#fdio-meeting: FD.io CSIT Project Meeting
Meeting started by mackonstan at 14:02:25 UTC
(full logs).
Meeting summary
-
- mackonstan (mackonstan,
14:02:31)
- Tibor Frank (tifrank,
14:02:32)
- Vratko Polak. (vrpolak,
14:02:35)
- Peter Mikus (pmikus,
14:02:35)
- Dave Wallace (dwallacelf,
14:02:41)
- Fan Zhang (fan_zhang,
14:02:51)
- Agenda bashing (mackonstan, 14:03:26)
- Inputs from projects VPP (mackonstan, 14:03:41)
- Andrew: rc1 postponed by 24hrs due to infra
issues (mackonstan,
14:03:57)
- Inputs from projects VSAP (mackonstan, 14:04:52)
- no update (mackonstan,
14:04:57)
- Inputs from projects (mackonstan, 14:05:01)
- TSC - 2022 budget discussion, more detail
later (mackonstan,
14:06:49)
- E810 flow "offload" patch (mackonstan, 14:06:57)
- functional device tests: onboarding, Fan
hitting issues with rxq configurations, Peter assisting with FW
upgrades and such (mackonstan,
14:07:45)
- perf: work-in-progress, Fa, delayed as need to
get functonal device tests working first. (mackonstan,
14:08:50)
- IPsec SPD flow cache patches (mackonstan, 14:09:19)
- Vratko: no progress, waiting for VPP
code (mackonstan,
14:09:36)
- TRex b2b loopback calibration tests (mackonstan, 14:09:52)
- Viliam: tests merged, now onboarding onto
jenkins and into trending (mackonstan,
14:13:01)
- IP scale tests (mackonstan, 14:14:16)
- Vratko: ip4, ip6 scale tests are back now with
exec CLI instead of VAT (VAT cmds removed from VPP); but tests are
taking longer making daily tests lasting >> 12hrs fixing that
by removing some scale tests from job_specs; next step is move to
VAT2; introducing /30 ip4 tests (mackonstan,
14:19:45)
- Releases (mackonstan, 14:20:19)
- CSIT-2110 - Tibor: waiting for vpp rc1 before
pulling the csit branch, jenkins jobs prepared (mackonstan,
14:21:27)
- rls2009_lts - jobs are broken anyways, no point
cherrypicking/fixing, removing the jobs. (mackonstan,
14:24:32)
- Physical Infrastructure (mackonstan, 14:26:12)
- Maciek: Trishan summary (mackonstan,
14:29:08)
- 1. AWS Performance Testing: Provide CSIT with
increased testing capabilities to execute test suite against AWS
Cloud server configurations and report results, significantly
increasing the testing profile of the lab, without investing in
physical hardware. Request for budget to execute tests in AWS
~12k/year. (mackonstan,
14:29:32)
- 2. Build machines: faster Arm, as current VPP
build takes 1hr on Arm, vs sub-20min on x86. (mackonstan,
14:29:43)
- 3. Performance testbeds: testbeds with i) fast
Arm servers (promised in the past, but still not confirmed; ii) low
power x86/Atom/SnowRidge servers as suggested by Matthew Smith; iii)
new Intel Xeon SapphireRapids SuperMicro servers that are confirmed
to ship in CY2022. Plus Xeon servers for TG/TRex per FD.io lab
funding agreement. (mackonstan,
14:29:56)
- 4. Replace software Router: Invest in physical
router/switch in DC. Virtual router currently used to manage
subnets, identified as the root cause of several underlying
issues. (mackonstan,
14:30:23)
- 5. Testing Data Analytics: Request for servers
and/or AWS services to process testing data for report generation.
Cisco provided existing data backend servers in lab. (mackonstan,
14:30:42)
- 6. Continue rack allocation, with 6th rack
allocation, as originally requested and approved in 2019.
(mackonstan,
14:31:00)
- Virtual Infrastructure, CI/CD (mackonstan, 14:32:16)
- Dave: debrief about the recent infra
instability; despite the move to YUL1 location, looks like Jenkins
is configured to use the public address for some transmission paths,
making Nomad clusters still using wireguard tunnels between sites;
working with Vexxhost and Vanessa, got full attention on it now;
however it is unclear what part of the configuration is the culprit;
need to monitor for few days, if instability reappears (mackonstan,
14:40:10)
Meeting ended at 15:01:03 UTC
(full logs).
Action items
- (none)
People present (lines said)
- mackonstan (33)
- collab-meetbot (5)
- tifrank (5)
- vrpolak (1)
- pmikus (1)
- dwallacelf (1)
- fan_zhang (1)
Generated by MeetBot 0.1.4.