstan, 14:25:09)
    1. Dave: vpp v19.01.3 published last Friday. (mackonstan, 14:25:43)
    2. Peter: heard from Jan that csit-1901.3 tests are complete. Report to be generated. (mackonstan, 14:26:19)
    3. Dave: vpp v19.04.2 is to be published by eow. (mackonstan, 14:26:40)
    4. Dave: v19.08 f0 milestone is next Wed 31-Jul (mackonstan, 14:27:07)
    5. The name of the new gating job is: vpp-csit-verify-api-crc-{stream} (vrpolak, 14:30:09)
    6. Dave: vpp call yesterday, discussed and agreed to start running vpp-csit-verify-api-crc*, with voting, job starts on 30-Jul, if false positives handle with manual override (mackonstan, 14:30:35)

  6. HC (mackonstan, 14:31:59)
    1. still open is how to handle orphaned L1 KWs used by HC, as HC tests got now out of sync with VPP master and not running - discussion on the list. (mackonstan, 14:34:08)
    2. Maciek to start thread on csit-dev (mackonstan, 14:34:42)

  7. TSC (mackonstan, 14:35:04)
    1. no updates (mackonstan, 14:35:13)

  8. Operational status (mackonstan, 14:35:24)
    1. trending analytics job is running out of memory, so trending graphs results not displayed (mackonstan, 14:36:37)
    2. Vratko: increased the memory few times but still out of memory, most likely due to excessive data produced by tests due to using PAPI vs. VAT (mackonstan, 14:37:24)
    3. waiting for Tibor to have a proper look, Tibor is back next week (mackonstan, 14:37:39)
    4. API coverage - Vratko: based on verify-api-crc work now have a good view of api's used by csit, but it's early days, once we enable api-crc gating job, this will drive api coverage evaluation (mackonstan, 14:39:46)

  9. VPP code performance (mackonstan, 14:40:17)
    1. need trending analytics to work again to compare and see if any anomalies (mackonstan, 14:40:33)

  10. Releases (mackonstan, 14:40:42)
    1. Dave: working on tcp tests for 1908, some risk due to PTO in the way (mackonstan, 14:42:15)
    2. Dave: refactoring wrk tests (mackonstan, 14:42:35)
    3. Dave: not planning QUIC tests. (mackonstan, 14:42:41)
    4. VAT-to-PAPI: Andrej: issues with LISP but fixed now (issue was with VAT not checking retval, and PAPI does, some refactoring) (mackonstan, 14:43:57)
    5. VAT-to-PAPI: policer wip, tap wip. (mackonstan, 14:47:00)
    6. VAT-to-PAPI: scale tests still running on VAT (mackonstan, 14:47:46)
    7. Arm - Juraj: dev and perf tests, not done yet, but very close, need to clean up images, aiming to have it working in time for 1908. Currently running arm mrr-daily, but failing at the moment, patch wip. (mackonstan, 14:49:31)
    8. Arm - Juraj: device tests, fixing executing image. (mackonstan, 14:50:39)
    9. Arm - Juraj: device tests, only one server right now, looking to add 2nd server. (mackonstan, 14:52:35)
    10. Arm - Juraj: device tests going ahead with 1 server for csit-vpp-verify non-voting jobs for now. Need 2nd server to do the same for vpp-csit-verify (mackonstan, 14:53:11)

  11. evelopments (mackonstan, 14:53:55)
  12. Developments (mackonstan, 14:54:02)
    1. move to Python 3.5 (post PAPI migration): need to start this post pulling the branch for 1908 (mackonstan, 14:55:29)
    2. Dave: VPP is already working on this: Paul, Naveen, Ole. (mackonstan, 14:55:52)
    3. New RF tags proposal - looking for finalize this shortly, but implementation post 1908 (mackonstan, 14:56:53)
    4. L1 KWs alignment between -func/-dev and -perf - Peter: setup, teardown aligned, most tests are too, few test specific Kws in progress, see updated google docs (mackonstan, 14:58:32)

  13. New tests (mackonstan, 14:59:39)
    1. IPsec scale tests - Peter: added 40k and 60k tests, after verifying vpp fix from Neale (mackonstan, 15:00:01)
    2. VXLAN tests on 2-node - Peter: tests running and reviewing the results (mackonstan, 15:00:19)
    3. NFV density - IPsec - Peter: in 3-node setup, also going up to 40k and 60k over multiple containers (mackonstan, 15:01:35)

  14. TRex (mackonstan, 15:02:15)
    1. Vratko: TRex running now with HDRhistogram, will be collecting the data, and in the future will process. (mackonstan, 15:02:46)
    2. https://drive.google.com/file/d/1wu-ZxrFoiaLFtWdxHrkQO-pzgk6ysa8I/view?usp=sharing (pmikus, 15:03:25)
    3. Maciek on PTO from next week for 2.5 weeks (mackonstan, 15:04:07)


Meeting ended at 15:04:11 UTC (full logs).

Action items

  1. (none)


People present (lines said)

  1. mackonstan (59)
  2. collabot` (4)
  3. pmikus (2)
  4. vrpolak (2)
  5. dwallacelf (1)
  6. jlinkes (1)


Generated by MeetBot 0.1.4.