►
From YouTube: Kubernetes kops office hours 20190412
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Hello:
everyone:
it
is
Friday
April
12th.
This
is
cop's
office
hours.
I
am
your
moderator,
just
in
Santa
Barbara
I
work
at
Google.
This
is
a
reminder.
This
meeting
is
going
to
be
recorded
and
put
on
the
internet,
and
please
be
mindful
of
our
code
of
conduct
would
be
a
good
person.
There
is
an
agenda
which
I
will
paste
a
link
to
in
the
notes,
and
we
actually
do
have
a
couple
of
great
items
on
there
already.
Please
do
feel
free
to
put
your
name
on
there.
B
A
C
I
just
want
to
bring
it
up
so
currently
today,
when
there's
a
change
to
an
ad
on
the
man
out,
there
there's
no
way
to
change
the
manifest
as
it
sits
in
the
state
store
on
without
rolling
the
version
number.
So
if
we
can't
use
any
other
configuration
options
inside
that
manifest,
is
it
roll
mush?
You
purge
the
labels
or
sorry
first
annotations
off
the
food
system
namespace
which
will
trick
it
and
force
it
to
pull
so
I
wanted
to
bring
up
some
ways
to
improve
that
you
can.
A
The
point
I
think
so.
Yes,
so
right
now
we
have
this
version
tag
on
the
add-ons
and
we
only
install
if
the
version
I
think
is
I
think
we
only
install
if
it's
currently
only
it's
newer.
Only
if
it's
different
I
think
we
only
know
if
it's
newer
and
yes
you.
We
now
have
the
ability
to
make
some
changes
to
the
add-on
that
our
configuration
changes
like
over
in
the
image,
but
that
doesn't
necessary
change.
A
They
that
version
I
thought
there
was
another
peace
in
their
cold
ID
or
something
that
that
we
could
maybe
use,
but
I
think
we
should
definitely
address
this
one
way
we
could
do.
It
would
be
so
I
think
there's
another
piece
in
there,
which
is
the
ID
the
other
way
to
do
it
would
be
to
have
a
sort
of,
and
if
the
idea
is
different
than
we
will
install
always
regardless
of
version.
Remember
this
ID
or
tag
or
something
like
that.
If
we
don't
have
that,
we
should
have
something
like
that.
A
I
think
and
that,
as
you
say,
it
could
be
that
the
hash
would
be
a
great
way
to
do
that.
There's
also
the
I
mean
soon
ish,
I,
I,
hope,
I
hope.
We
will
soon
start
to
adopt
the
add-on
operators
in
our
management
so
that
that
is
coming
soon
and
I.
Think
like
we're
clearing
the
decks
on
the
branches
for
a
bunch
of
those
bigger
changes.
So
we
can
talk
about
that
a
little
bit
but
yes
seems
full
term.
A
A
C
A
C
A
C
A
Okay
and
I
am
next
on
the
agenda,
and
this
is
about
so.
We've
talked
about
this
a
lot
in
the
past
that
I
have
finally
started
to
pull
the
trigger,
so
we
are
going
to
catch
up
on
the
cops
version
numbers
we
have
talked
in
the
past
of
how
the
cops
dot
zero
release
is
supposed
to
be
a
release
that
we
believe
to
be
good
and
ready
to
run
in
a
production,
or
you
know
close
to
production
environment
with
your
own
verification,
but
that
leads
to
a
delay
versus
the
kubernetes
release.
A
A
We
thanks
Ryan
for
the
PRI
thing,
which
is
that,
like
the
one
we're
gonna
put
in
that
and
then
I
this
morning
cut
one
thirteen
zero
alpha,
one
which
is
a
genuine
alpha,
but
is
up
on
github
and
there's
a
PR
for
114
zero
alpha
one
and
I
guess
I
will
do
cherry
picks
from
anyone
talking
through
cherry
picks,
but
I
will
also
do
cherry
pit.
I
will
drive
the
cherry
picking
to
those
release
branches
and
we
will
maintain
them
until
well,
maintain
them
and
we
will
for
115.
A
We
will
release
one
15
alpha
something
half
of
one
whenever
I
guess,
whenever
the
first
date
of
cudeman
eyes,
1:15
comes
out
my
camera
exact,
what
we
said,
but
that
sort
of
timeframe
and
I
guess
we
could
also
actually
have
that
be
the
master
branch
for
now.
But
this
also
means
that
now
we
can
put,
we
can
put
more,
we
can
put
bigger
changes
onto
the
master
branch
without
risking.
A
A
That
are
that
are
riskier
without
like
pushing
our
delay
further
behind,
because
then
was
it
Peter
that
was
a
user
to
the
the
cool
graph.
Thank
you.
Yes,
that
graph
was
pretty
compelling,
and
so
we
should
be
able
to
address
that
and
I
think
with
1:15.
We
should
have
cops
once
I
think
the
goal
should
be
that
we
release
cops
115
beta
1
at
the
same
time
as
kubernetes
115
zero
as
a
lot
of
numbers.
That's.
D
A
So
well,
then,
were
there
I
cut
them
this
morning
and
I
tagged
113
on
the
branch
and
I
did
114
as
a
PR
to
the
master
branch,
because
I
wanted
to
get
the
node
up
change.
Cuz.
You
know,
we
sure
don't
want
to
leave
that
isolated.
I
probably
actually
should
have
made
that
115,
but
I
don't
know
we
can
figure
it
out.
I
heard
yeah.
It
feels
a
little
weird
got
115
now,
but
I
guess
I
guess
I
said
message
is
so
yeah
and
alright.
D
D
So
yeah,
so
the
this
I
think
you
but
I
think
you've
had
a
busy
week
guys,
given
all
the
Google
stuff
going
on.
So
the
the
112
beta
1
release
notes
don't
line
up
with
actual
what
was
in
there,
because
I
clicked
on
I
was
like
oh
I,
have
a
PR
here,
but
that
was
from
2018.
So
I'm
not
sure
what
happened,
but
more
just
want
to
make
sure
it
looked
like
they
looked
like.
D
A
D
Then
the
next
one
is
also
me:
I
I,
just
since
Justin
you're
asking
about
your
doing
you're
talking
about
the
cherry-pick
script.
This
is
the
cherry
pick
script
that
I
use
and
it's
the
official
kubernetes
one.
So
this
assumes
that
you're
in
a
go.
You
know
path
in
cops
and
then
it
goes
up
one
directory
to
kubernetes.
D
A
Yes,
because
so
I
guess
one
of
the
things
we
should
figure
out
is
one
sorry
in
someone's
know
when
when
people
are
doing
more
PRS
today,
please
do
if
you
can,
if
you
can
highlight
whether
you
think
it
should
go
into
the
112
113
one
14
brunches
I
will
try
to
figure
it
out,
but
yeah.
It's
certainly
worthy.
If
you
have
an
opinion
do
play
in
it.
The
commit
not.
A
D
A
It
doesn't
okay,
yes,
the
if
we
if
there
are
features.
Yes,
we
have
so
you
should
put
them
on
milestones
the
milestones
they
were
more
forward-looking,
so
I
wouldn't
I,
wouldn't
try
to
use
milestones
to
donate.
You
don't
also
have
one
milestone
and
get
as
far
as
get
and
github.
As
far
as
I
know,
I
don't
know
if
we
could
do
a
label
yeah,
that's
what
you're
saying
yes,
I,
don't
think
we
should
have
milestones,
but
maybe
I
don't
know.
A
We
I
think
just
a
comment
to
start
with
and
we
can
sort
of
try
to
figure
it
out
and
then
a
green
I'm,
alright
cool
and
then
I'm
gonna.
Try
that
I
think
it's
important
to
try
to
cherry
pick
the
things
in
order.
Otherwise
I'm
worried
that
the
cherry
picks
will
start
to
break
so
yeah.
If
people
want
to
open
their
own
triptychs,
that's
wonderful,
but
if
not,
then
I
will
do
them
as
needed,
because
I
don't
want
to
miss
any
any
PRS
and
then
have
conflicts.
A
Well,
we
tore
through
that
agenda
all
right,
I,
don't
know
if
anyone
else
has
anything
they
would
like
to
discuss.
Otherwise,
you
know
we
have
looks
like
a
couple
of
things
to
do
and
lots
of
exciting
stuff.
So
the
exciting
thing
is
I
think
if,
if
people
want
to
do
bigger
things,
we
are,
we
are
hope.
We
are
certainly
at
the
point
where
you
can
do
that
and
it
might
not
land
until
cops
1
in
15.
A
Pickups
115
is
maybe
now
three
months
away,
at
least
for
beta
one
so
or
less
than
three
months
or
whatever.
Whenever
the
release
is
two
months
away,
so
please
do
send
exciting
stuff.
I
think
it's
that
don't
feel
that
you
should
be
constrained
by
that
we've
been
a
certain
sort
of
mini
mini
hold
because
of
the
one
Heather
of
the
backlog.
I
think
that
is,
that
is
no
longer
going
to
be
the
case
because
the
branches
are
cut
so
chunks
away.
We.
A
Or
that
are
required,
I
think
it's
just
building
more
confidence
in
that
release.
I
think
we
had
a,
we
had
a
I
think
Ryan
and
I
don't
know.
Maybe
some
other
people
found
some
issues
around
mixed
instance
policy,
which
is
new
functionality,
112,
but
still
important,
and
so
we
had
to
fix
that.
But
I
guess
it's
reassuring
that
the
things
that
are
broken
are
selling
or
a
new
functionality.
C
Yeah
I
would
add
that,
from
the
midst
instance
policy
standpoint
that
seems
solid
at
this
point,
I've
been
testing
out
on
our
test
clusters
and
both
new
to
mixed
instance,
policies
and
switch
path,
KS
cheese
and
documents
as
Falls
these
a
few
times
with
the
terraform
output
mode
and
I,
didn't
know
issues.
That's.
F
Great
I
got
a
chance
to
test
it
out
to,
and
everything
was
working.
Fine
I
found
a
few
issues,
but
that's
mostly
on
how
AWS
sets
things
up
and
not
on
cots
itself.
So
are
there
things
we
can
things?
We
should
fix
all
right,
yeah?
Oh
it's
just
how
AWS
launch
templates
kind
of
work
dusted
them
out
manually
as
well
to
verify,
and
there
was
nothing
weakened.
Specifically
they,
okay,
okay,.
A
G
G
So
I
did
some
playing
around
earlier
this
week
and
determined
that
yeah
I
do
need
to
use
that
CD
manager
for
that
I
also
went
into
the
cops
dev
channel
yesterday
and
kind
of
scrolled
up
and
saw
a
few
weeks
ago
that
this
was
already
discussed,
and
there
was
a
mention
of
updating
the
documentation
to
kind
of
include
that
and
be
more
explicit
I'm
happy
to
update
that.
If
that
would
be
helpful,
because
that
was
something
that
was
kind
of
confusing
for
me.
G
A
A
A
Thing
I,
thank
you
and
that's
a
very
light
way
of
giving
the
feedback
the
it
is.
It
is
hard
to
like
to
find
the
time
and
it's
hard
to
sort
of
see
the
documentation
with
fresh
eyes.
So
yes,
if
you
see
anywhere
that
the
dog
should
be
improved,
it
yeah
that
would
be
a
PR
is
greatly
appreciated.
Even
if
it's
not
perfect,
even
it's
just
like
you
know
getting
it
going.
A
Yeah,
please,
please
do
put
things
on
on
that.
I
guess
yeah
it's
hard
to
keep
up
with
the
flow,
but
on
slack
and
everything.
But
yes,
if
you
do
put,
did
have
issues
or
anybody,
she
seemed
to
be
great
way
for
permanent
longer
running
discussions
and
slack
seems
to
be
a
great
way
for
faster
ones,
but
the
slack
ones
do
tend
to
are
more
I
couldn't
get
missed,
but
do
you
open
those
and
that
that
would
be
anything
you
do?
There
is
greatly
appreciated.
A
F
Yeah
thanks
just
since
we're
on
the
at
CD
manager,
topic
I've
tried
to
use
it
a
few
times
and
I
keep
hitting
issues
at
points
and
I'd
love
to
debug
them.
I,
don't
even
know
if
they're
on
my
side
or
if
it's
something
with
that
CD
manager,
the
documentation
around
it
just
is
not
the
greatest
currently
and
even
digging
through
the
code.
It
can
be
hard
to
find
exactly
what's
going
on
or
where
I
should
even
be
looking,
so
just
giving
people
a
little
more
information
around.
A
I
can
try
to
write
some
Docs
to
be
more
like
this
is
how
it
should
work,
and
all
this
sort
of
thing
like
I
think
like
this
is
the
basic
flow
I.
Think
the
if
you
are
looking
to
understand,
what's
going
on
the
the
output
of
the
sed
manager,
pods
is
supposed
to
be
comprehensive.
I
think
that's
overly
comprehensive
there.
There
should.
B
A
A
leader
elected
or
leader,
determined
amongst
the
like
three
I'm
guessing
in
your
case
at
CD
manager,
five
CD
manager,
pods,
and
so
that
one
is
going
to
be
the
most
interesting
because
it
sort
of
drives
the
decisions
of
the
others,
but
yeah
you
wanna.
If
you
want
to
stack
with
them
or
something
or
put
them
up,
I
was.
F
A
Understand
I
know
you're
super
busy
busy.
It's
been
a
busy
time,
but
I
should
now
have
more
a
little
bit
of
a
little
bit
more
slack,
but
yea
be
yea
and
I.
Think
I
think
the
city
manager,
the
long-term
plan
for
that
is,
we
are
gonna.
We've
we've
kicked
off
a
project
joining
in
with
the
platform.
9
have
a
similar
tool,
that's
gonna,
be
an
official
sequester
lifecycle
project,
I
think
based
on
timings.
We're
gonna
try
to
get
everyone
onto
at
CD
3.
A
But
then
yes,
we're
basically
to
get
a
functionality,
that's
key
manager
in
there
and
have
it
be
hopefully
better
documented,
and
one
of
the
things
we're
working
on
is
the
idea
that
NCD
manager
is
an
automated
way
of
running
at
CD
ATM
commands.
So
it
should
be
more
self-explanatory.
What
is
happening.
F
D
Yeah
actually,
now
that
we're
talking
about
this
I
I
had
one
thing
that
I
forgot
to
actually
create
an
issue
for
or
anything
we
had
hard-coded
at
CD
versions.
We
were
already
on
that's
83,
but
I
think
we
were
on
3.0
dot,
something
and
it
just
didn't
work
at
all.
It
worked
once
I,
upgraded
the
3.1
and
then
3.2
and
3.2
was
good,
but
so
is
there
a
minimum
at
CD
version,
and
you
know,
if
that's
the
case,
I'm
happy
to
open
P
hard
to
document
that,
yes,
the.
A
A
We
should
also
expand
NCD
the
set
of
SED
images
included
in
SVG
managers,
the
ones
that
are
actually
supported
by
kubernetes
at
any
point
state
but
I'm.
Guessing
we
missed
one
if
you,
but
no
anyway,
yeah
exactly
don't
tell
my
boss,
I,
don't
know
I
I,
think
I
think
we
could
have
missed
one
because
it
I
think
there
was
one
which
I
didn't
put
in
there,
because
by
the
time
cops
got
to
it.
A
There
was
a
new
recommended,
like
kubernetes,
did
a
midstream
or
out-of-band
update
so
I'm
guessing
anyway,
we
could
put
all
the
SED
images
all
the
SED
versions
in
that
HCD
manager
image.
That's
one
approach.
There
are
a
lot,
but
we
can
certainly
put
more
in.
If
there's
one
missing
that
everyone
wants,
we
can
definitely
pop
it
in
there
right.
It's
not
that
it's
not
that
heavy.
We
can
make
it
that
at
City.
Manager
is
better
about
not
requiring
all
the
images,
older
versions
and
start
to
have
per
like
an
entity
manager
version
/
@
çd.
A
What
NCD
ADM
does
is
actually
downloads
at
CD
into
the
pod,
which
I'm
imagining
or
into
it,
which
I'm
imagining
will
make
people
a
little
bit
unhappy,
but
maybe
that's
a
better
option
than
if
you
meant,
if
like,
if
you
don't
have
it
at
all,
maybe
that's
a
better
option,
so
I
think
there's
stuff
we
can
do
an
entity
manager
I.
Think
the
I
think
we
should
at
least
start
with
validating
in
cops
that
you're
not
running,
something
which
is
just
not
gonna
work
right,
I!
A
D
G
E
G
A
E
B
I
hope
it's
okay,
I
did
it
last
minute,
so
some
of
you
know
I
added
off
support
for
digital
ocean
last
year.
I
don't
work
there
anymore,
so
I
don't
have
time
to
keep
supporting
it.
I
know
a
few
months
ago,
I
asked
if
anyone
else
want
to
maintain
it.
I
don't
think
there
is
anyone
that
wants
to
keep
supporting
it
and
push
it
to
the
data
so
I'm
at
least
willing
to
do
the
work
to
undo
the
Alpha
support
and
just
remove
it
entirely.
B
B
B
Yes,
they
do,
and
that's
one
of
the
larger
reasons
why
you
know
it's
it's
not
too
many
people
using
in
production
I
mean
if
it's
good
enough,
like
we
see
the
same
thing
with
Ezio
right,
so
no
one
is
willing
to
actually
build
it.
So
I
think
removing
something
that's
feature
gated
that
no
one
wants
to
maintain.
A
A
A
B
I
mean
it
could
be
a
nice
like
experimental
thing,
I
mean
it
is
also
the
only
supported
provider.
That's
using
the
cloud
controller
manager
to
which
was
kind
of
a
nice
way
to
see
how
feasible
it
is
and
cops
so
that
was
nice,
but
yeah
I
mean
I,
could
talk
to
Marco
and
see
if
he
wants
to
take
that
on
that
it
well
yeah
I'll
follow
up
I
just
wanted
to
kind
of
see
like
what
the
what
the
sentiment
is
around
this
and
if
we
rather
delete
it,
not
I.
B
G
B
A
Think,
okay,
either
way,
I
think
it'll
be
great
to
have
to
see
if
anyone
is
interested
in
in
doing
it,
and
also
because
it
is
feature
gated
with
the
understanding
they
can.
Try
like
someone
wants
to
try
something
cooler
with
it
like
you
know
like
let's
go
to
cluster
API
or
let's
do
something
like
that,
and
that's
actually
something
like
it's.