#fdio-csit: FD.io CSIT Project Meeting
Meeting started by mackonstan at 15:02:58 UTC
(full logs).
Meeting summary
-
- Vratko Polak. (vrpolak,
15:03:00)
- Juraj Linkeš (jlinkes,
15:03:01)
- Dave Wallace (dwallacelf,
15:03:04)
- Agenda bashing (mackonstan, 15:03:07)
- Tibor Frank (tifrank,
15:03:13)
- Maciek, base freq vs turbo freq performance
testing (mackonstan,
15:03:43)
- Inputs from VPP (mackonstan, 15:04:30)
- Andrew, onboarding vpp make test multi-worker
in progress, increasing load on nomad clients (mackonstan,
15:06:52)
- Dave, v20.09.01 maintenance referred till after
DC move (mackonstan,
15:07:35)
- Inputs from TSC (mackonstan, 15:07:57)
- Vratko, ongoing discussions about Scapy induced
GPL re-licensing for VPP make test (mackonstan,
15:08:47)
- Releases CSIT-2101 (mackonstan, 15:09:02)
- Tibor, testing wise, almost all runs are
finished. But MLX cx556a got excluded by mistake. All 10 runs should
be completed by this Friday. (mackonstan,
15:10:25)
- Tibor, coverage tests, some are still ongoing,
should be done by this Friday. the long pole is 2n-tsh, will take
until next week. (mackonstan,
15:11:20)
- Tibor, report static content, open patch,
methodology on randomized ip scale tests by Vratko (mackonstan,
15:12:17)
- https://gerrit.fd.io/r/c/csit/+/31165
(mackonstan,
15:12:19)
- Tibor, another open patch by Juraj, rls notes,
known issue, merging now (mackonstan,
15:14:05)
- https://gerrit.fd.io/r/c/csit/+/31213
(mackonstan,
15:14:24)
- Agreed to push the report publish to
Friday (mackonstan,
15:19:55)
- Physical Infrastructure (mackonstan, 15:20:20)
- FD.io DC move, ping to the new YUL1 expected
this week (mackonstan,
15:24:42)
- Peter, would like to gracefully shut down all
the machine in MTL2 before move (once YUL1 is up and running)
(mackonstan,
15:25:42)
- Andrew .17 Arm server is down (Nomad)
(mackonstan,
15:25:57)
- Peter, monitoring solution is working well,
added Jenkins monitoring incl. the build queue, raising alarm if
more than 10 execution fail in the row! (mackonstan,
15:29:37)
- Environment, Framework (mackonstan, 15:32:35)
- Peter, working on updating pixi scripts, will
be dry running ubuntu 20.04 on hsw testbed(s) (mackonstan,
15:33:33)
- Maciek, TRex has now a partial support for
Intel E800 100GbE NICs, but quite a few caveats, evaluating impact
on CSIT, collaborating with RayK on that, as contribution to Trex is
from intel (mackonstan,
15:36:18)
- Performance Tests (mackonstan, 15:36:35)
- Maciek, All performance tests are done with
processor cores set to base frequency (re Arm, Juraj to check and
re-confirm this is the case). (mackonstan,
15:40:11)
- Maciek, some vendors (AMD) are asking to test
turbo frequency instead (mackonstan,
15:40:57)
- Challenges include: changing between base and
turbo frequency without reboot, repeatability of tests (mackonstan,
15:44:57)
- AI: Need to come up with a scheme to address
it: base frequency is here to stay, the question is to if, how and
how often to run tests with turbo enabled? (mackonstan,
15:48:11)
Meeting ended at 15:52:33 UTC
(full logs).
Action items
- (none)
People present (lines said)
- mackonstan (30)
- collab-meetbot (3)
- vrpolak (1)
- jlinkes (1)
- dwallacelf (1)
- tifrank (1)
Generated by MeetBot 0.1.4.