#fdio-csit: 17-Jan FD.io CSIT Project Meeting

Meeting started by mackonstan1 at 15:03:26 UTC (full logs).

Meeting summary

    1. Peter Mikus (pmikus, 15:03:30)
    2. Jan Gelety (jgelety, 15:03:31)
    3. Tibor Frank (tifrank, 15:03:32)
    4. Marek Gradzki (mgradzki, 15:03:44)
    5. Dave Wallace (dwallacelf, 15:03:45)

  1. agenda bashing (mackonstan1, 15:03:55)
    1. Ed Kern (snergster, 15:04:16)

  2. project updates (mackonstan1, 15:04:31)
    1. vpp rc2 label dropped last wed per schedule (mackonstan1, 15:04:50)
    2. dave: vpp rls on target for next week 24-feb (mackonstan1, 15:05:15)
    3. Tom Herbert (tfherbert, 15:07:08)
    4. honeycomb marek: rc2 milestone today, wip on target, no new issues for csit jobs, report rls1801 related manual sections updated (mackonstan1, 15:07:41)
    5. nsh_sfc: peter nsh_sfc 18.01 jobs are working, tests are passing (mackonstan1, 15:08:59)
    6. some emails about cross-project collaboration within LFN - Maciek to fwd to csit-dev. (mackonstan1, 15:12:09)

  3. Operational status (mackonstan1, 15:12:27)
    1. PeteR: tldk jobs are failing (mackonstan1, 15:13:03)
    2. tldk - is project alive, but no new code, not much activity on the list (mackonstan1, 15:14:21)
    3. https://lists.fd.io/pipermail/tldk-dev/2018-January/thread.html (mackonstan1, 15:14:29)

  4. rls1801 (mackonstan1, 15:23:25)
    1. https://wiki.fd.io/view/CSIT/csit1801_plan (mackonstan1, 15:23:50)
    2. csit rls1801 deliverables status updated at the link above (mackonstan1, 15:30:12)
    3. netgate cpe - physical connectivity still only partial - need 2x 10GE TRex ports per netgate testbed (mackonstan1, 15:32:29)
    4. negtage testbed target specificaiton should be documented in csit wiki (mackonstan1, 15:32:48)
    5. https://wiki.fd.io/view/CSIT/CSIT_LF_testbed (mackonstan1, 15:32:52)
    6. IPSEC+LISP patch fixing all ipsec lisp tests (mackonstan1, 15:37:55)
    7. https://gerrit.fd.io/r/#/c/10121/ (mackonstan1, 15:37:59)
    8. IPSEC+LISP bug VPP-1122 (pmikus, 15:40:54)
    9. another bug - ipsec multi-thread/multi-core and scale tests failing (mackonstan1, 15:41:36)
    10. VPP-1122 (mackonstan1, 15:42:06)
    11. draft rls1801 report generation - to be done now (mackonstan1, 15:48:10)
    12. dry run tests also can be started now (mackonstan1, 15:48:19)
    13. https://docs.fd.io/csit/master/report/ (mackonstan1, 15:48:21)

  5. Meltdown/Spectre (mackonstan1, 15:51:04)
    1. Fixed in Ubuntu Linux kernel 4.4.0-108.131(4.4.0-109.114) (mackonstan1, 15:51:08)
    2. Upgraded Cisco mirror trending testbed last Sunday to 4.4.0-109.114 and run trending jobs (mackonstan1, 15:51:26)
    3. Same/similar Haswell servers as in FD.io - executed daily performance trending jobs to get indicative numbers (mackonstan1, 15:51:35)
    4. Manual inspectiona and comparison of results from first run does not show any significant performance change (mackonstan1, 15:51:53)
    5. Some small fluctuations outside of expected stdev observed for vhost and memif tests between the runs, but need more runs to have statistically meaningful data. Note fluctuations are also observed with previous kernel, especially for NDR. (mackonstan1, 15:52:07)
    6. Execute rls1801 full set of tests - performance and functional - on existing kernel - publish rls1801 report as per published plan. (mackonstan1, 15:53:15)
    7. - include a note about indicative results with Meltdown kernel fix from internal testing. (mackonstan1, 15:53:24)
    8. Upgrade FD.io testbeds with Haswell servers to new fixed kernel ver., rerun full set of tests (mackonstan1, 15:53:35)
    9. - update the report with a separate section, providing statistically valid before/after comparison (for report each perf test is run 10 times) (mackonstan1, 15:54:00)
    10. - this will constitute public referenceable data for Meltdown performance impact. (mackonstan1, 15:54:15)
    11. Spectre - requires Intel Processor microcode upgrades (mackonstan1, 15:54:28)
    12. Published reports that Intel microcode published last week is introducing system instabilities per Intel public blog and press. (mackonstan1, 15:54:57)
    13. https://newsroom.intel.com/news/intel-security-issue-update-addressing-reboot-issues/ (mackonstan1, 15:55:10)
    14. Informed by Intel that BIOS update is required in many systems, but possibly not all. No UCS specific info, as the Spectre thing is still under info emargo within Intel. (mackonstan1, 15:55:35)

  6. FD.io lab updates - ARM (mackonstan1, 15:58:40)
    1. machhiatobin devices in labs, but not running any FD.io jobs (mackonstan1, 15:59:35)

  7. CSIT TOIs (mackonstan1, 16:00:55)
    1. TOI1: Tue 23-Jan 07:00-08:00 PST / 15:00-16:00 UTC / 16:00-17:00 CET (mackonstan1, 16:01:21)
    2. TOI2: Fri 31-Jan 07:00-08:00 PST / 15:00-16:00 UTC / 16:00-17:00 CET (mackonstan1, 16:01:35)


Meeting ended at 16:02:59 UTC (full logs).

Action items

  1. (none)


People present (lines said)

  1. mackonstan1 (47)
  2. collabot (4)
  3. pmikus (3)
  4. mgradzki (1)
  5. tfherbert (1)
  6. jgelety (1)
  7. dwallacelf (1)
  8. snergster (1)
  9. tifrank (1)


Generated by MeetBot 0.1.4.