►
From YouTube: OKD Working Group 2020 05 26 Full Meeting Recording
Description
OKD Working Group 2020 05 26 Full Meeting Recording
A
And
if
you
haven't
yet,
if
you
could
yeah
there,
you
go
Craig
you're
being
better
than
me
and
faster
than
me
and
I'm
loving,
Joseph,
the
okd
for
GA.
It's
the
most
hopeful
shirt
I've
seen
all
day.
So
thank
you
for
that
and
we'll
figure
out
getting
a
panda
logo
on
that
t-shirt
and
when
we
go
GA
I'll
at
least
get
some
stickers
printed
from
the.
B
A
My
furthest
trip
so
far
has
been
from
a
ferry
ride
over
to
Vancouver
and
back
so
I'm
totally
jealous
that
you
got
a
real
vacation
in
well
played
so
I'm
gonna
share
my
screen
and
we're
gonna
put
the
Craig
put
the
list
in
there.
If
you
can
add
your
name
to
that
and
we're
gonna
drive
through
today
and
see
what
we're
doing
I'll
share
my
whole
screen
here.
That's
why
not
I
think
I
have
a
clean
thing:
yeah
yeah,
there's
still
some
stuff.
C
Okay,
so
yesterday
we
released
better
five
I,
don't
think
it
has
any
weight,
it
has
all
the
fixes
merged
to
run
a
single
master
install.
You
would
need
a
beefy
machine
for
that,
like
16
gigabytes
of
RAM,
and
you
still
need
a
bootstrap
node,
but
other
than
that
I
give
it
a
try
on
a
few
platforms
and
it
seems
to
work
so
I
would
appreciate
some
tests.
You
won't
be
able
to
upgrade
and
you
won't
be
able
to
make
host
modifications
using
MCO,
but
other
than
that.
It's
fully
functional.
C
C
C
A
A
A
A
A
Couple
of
dead
links,
or
at
least
very
old
links-
yes,
probably
so,
if
you
point
them
out
to
me,
I'll
either
remove
them
or
find
a
more
relevant
one
and
hopefully
we'll
get.
There
he's
also
been
working
on
updating
the
logo,
for
so
it
has
a
an
actual
four
on
the
camera
on
the
side,
so
stay
tuned
for
a
new,
ok,
D,
panda
imaged
to
go
out
with,
for
hopefully
that's
good,
so
cool.
A
C
A
The
only
other
thing
that
I
have
is
we.
We
moved
yes,
yesterday
what
we
were
supposed
to
do,
an
okay,
D
working
group
am
a
session
on
and
we've
moved
it
to
I
think
June.
First
next
Monday
as
part
of
the
briefing
series
that
I
do
at
I.
Guess
it's
1600,
UTC
or
1,700
1,600,
usually
just
the
hour
before
this
and
I'll,
send
a
note
out.
If
people
want
to
join
that
just
to
try
and
drive
some
more
interest
in
an
okay,
d4
and
answer
any
questions
that
are
out
there.
A
There
have
been
a
lot
of
questions
that
I've
been
noticing
and
people
participating
in
the
kubernetes
OpenShift
dev
channel
from
a
number
of
different
spaces.
So
there
seems
to
be
quite
a
few
people
who
are
testing
on
working
with
ok
d4,
but
not
participating
in
the
working
group,
so
I'd
like
to
to
get
some
of
those
folks
enrolled
and
engaged
in
the
community
conversations
here.
A
F
E
Yep
cool
cool
I
just
wanted
to
share
that
for
the
past
weekend,
I'll
be
testing
okay
deal
on
vSphere,
trying
to
use
IP
I,
not
EOP,
I
and
I
haven't
been
successful.
The
the
Fedora
machine
that
I
got
from
the
beta
4.4.
It's
not
booting
up
and
get
some
some
with
error
or
on
the
cardinal
walk
during
the
initialization
and
do
the
investigation
I
realize
that
there
are
few
Bugzilla
is
opened
for
for
I.
E
C
C
Assembling
official
Knightley's
would
take
some
time,
but
we
could
share
something
we
have
right
now
and
hopefully
that
will
work
if
it
won't
and
we
would
find
a
problem
in
fedora
course.
We
would
have
to
wait
for
official
nightly
so
that
we
could
have
something
to
share
with
our
engineers,
and
that
would
be
something
more
complex
than
just
some
random
things
I
built
over
the
weekend,
but
yes,
giving
it
a
preview
for
the
for
the
nightly
definitely
possible.
Yes,.
B
So
I'd
say
that
there's
still
two
things
here,
so
we
want
to
get
everything
merged
into
master,
but
for
the
4.5
release
timeframe
we
will
still
be
building
off
of
a
branch,
so
I
think
the
more
important
ones
are
getting
that
into
master,
but
on
the
Atlanta
doesn't
necessarily
block
GA.
We
just
really
want
to
get
ya
the
certainty
that
it's
eventually
going
to
merge,
but
before
we
call
it
a
GA,
it's
yeah,
it's
only
well.
Actually
buddy
made
a
good
great
list
of
things.
B
A
B
A
Alright,
this
may
be
a
short
meeting.
Cuz,
you
guys
aren't
his
talkative,
as
you
usually
all
are.
A
So
the
young,
the
only
other
thing
besides
these
getting
to
GA,
which
is
sort
of
the
end
all
and
be
all
of
everything
in
my
world
at
least
you
know.
Add
these
four
to
here
is
the
coop
con
North
America
is
has
their
call
for
papers
out
and
I
would
love
to
see
a
talk
submitted
by
maybe
some
of
us
here
to
talk
about
building
an
opinionated,
kubernetes
distro.
A
There
are
lots
of
opinionated
things
out
there,
so
I'm
wondering
if
anyone's
game
to
do
to
that
submission,
it'll,
probably
more
and
likely
be
a
virtual
event,
but
it'll
still
be
be
an
interesting
one.
So
if
you're
a
game
for
that,
let
me
know,
and
we
can
collaborate
on
it
and
get
it
in
and
in
a
timely
fashion.
A
A
There
will
be
I,
don't
think
we'll
ever
let
Vadim
or
Christian
off
the
hook
on
this
one
just
because
it
hits
GA.
It's
still
gonna
keep
in
sync
with
OCPs
release
cycles
and
the
new
features
and
functions.
So
this
this
project
is
gonna,
be
on
going
on
no
Vadim
for
CEO
I.
Unfortunately,
Paul
got
the
role,
but
you
know
there's
always
a
possibility
that
Paul
Cormier
will
retire
and
we
can.
You
know,
put
that
email
for
CEO.
You.
A
A
You
go
all
right,
so
anyone
else
got
stuff
that
we
need
to
to
get
done
today.
I
know
this.
This
is
shockingly
quick
for
an
okay
D
meeting.
I
can't
believe
it
must
be,
is
Neil
not
on.
You
know.
A
No
I
don't
want
to
cut
you
off
at
the
at
the
bit,
but
if
anyone
wants
to
to
work
on
the
a
draft
of
a
call
for
papers
for
coop
con,
just
reach
out
to
me
and
I
will
help
craft
it
and
hopefully
get
it,
get
it
accepted
and
we
can
go
on
and
we
might
actually
I've
been
trying
to
get
a
white
paper
done
sort
of
categorizing
kubernetes
distributions,
with
a
couple
of
other
folks,
so
I'm
looking
to
see
about
getting
some
more
visibility
for
okay,
be
out
there
in
the
world
once
it's
GA.
A
D
Answer
it
I
can
answer
a
question.
Joseph
brought
up
earlier,
just
to
add
more
conversation
here,
so
just
if
you'd
asked
about
if
the
autoscaler
would
work
with
these
fear,
once
we've
got
everything
together.
So
just
as
a
as
a
blanket
a
statement
in
OpenShift
cluster
autoscaler
uses
our
version
of
cluster
epi,
which
we
call
machine
api,
so
anything
that
the
machine
api
operator
is
able
to
control.
D
D
Yes,
yeah,
okay,.
D
If
you
look
at
the
way
we
deployed
the
cluster
autoscaler,
we're
using
in
fort
this
should
be
in
4.4,
we
use
a
provider
called
open
shift
machine
api,
that's
the
name
for
the
provider
we
give
and
that
you
know
so.
The
autoscaler
is
just
creating
like
machine
sets
and
machine
deployments.
It's
modifying
those.
So
it's
not
actually
speaking
directly
to
the
provider.
The
machine
api
is
speaking
to
the
provider,
so
any
any
provider
implementation
we
have.
That
would
work
through
ipi
or
through
UPI
I.
D
E
D
Need
to
so
there's
few
different
types
of
resources
that
you
want
to
be
aware
of.
One
is
called
the
cluster
autoscaler
resource
and
the
other
one
are
called
machine,
autoscaler
resource,
a
cluster
autoscaler
resource.
There
should
only
be
one
of
them
and
you
create
that
to
tell
openshift
to
make
you
the
autoscaler,
so
it
will
start
the
autoscaler
operator
or
the
autoscaler
yeah
operator
them
a
process.
D
I
guess
it's
just
a
controller
really,
so
you
create
a
cluster
autoscaler,
and
this
is
all
documented
that
will
launch
the
the
autoscaler
and
then
you
create
machine
autoscaler
resources
to
associate
a
machine
set
with
the
autoscaler.
So
if
you
have
a
bunch
of
machine
sets,
you
need
to
create
a
machine
autoscaler
to
tell
the
cluster
autoscaler
to
watch
that
machine
set,
and
then
you
can
set
kind
of
the
upper
and
lower
bounds
on
you
know
how
much
it
should
scale
up
or
scale
down.
Basically,.
D
A
I
have
one
more
thing,
because
there's
always
one
more
thing
and
just
when
you
think
you're
going
to
get
off
early
I
did
a
talk.
The
other
day
with
Alvin
crickey
from
kinfolk
IO,
and
the
folks
who
have
flatcar,
along
with
a
whole
lot
of
Inspector,
Gadget
and
other
kind
of
cool,
tooling
and
things
and
I
was
curious.
If
anyone
had
looked
at
flatcar
at
all
in
terms
of
ok
d4
and
what
what
this,
if
that
was
a
possibility
of
something
to
collaborate
with
them
on.
C
C
C
A
Could
add
a
comma
at
one
and
on
the
community
board
and
and
they
weren't
they
weren't
thinking
about
it?
It
was
just
something
came
up
in
a
conversation
so
and
I,
don't
I
haven't
heard
of
anybody
who
wants
it.
So
it's
yeah,
it
looks
like
it's
pick
three,
but
I'm
just
I'd
be
curious,
so
maybe
on
the
community
board,
we
have
a
column
of
areas
for
experimentation
and
just
to
keep
track
of
that
in
case
that
so
we
can
add
something
there.
So
thanks.
That
was
great
and
thank
you
for
those.
A
D
A
E
There
sometimes
you
to
add
to
the
conversations
that
a
lot
of
the
traffic
we
see
in
the
opposite
channel
or
urban
areas
lags
that
most
of
the
people
that
are
trying.
Ok
t
they
went
through
the
the
user-provided
infrastructure
side.
So
if,
if
you
would
be
able
to
use
in
your
demo,
no
examples
from
user,
provided
that
would
be
very
helpful
for
the
community.
D
All
right,
I'll
take
some
time
to
see
if
I
can
get
ok,
D
up
and
running,
if
I
can
use,
vSphere
I
will
but
I'd,
you
know,
I,
don't
have
a
vSphere
account
right
now,
so
the
main
things
available
to
me
are
like
AWS
and
GCP
are
what
I
work
on
most
of
the
time.
So
if
I
can
just
do
like
an
oak
ad
install
the
AWS,
that
would
probably
be
easiest
for
me,
but
the
mechanics
of
what
I'm
going
to
show
they
won't.
D
A
Because
I
have
I
have
a
slot
on
Wednesday
the
3rd
of
June.
At
my
9:00
a.m.
you,
I
think
your
your
East
Coast
aren't
you
noon,
I
actually
think
I,
because
this
is
a
reoccurring
thing.
People
asking
this
question:
if
you're
asking
in
the
working
group
I'm
sure
there's
people
out
there
asking
as
well,
so
it
might
be.
Let's
let's
aim
for
that:
the
3rd
of
June
and
we'll
do.
It
seems
like
a
good
good
topic
and
is.
A
I'm
on
I'm,
on
line
now,
four
days
a
week,
I
found
someone
to
do
one
of
the
day's
Tuesday's
for
me
doing,
briefings
from
on
all
kinds
of
topics
by
Commons
members
and
different
working
groups
and
Monday's.
The
plan
is
to
always
do
an
with
one
of
the
upstream
projects.
So
you
guys
are
the
guinea
pig
for
next
Monday
at
9:00
a.m.
A
A
That
was
was
that
Joseph
or
that
was
Charl,
so
I'm
from
Old
Dominion
to
see
if
they
would
join
so
maybe
doing
a
couple
of
little
demos.
I,
don't
know
how
long
it
takes
you
to
do
an
auto
scale
or
a
demo,
and
a
single
cluster
node
thing,
but
just
I
have
I
can
go
as
long
as
I
want
on
Monday.
We
could
go
two
hours
if
we
need
to,
and
maybe
that's
what
we
do
on
Monday
is
have
the
AMA
do
a
quick
here.
A
You
hear
you
do
auto-scaling,
and
here
you
know,
here's
how
you
do
it
on
single
cluster
just
to
have
things
to
talk
about.
If
that
sounds
reasonable
for
folks
and
then
I
can
clip
it
all
into
the
little
videos
that
I
like
for
my
youtube
keeping
and
so
people
don't
have
to
watch
the
whole
two
hours
and
listen
to
me
blather
on
in
between
when
you
guys
go
silent
on
me,
sounds.
D
Good
to
me,
like
so
yeah,
send
me
a
send
me.
Some
info
I
would
probably
want
like
10
to
15
minutes
for
mine,
because
it
just
getting
the
autoscaler
working
is
one
thing,
but
then
to
show
it
scaling
out
and
scaling
it
back
in
can
take.
It
could
take
four
or
five
minutes,
because
you
need
to
wait
for
them.
Yeah.
G
D
Yeah,
like
I,
would
probably
just
do
that.
I
would
show
people
you
know
will
will
turn
the
autoscaler
on
will
make
a
workload
and
force
it
to
scale
up
we'll
watch
it
scale
out.
Then
we'll
delete
the
work
known
and
we'll
watch
it
scale
down.
You
know
that
should
be
enough,
and
then
we
could
do
like
QA
or.
D
B
D
So
I
just
wanted
to
respond
a
little
bit
to
some
of
the
things
that
Joseph
is
saying
here.
So
you
saying
like
you,
want
to
see
a
machine
deletion
and
then
the
autoscaler
kind
of
fill
in.
Actually
that
doesn't
even
take
the
autoscaler.
The
machine
api
will
do
that
for
you
so
like
as
it
stands.
Currently,
if
you
do
not
scale
down
your
machine
set,
but
you
were
to
delete
a
machine,
not
the
node.
Please
don't
delete
nodes
that
that
disrupts
things.
D
But
if
you
delete
the
machine
object,
then
the
art,
then
the
machine
API,
will
actually
take
care
of
removing
that
and
then
creating
a
new
one
for
you.
So
like
the
autoscaler
to
exercise
the
autoscaler,
we
oftentimes
look
more
at
creating
workload
in
the
cluster
because
what
the
autoscaler
wants
to
look
for
it
doesn't
look
to
see
if
the
size
of
those
groups
have
changed.
D
What
it
looks
to
see
is
are
they're
underutilized,
nodes,
meaning
nodes
that
could
be
reaped
from
the
cluster
and
are
there
unscheduled
pods
and
if
there
are
unscheduled
pods
and
there's
availability
that
will
increase
the
size
of
the
cluster.
Otherwise,
it
won't,
but
just
doing
things
like
deleting
machines
and
watching
it
fill
back
in
that's
actually
the
Machine
API.
That's
doing
that
for
you,
not
the
autoscaler.
D
D
Right,
so
what
what
I'll?
Probably
what
I
do
in
these
situations
is
like
I'll,
probably
turn
the
autoscaler
down,
so
it
only
looks
for
like
a
very
short
time
to
respond
and
then
yeah
we
make
workloads
that
basically
have
such
large
memory
requests
that
we
can
only
fit
one
on
a
node
and
then
it's
real
easy
to
tell
the
autoscaler
like
okay
I
want
to
rep.
You
know
make
ten
replicas
of
this
workload.
It
should
make
10
machines
to
fill
it
out.
So
what
will
do
something
like
that?
Probably
ok,.
A
So
I
will
add
you
to
the
the
call
on
Monday,
so
you'll
get
an
invite
and
I
will
post
a
note
on
the
mailing
list
with
the
details
for
that.
Funnily
enough,
it's
exactly
the
same
number
that
you've
dialed
in
for
blue
jeans
here
today.
So
you
don't
have
to
worry
too
much
about
that.
But
so
that's
that's
how
this
whole
thing
works.
F
Because
my
idea
of
the
countdown
page,
because
a
dota
I
think
we
should
not
underestimate
the
importance
of
this
G,
a
thing
I
would
say
because
yeah
the
time
frame
was
so
long.
People
are
waiting
a
lot
of
that
I
think
we
should
yeah
they
should.
We
could
do
something
to
make
it
more
interesting
to
gain
more
awareness,
that
there
is
something
cool
coming
in
the
pipeline
and
yeah
I.
Don't
know
no
a
PR
counter
is
the
right
thing
here
or
if
we
find
something
something
better
but
I,
think
it's
worse.
Yeah.
A
B
Jump
in
here,
real
quick,
so
I
really
like
to
set
a
date
once
we
are
able
to
do
that.
So
I
think
we're
getting
very
close
to
to
actually
being
able
to
do
that.
But
right
now
there's
a
few
more
things
that
need
to
happen
internally
the
planning
and
then
we
can
really
commit
to
a
date.
I
think
we're
getting
close.
But
I
would
like
to
push
that
conversation
out
at
least
till
next
meeting
and
then
be
where
we're
at
at
that
time.
Isn't
yeah
yeah.
F
A
That's
what's
top
of
mind
for
me,
so
I
think
we
have
a
little
bit
of
time,
so
maybe
between
getting
a
logo
and
a
launch-
and
we
can
make
this
celebrate
this
in
a
couple
of
ways-
it'd
probably
go
back
to-
we
can
do
a
blog
post
about
it
on
openshift
comm
I'm
happy
to
do
that
and
write
that
up
and
I'm
get
the
word
out
that
way.
I,
don't
think
you're
gonna
see
Red
Hat,
do
any
big
press
release
or
announcement
about
it.
A
It's
gonna
have
to
be
totally
community,
driven
as
it
does
and
compete
with
OCP
and
all
the
other
product
offerings.
So
we
kind
of
a
it
company
kind
of
play
under
the
radar
from
corporate
marketing.
So
it
can
happen.
I
can
make
that
happen
and
then
we'll
just
I'll
use
my
platform
on
OpenShift
Commons
briefings.
We
can
host
an
open,
Schiff
Commons
briefing
announcing
the
GA
and
we
can
start
the
cadence
of
okd
working
group
AMA
session.
So
we
did
an
ok
D
for
AMA
next
week.
So
do
we
have
it?
A
A
B
Yeah,
we
will
be
merging
all
the
changes
in
the
four
point:
six
development
timeframe,
which
is
open
now
and
then
for
four
point:
five,
who
will
have
a
separate
branch,
so
we
will
be
basing
off
of
the
four
point:
five
release
branch,
but
then
add
a
few
additional
comments
on
top
of
that
to
enable
everything
to
work
with
ignitions
back
three
and
Fedora
core
OS.
From
the
four
point,
six
release
on
it'll
be
built
off
of
the
same
branches
of
the
release.
B
Four
point
x:
branch,
just
like
OCP,
that
was
our
our
goal
for
GA
and
now
we've
sort
of
to
not
have
to
wait
for
that,
we'll
be
releasing
GA.
Probably
off
of
the
F
cost.
Four
point:
five
branch,
but
we
want
to
be
sure
that
all
the
changes
will
then
land
in
four
point
six,
so
we
don't
regress
in
any
way
from
four
point:
five
to
four
point:
six,
because
once
we
go
GA,
we
can't
just
you
know
there.
B
A
A
A
So
that
that's
I
think
that's
the
other
thing.
The
other
thing
that
I
was
thinking
about
too,
is
that
maybe
I
should
try
and
see
if
I
could
find
a
twitter
if
okay
di
Oh
was
a
Twitter
handle
that
I
could
steal
to
start
tweeting
out
some
stuff
too,
but
let
me
see
if
I
can
get
that
done.
So
with
all
of
that
anything
else,
we
should
be
there's
nothing
else
in
the
chat
here
then
I
see
that
I
missed
it
see.
Sample
operator
also.
C
C
A
C
G
C
B
Yeah
I
think
it's
it's
not
too
far
off.
We
just
need
buy-in
from
from
those
teams
that
are
actually
concerned
with
maintaining
what
we
propose
an
outline
there
yeah,
but
I
think
it'll
go
through
and
we'll
we'll
have
that
very
soon
to
merge
and
yeah,
and
then
we
can
really
look
at
implementing
everything.
I
mean
folk
for
most
the
things
we
have
done
the
code
already
samples
operator
certainly
is
the
thing
we
have
to
look
at
fixing
actually
changing
up
a
few
things,
but
we
need
you
know
before
we
put
in
any
effort
there.
F
Would
very
appreciate
to
hear
in
some
not
today,
but
in
a
later
meeting
of
how
the
relationship
between
OCP
and
Oh
kitties,
because
I
always
thought.
Okay,
Dee
is
some
something
to
get
used
to
OCP.
And
if
you
have
enough
traffic
on
your
OPD
cluster,
some
you
can
switch
over
to
OCP,
because
this
is
a
set.
What
we
plan
to
do
and
I
think
lots
of.
B
Now
we've
been
calling
it
because
it's
it's
it's
the
same
codebase,
but
then
it
has
a
different
base
operating
system
and
with
the
with
the
base
operating
system,
there
is
a
upstream
downstream
relationship,
but
it
is
not
in
the
actual
cluster
code,
so
the
cluster
codes
same,
which
is
what
yeah
it's
a
sibling
distribution
and
then,
in
the
future
of
course,
you'll
be
able
to
upgrade
from
okay
d2l
CP,
with
the
commands
you've
been
using
to
upgrade
from
one
beta
to
the
next
and
just
switch
out
the
base
image
and
you
know,
become
a
customer
that
way
definitely.
F
C
B
I'm
sure
there
will
be
limitations
on
the
you
know
on
the
workloads
we
can
migrate.
But
if
it's
you
know
if
there's
survey
a
supported
product
or
operator
on
the
OCP
side-
and
you
do
that
beforehand
on
ok
d,
you'll
be
able
to
transition
that
workload
into
OCP
I
mean
that
I
think
legal
is
not
gonna,
be
against.
That
I
mean
I,
can't
really
speak
to
that,
but
just
make
sense
for
Red
Hat.
B
A
Alright
then
I'm
multitasking,
as
everyone
can
probably
tell,
and
trying
to
update
a
road
to
GA
column
there
in
the
community
page
with
the
list
of
the
open
issues
and
links
to
them.
So
if
I
didn't,
if
I
missed
something,
please
add
it
there
Christian
everybody
I,
think
there's
one
more
but
I
losing
the
thread
here
today.
So
I.
A
Anything
else
we
should
capture
today,
if
not,
hang
me
on
the
kubernetes
slack,
if
you're
interested
in
collaborating
on
the
submission
for
coop
con
north
america
and
we
can
write
something
up
and
let
me
know
I'll
reach
out
to
Charl
and
see
how
long
his
talk
on
the
single
cluster
install
would
be
for
Monday.
And
if
he's
available
to
do
that
too
and
Mike,
all
I
will
add
you
to
the
blue
jeans.