#opnfv-testperf: Weekly Test WG Meeting
Meeting started by mbeierl at 15:08:40 UTC
(full logs).
Meeting summary
- Roll Call (mbeierl, 15:08:54)
- Mark Beierl (mbeierl,
15:09:07)
- Emma Foley (efoley,
15:10:00)
- Alec Hothan (Alec82,
15:12:58)
- k8s testing (efoley, 15:13:58)
- Yardstick does not currently have K8s
performance tests - K8s contexts are being added (mbeierl,
15:14:53)
- Will be equivalent scenarios but different
context (ie use K8s instead of OpenStack) (mbeierl,
15:15:58)
- Storperf is tightly coupled with Heat, so k8s
support is a while away (efoley,
15:18:01)
- Discussion of test document and overall cohesion of test tools (mbeierl, 15:18:36)
- At Plugfest a summary was put together for all
the test tools (mbeierl,
15:19:17)
- Discussions about how to share the test tools
beyond OPNFV (mbeierl,
15:19:29)
- Emma and Georg submitted talks to 3
conferences (mbeierl,
15:19:45)
- More documentation might be "forced" to occur
if talks are accepted (mbeierl,
15:20:03)
- Do we attempt to standardize on a single test
result db, or have each tool rely on its own method to report
results (mbeierl,
15:24:02)
- Al mentions that own results are typically the
most reliable way (mbeierl,
15:24:27)
- test results DB is a central point of
failure (mbeierl,
15:24:47)
- test results db is becoming orphaned in that
there is no real PTL, don't know where to open JIRA tickets,
etc (mbeierl,
15:26:06)
- test results DB is being used as part of the
release process: example
http://testresults.opnfv.org/master/functest/status-apex.html
(mbeierl,
15:27:37)
- the fact that it is part of the release process
and no identified PTL is a concern (mbeierl,
15:27:56)
- Barometer running as a core part of this
'central test project' so that it is always collecting data and
anyone could examine it to see what happened during a test
run (mbeierl,
15:35:25)
- Barometer can use influx as a write source for
all metrics and any read plugin (prometheus, collectd, etc,) can be
used (mbeierl,
15:38:16)
- collectd needs to be manually installed on
compute, control, etc, (mbeierl,
15:40:56)
- read plugins (metrics) are configured on each
node and configured to send to barometer (mbeierl,
15:42:02)
- compute nodes KVM metrics are more interesting,
than say control nodes (mbeierl,
15:43:10)
- Barometer upstreamed ansible scipts for the
containers and made plugins (mbeierl,
15:44:05)
- Yardstick and Barometer could be looked at as
key areas for standardisation/integration between the test
projects (efoley,
16:01:58)
Meeting ended at 16:03:25 UTC
(full logs).
Action items
- (none)
People present (lines said)
- mbeierl (24)
- efoley (4)
- poutine15 (4)
- collabot (4)
- Our (4)
- issyl018 (4)
- Alec82 (1)
Generated by MeetBot 0.1.4.