#opnfv-cperf: cperf
Meeting started by dfarrell07 at 18:21:57 UTC
(full logs).
Meeting summary
- 
  - Jamo has cbench job running, w00t :D
    (dfarrell07,
    18:22:09)
 
  - dfarrell07 looked at RFC for controller perf
    testing, looks good, like hard work, no major feedback (dfarrell07,
    18:22:44)
 
  - Bluejeans is somehow showing mat as
    marcus (dfarrell07,
    18:24:48)
 
  - Mat has been working on containers for
    s3p (dfarrell07,
    18:25:09)
 
  - both control and compute containers seem to be
    in pretty good shape (dfarrell07,
    18:25:21)
 
  - both containers are on 16.04 now (dfarrell07,
    18:25:41)
 
  - mat thinks containers are consumable now
    (dfarrell07,
    18:25:49)
 
  - mat has up to 500 compute containers running,
    then starting to actually hit scale issues, like can create
    instances and they don't ping (dfarrell07,
    18:26:32)
 
  - this is good, now actually hitting real issues,
    not devstack or deployment issues (dfarrell07,
    18:26:45)
 
  - marcus' kolla work will combine with this
    soon (dfarrell07,
    18:27:00)
 
  - looking at two approaches for tests - one with
    all 500 started, other with slowly starting (like 10 at a
    time) (dfarrell07,
    18:27:38)
 
  - jamoluhrsen asks mat if can use compute docker
    containers one-off, how hard it is to point at some random
    controller IP (dfarrell07,
    18:27:58)
 
  - mat says they have no major deps on the
    network, has done PoCs that seem to show they move easily, robust in
    that situation, only complexity is in devstack (dfarrell07,
    18:28:35)
 
  - jamo has openstack+odl in POD 2, can he add a
    container as compute node in this env (dfarrell07,
    18:29:17)
 
  - mat says yeah, just need to point at service
    node for whoever will be doing openstack services, and odl control
    node (dfarrell07,
    18:29:37)
 
  - mat has put systemd in these containers, so
    more like LXC containers than normal single-process container
    (dfarrell07,
    18:30:14)
 
  - with systemd, kinda need to give it some time
    to startup, can't be instant, maybe 20 at a time (dfarrell07,
    18:30:30)
 
  - mat has commits to gerrit (int/pack/docker)
    inbound soon (dfarrell07,
    18:30:57)
 
 
Meeting ended at 18:31:40 UTC
(full logs).
Action items
  - (none)
 
People present (lines said)
  - dfarrell07 (20)
 
  - collabot (3)
 
  - mgkwill (1)
 
Generated by MeetBot 0.1.4.