►
From YouTube: Kubernetes Kops Office Hours 20180511
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
Good
morning,
everyone
or
good
afternoon
it
is
May
11th.
This
is
the
cop's
office
hours.
We
have
a
couple
of
items
on
the
agenda,
please
feel
free
to
add
your
agenda
items,
and
so
we
can
be
sure
to
get
to
them
and
your
names
in
the
attendee
section.
Otherwise,
I
guess
we
should
just
jump
right
into
it.
I
want
to
think
that
one,
a
quick
update
last
last
week
was
coupon
in
guca
vu
in
Copenhagen
I
think
it
went
great.
A
It
was
great
to
see
everyone
or
some
people
there
and
lots
of
interest
in
cabanas,
of
course,
and
in
cops
and
lots
of
interesting
I
think
what
we're
working
on
being
the
right
things
in
terms
of
you
know:
@cd
adoption,
add-on
management.
Those
seem
to
be
the
two
like
hot-button
topics,
so
in
general,
I
think,
oh
and
more
more
clouds
and
I
think
those
are
those
are
the
sort
of
things
between
you
know.
A
We
should
be
working
on
and
the
feedback
I
think
we
got
was
generally
positive,
that
those
are
the
things
for
important
to
people.
So
yeah
I,
don't
know
anyone
has
anything
else.
They
want
to
add
about
cube
con,
but
a
bigger
and
bigger
every
year.
I
know
it
seems
like
I
haven't
been
talking
pony
you
before
I
think
it
was
about
the
same
size
as
coop.
A
Con
us
was
I,
guess
like
about
a
year
ago,
so
not
the
most
recent
one
but
previously,
and
so
they
just
keep
getting
bigger
and
bigger,
and
it's
awesome
to
see
the
community
other.
The
next
thing
on
the
agenda
was
I
want
to
give
an
update
on
110,
as
in
case
for
people
that
don't
know
we're
trying
to
get
the
sed
at
see.
The
upgrades
working
from
entity
to
two
at
City
three.
The
goal
is
to
get
that
as
an
opt-in
in
its
in
cops
110.
That
is
going
well.
A
The
last
big
thing
is
the
adoption
of
existing
volumes,
so
that
so
I
think
I
have
it
at
least
on
my
mic
testing
working
where
you
can
switch
between
sed
versions
and
upgrade
from
its
attitude
at
c3.
If
you're
using
the
manager,
you
can
opt
in
to
the
manager,
the
missing
is
right.
Now
are,
if
you
were
not
using
the
manager.
If
you
have
an
existing
cluster
and
you
start
using
the
manager,
we
need
to
adopt
those
that
data,
that's
obviously
pretty
important
and
that
just
needs
either.
A
A
There
is
one
challenge
which
is:
we
were
talking
about:
switching
to
Etsy,
DV
3
by
default
for
new
clusters.
The
problem
is
I,
don't
think
we
can
do
that
because
we
would
have
to
set
a
net
CD
version
of
let's
say,
3,
to
write
in
this
back
and
then
the
problem
is
when
we
going
forwards.
We
don't
know
whether
you
actually
want
to
stay
on
three
already
want
to
use
the
latest
version.
A
I
think
we
in
default
dead,
CD
3,
but
I
yeah
I'm
not
entirely
sure
we're
gonna
be
able
do
the
creation
of
new
clusters
defaulting
to
at
CD
3
if
you
are
not
using
a
CD
manager,
but
we'll
see
I
need
to
play
around
with
that.
So
hopefully
this
weekend
address
the
PR
backlog.
There
are
some
great
PRS
in
there.
I
can
see
one
of
them
highlighted
and
next
up
and
then
cut
the
110
alpha.
We
don't
have
to
obviously
get
all
the
pr's
in
there.
A
B
B
Think
Shane
is
on
the
call
to
essentially
my
concern
is
more
so
around
how
we
want
to
approve
these
PRS.
So
I,
don't
think
anyone
on
the
owners
list
is
an
expert
and
not
using
a
driver
cloud,
but
I
mean
looking
at
the
code.
I
mean
the
coast.
Is
fine,
there's
nothing
in
there.
That
would
break
any
other
functionalities
and
cops.
So
it's
a
mentality
going
forward
going
to
be.
B
A
I
think
that's
that's
a
very
accurate,
so
we
obviously
would
be
great
to
have
like
people
from
Alibaba
into
the
owners
file
to
be
like
the
relevant
reviewers
as
more
people
contribute
to
that.
But
yes,
I
mean
I.
Think
like
when
I
look
at
them,
I
look
for,
is
it
gonna
break
other
clouds
and,
like
obviously,
people
that
are
contributing
new
clouds
might
not
be
as
familiar
with
cops
in
general
and
so
like?
Are
there
tricks?
Are
there
like
one
of
the
other?
A
The
people
are
falling
into
one
of
the
pitfalls
that
sort
of
thing,
but
yes,
the
one
thing
we
could
do
is
I,
don't
know
Shane.
If
you
want
to
consider
creating
a
fire
if
it's
currently
a
future,
flagged
Alibaba
or
not,
but
a
feature
flag.
So
GC
is
still
a
feature
fact,
and
what
that
means
is
that
technically
we
can
make
changes,
make
breaking
changes
to
it.
A
Obviously
we
don't
like
to,
but
we
can
do
that,
and
so
it
may
be
handy
to
have
a
feature
flag
just
so
that
you
can
feel
very
comfortable
making
changes
in
future,
but
that
doesn't
really
impact
yes
and
answers.
Your
question,
the
pr,
if
it
code,
looks
good,
isn't
breaking
other
clouds,
let's
get
it
in
there.
Yeah
I
was
gonna,
I
was
gonna,
look
through
in
there.
I
ran
out
of
time,
so,
but
yes,
that
would
be
the
standard
I
would
also
apply.
Okay,.
C
I
just
heard
this
on
there,
just
as
a
talking
point,
it
kind
of
relates
to
what
you
were
just
saying:
deficit
about
110,
and
you
know,
depending
on
how
far
out
the
Alpha
is.
Is
it
worth
considering
a
1:9
one
for
a
bunch
of
the
stuff?
We
talked
about
last
time
because
I
brought
up
at
our
last
meeting
cubed
ENS.
There
was
a
big
issue
that
I've
patched
on
ours
on
our
clusters
manually,
but
you
know
I
read
patch
it
all
the
time.
Sorry.
A
That's
my
god.
We
should
talk
good.
We
should
definitely
do
that
and
do
one
nine
one
goes
in
there.
But
yes,
certainly
the
DNS
one
is
worthy
and
I.
Don't
recall
any
PRS
that
were
like
warranted
going
into
one
least
they're,
certainly
like
I,
think
there's
a
PR
for
a
new
version
of
calico,
but
my
inclination
would
be
it's
a
new
major
version,
and
so
we
should
go
into
cops
110
rather
than
going
into
one
nine
one.
A
A
A
A
You
could
change
the
image,
it's
not
the
best
idea
and
we're
gonna
try
to
make
it
more
things.
Nothing
that
looks
like
we're.
Gonna
try
to
make
it
and
try
to
figure
out
or
I'm
trying
to
figure
out
how
to
get
it
so
that,
in
this
case,
cube
DNS
would
publish
their
own
base
customization
resources
and
we
would
reference
them,
which
would
be
a
little
easier.
A
So
yeah,
that's
that's
sort
of
I'm
trying
to
get
that
mic
hack
working
and
then
we
can
sort
of
make
it
nicer.
It
would
be
great
I'm
playing
with
core
DNS
right
now,
which
we
merged
the
PRF.
For
that
and
so
accordion
s
is
now
opt
in
and
as
it
is
in
kubernetes
that
I
would
like
to
get
that
I
expect
there
will
be
more
revisions
of
core
DNS.
You
know
it's
new
code.
We
expect
lots
of
versions
of
it
and
I
would
love
to
not
have
to
do
tops
revisions
whenever
that
happens.
A
You
Peter
existing
I
am
rolls
aah
existing
animals
yeah
everyone's
favourite,
yes,
so
I
went
through
your
comments
and
I
noticed
with
the
exist
and
warned
if
changes
lifecycle
that
it
is
not
failing
if
it
doesn't
exist.
Okay,
I
said
that,
yes,
so
people
don't
know,
this
is
the
idea
of
using
so
cops
currently
creates.
I
am
objects,
in
particular,
I
am
roles
and
a
lot
of
people
in
enterprise
type
scenarios
want
to
pre,
create
them
or
create
them
a
little
differently.
A
We
want
to
make
that
possible
and
easy,
but
we
don't
necessary
want
to
give
up.
We
don't
to
be
completely
like
you
can
do
whatever
you
want,
and
we
won't
even
mention
it
if
it's
wrong,
and
so
we
have
this
life
cycle
mode
called
a
life
cycle
exists
and
warn
if
changes
which
will
check
that
the
should
check
that
the
I
am
role,
objects
that
you
are
specifying
do
exist
and
if
they
are
not
exactly
as
we
expect
it
will
be
a
warning.
We
have
a
different
mode
where
it
is
a
error.
A
If
things
are
not
exactly
as
you
expect
so
this
is
sort
of
it
should
be
a
in
my
opinion,
a
good
UX
in
that
it
will
enable
you
to
make
the
change
and
it
will
enable
you
to
create
your
own.
I
am
I,
am
objects
and
it
will
do
some
validation
and
it
will
give
you
a
warning,
but
not
an
error
if
they
aren't
quite
right,
which
I
don't
know
seems
like
a
good
compromise
to
me.
A
D
We
assume
that
we
could
run
one
of
the
life
cycles
and
it
would
just
check
to
see
if
something
exists
and
it
wouldn't
fail
and
the
life
cycle
is
checking
to
see
if
exists,
and
it
will
continue
to
check
to
see
if
it
exists
and
it
doesn't
like
gracefully
say:
no,
doesn't
it
exist
and
goes
away
it
just
loops
and
or
fails
so
I
think
this
is
a
problem
in
our
life
cycles
that
we're
using
the
wrong
when
we
set
the
life
cycle
to
a
specific
warning,
it's
the
wrong
one.
That
makes
sense.
D
A
A
To
mind
using
the
doc
aids
that
local
to
not
create
any
run
53
resources,
but
then
I'm
wondering
about
this.
The
API
servers
served
so
the
API
servers
cert
will
always
be
that's
fun
yeah.
It
would
always
be
like
it's.
It's
I
am
promotions,
isn't
it
Justin,
so
the
egg?
That's
the
API
server
cert
will
actually
for
Kate's.
That
Luca
will
be
the
low
bouncers
name
so
that
shouldn't
really
matter.
That's
an
AWS
name,
and
it's
a
since
also
insert
the.
A
There
is
a
way
to
specify
a
hosted
zone.
Id
I
think
that's
global,
so
I
think.
If
you
specify
the
host
zone
ID
correctly
of
the
other
hosted
zone,
it
might
just
work
as
long
as
you
had
permissions
in
both
directions
or
permissions
both
for
the
account
you're
running
under
when
you
do
cops,
update,
tester
and
for
the
I
could
give
permissions
to
that
hosted
zone
ID
in
the
account
the
cluster
lives
in,
but
I'm,
not
sure
how
you
would
specify
in
the
account
the
zone
lives
in
that
this
other
account
can
modify
this.
A
But
what
what
may
be
an
easier
trick
is
you
can
create
a
hosted
zone
in
your
account,
a
sub
hosted
zone
in
Europe
in
your
own
account
and
delegate
from
the
parent
to
the
child
as
it
were.
So
if
your
parent
was
example.com,
you
could
have
other
account
example.com
and
you
would
setup
NS
records
for
other
other
accounts
on
example.com,
in
example,
calm,
and
then
you
would
create
your
cluster
as
cluster
to
other
accounts
on
example.com,
so
that
that's
it's
a
slightly
I
guess
it
depends
what
the
motivation
is
right.
A
A
So
what
if
I
were
to
use
a
lifecycle
override
card
and
give
the
post-event
zone
ID
of
the
other
account
and
then
specify
a
hosted
or
a
lifecycle
override
to
not
create
the
records
because
of
the
I
am
permissions
yeah
the
I
could
make
them
myself.
Yes,
most
of
the
record
creation
is
going
to
be
done
by
DNS
controller,
so
I
think
cops
does
try
to
create
some
sort
of
placeholder
ip's,
but
they're,
not
important
and
I.
Don't
think
it
I,
don't
think
it's
a
hard
error.
A
If
there
was
Croatians
fail,
you
certainly
could
create
them
yourself.
I
I
don't
know
whether
it's
gonna
be
worth
the
effort.
I
mean
you.
Could
you
could
also
like
create
a
host
zone
in
your
local
account
and
your
other
account
and
then
cops
would
populate
it,
and
then
you
could
mirror
those
DNS
settings
right
like
at
least
then
everyone
will
be
happy
and
would
you'd
have
a
reference
source,
and
you
know
the
mirroring
is
something
you
could
you
could
write
without
you
know
having
to
like
reinvent
DNS
population,
but
I
would.
A
A
Oh,
if
you
haven't
seen
it
I
did
at
PR
very
recently,
which
lets
people
opt
out
of.
What's
a
proof
of
concept
right,
it
lets
people
opt
out
of
networking.
This
is
similar
external,
which
opts
out
of
configuring
routes
on
a
particular
subnet,
so
it
won't
configure
routes
among
configure
about
tables.
It
won't
configure
Internet
gateways
and
won't
configure
neck
gateways
and
configure
anything.
A
It's
basically
saying
like
I
know
what
I'm
doing
I
have
a
vgw
I
have
a
direct
connect
to
have
it,
whatever
the
idea
being
that
we
can
still
see
it
in
the
cluster
spec,
but
we
don't
necessarily
have
to
go
through
each
one.
This
was
sort
of
prompted
by
someone
that
did
in
that
instance
Peter,
which
is
was
great.
So
we
already.
A
We
now
have
the
ability
specify
in
that
gateway,
or
in
that
instance,
I
think
someone
was
asking
whether
they
could
specify
a
a
vgw,
a
VPN
gateway
and
then
but
then
they
also
had
a
use
case
for
for
specifying
like
just
a
wreck
or
for
using
Direct
Connect.
So,
in
other
words,
the
opt-out
mechanism,
I
have
I
am
wary
that
it
will
cause
people
to
mess
up,
but
we
will
see
I
guess
we
have
99
PR
is
open.
We
have
90
90
hours,
open
we're
starting
to
look
like
other
projects,
we're
doing
great.
E
A
A
A
E
D
Technically
and
who's
in
charge
of
it
is,
let's
be
frank:
we
have
an
owner's
file
which
has
multiple
members
that
have
PR
permissions
in
different
levels
of
peer
right.
It's
peer
review
rights.
If
you
go
into
honors
you'll
notice,
my
name
Justin
and
Andrew
Kim,
Mike
I,
think
there's
three
owners
on
the
call
today.
E
D
B
Going
okay,
yeah
yeah
bugs
just
kind
of
like
in
the
alpha
stage,
I'm
hoping
someone
will
there
has
been
a
couple:
she's
open,
so
I'm,
just
kind
of
waiting
so
like
it
doesn't
support
multi
masters,
yet,
which
is
probably
the
biggest
blocker
for
pushing
for
beta
or
anything,
but
we'll
get
there
soon.
Yeah.