#opnfv-cperf: cperf
Meeting started by dfarrell07 at 18:12:13 UTC
(full logs).
Meeting summary
-
- dfarrell07 was on PTO last week, no meeting I
think (dfarrell07,
18:12:27)
- https://docs.google.com/document/d/1P4Zl8TOegWnDEpAAamBZcqQMipZYp1HRs3Wn4PIVFbA
NetVirt Scale doc (dfarrell07,
18:22:55)
- acm has been working with latte this week,
getting it running against ODL, has been talking with Nikos about
measurements, has updated test env from Helium SR2 (lol!) to
something more recent (dfarrell07,
18:23:44)
- Nikos and acm have ideas for ways to improve
latte (dfarrell07,
18:24:34)
- wondering about ODL 6633 vs 6653, LuisGomez
reports that both are wired to same place internally, 33 is legecy
and 53 is IETF (dfarrell07,
18:25:36)
- acm may be seeing somewhat different behavior
for two ports, which shouldn't be the case, will try to reproduce
and report bug if can do so (dfarrell07,
18:26:05)
- Nikos has been on PTO (EU break), has been
fixing some bugs in lattee, seems to correctly capture pkt ins and
flow mods from multinet, next is to test with large topos and high
pkt rate (dfarrell07,
18:27:08)
- Some bugs in traffic gen multinet, working on
it (dfarrell07,
18:27:21)
- Nikos reports that acm's feedback about latte
stats have been very valuable, will work on (dfarrell07,
18:27:44)
- Nikos has been working on Dockerfile for all
components in NSTAT test, controller, *Cbench, multinet, etc
(dfarrell07,
18:28:16)
- This NSTAT Dockerfile is on DockerHub, can pull
image (dfarrell07,
18:28:29)
- Jamo has intel pod connecting to OPNFV Jenkins,
failing but seem to need Apex prereqs on jumphost, progress!
(dfarrell07,
18:31:33)
- Luis has been working on perf tests for
whitepaper, has sandbox job working for bulk-o-matic, still refining
but petty good (dfarrell07,
18:32:26)
- Sandeep has 3-node-cluster bulk-o job too, but
many issues, job under review (dfarrell07,
18:32:44)
- https://hub.docker.com/r/intracom/nstat/~/dockerfile/
link for NSTAT Dockerfile (dfarrell07,
18:33:02)
- talks to Nikos about adding that^^Dockerfile to
CPerf repo, building on OPNFV account on their infra (dfarrell07,
18:33:59)
- Al (acm,
18:34:28)
- Be is faster at pushing flows to switch, but
much slower at stats, inverse with Li plugin in B (dfarrell07,
18:35:37)
- https://jenkins.opendaylight.org/sandbox/job/openflowplugin-csit-1node-periodic-bulkomatic-perf-daily-only-beryllium/plot/
(LuisGomez,
18:35:52)
- https://jenkins.opendaylight.org/sandbox/job/openflowplugin-csit-1node-periodic-bulkomatic-perf-daily-only-boron/
(LuisGomez,
18:36:56)
- https://jenkins.opendaylight.org/sandbox/job/openflowplugin-csit-1node-periodic-bulkomatic-perf-daily-only-boron/plot/
(LuisGomez,
18:37:16)
- Detailed discussion about above test
results (dfarrell07,
18:41:41)
- Question about this tool - it lives in ODL,
Java-based, not talking to NB REST APIs, will compare to test that
uses NB REST (dfarrell07,
18:42:11)
- Welcome Bhuvan to the call, Veryex
person (dfarrell07,
18:44:09)
- Have ported tool install from Ubuntu-based to
CentOS-based, have it installing in CPerf Dockerfile, still working
on getting licensing software working (dfarrell07,
18:44:44)
- Options to give to CPerf - push install logic
to Dockerfile, or create RPM that we use in our Dockerfile
(dfarrell07,
18:45:25)
- Discussion about RPM option, would be nice
because reusable (dfarrell07,
18:46:57)
- Luis have been giving feedback about better
testing, script improvements, those have been created/updated, tool
made better (dfarrell07,
18:50:01)
- dfarrell07 discusses options for hosting
Varyx-tool RPM, can't put in git repo but can find somewhere for
it (dfarrell07,
18:50:22)
Meeting ended at 18:29:31 UTC
(full logs).
Action items
- (none)
People present (lines said)
- dfarrell07 (27)
- LuisGomez (3)
- collabot (3)
- acm (1)
Generated by MeetBot 0.1.4.