►
Description
For more Continuous Delivery Foundation content, check out our blog: https://cd.foundation/blog/
Kubernetes SIG & Cloud SIG - AWS, GCP, Azure - Gopinath Rebala, OpsMx
A
So
mainly
we'll
do
it
similar
to
what
we
are
done
in
the
morning.
If
there
are
any
big
issues
that
we
want
to
discuss,
we
can
touch
on
those,
and
then
we
can
discuss
what
has
been
going
on
for
the
last
one
year
or
six
months
with
the
specific
Cloud
drivers.
And
what
are
the
things
that
we
want
to
discuss
going
forward
in
terms
of
roadmap.
C
E
A
Think
it
has
not
been
well
well
documented
on
the
website.
There
are
a
bunch
of
PR
so
that
are
raised
in
last
three
four
months
on
a
continuous
basis
for
them.
So
one
of
the
engineers
is
also
online.
Kiran
he's
been
active
actually
working
on
those,
so
there
is
a
new
support
for
cloud
run
for
Google
and
there
were
enhancements
with
the
load,
balancer
settings
and
some
option
additional
options
for
deploying
VMS.
E
B
Yeah
but
the
other
ones
were
okay,
okay,
yeah
yeah.
Well,
it's
always
hard
to
like
in,
in
my
opinion,
an
open
source
when
you
release
something
it's
very
hard
to
deprecate.
Anything
like
you
could
say
someone's
not
using
something,
but
a
lot
of
the
times
like
someone
in
the
Far
Far
Corners
of
the
universe
is
probably
using
it.
B
So
it's
difficult
right,
so
you
know
because
we
have
released
it
I'm
more
prone
to
say
we
probably
should
just
support
it.
Unless
there's
no
maintainers
or
active
yeah.
If
that's
the
issue,
then
we
probably
should
look
for
someone.
That's
willing
to
do
that
and
if
not,
then
we
could
probably
just
say
hey.
This
is
unmaintained,
but
you
know
it's
not
deprecated.
You
know
I'd
rather
go
that
route
because
and
then,
if
no
one
decides
to
actually
ever
pick
that
up
then
I
feel
like.
B
E
B
But
yeah,
like
I
said
due
to
the
nature
of
Open
Source
I
like
to
keep
things
like
as
when
they're
released
to
be.
You
know
as
supported
as
it
can
be,
and
if
we
do
deprecate
give
customers
at
least
a
good
amount
of
time
to
say,
hey
we're
thinking
about
deprecating.
This
at
least
create
a
GitHub
issue.
B
C
B
Definitely
I
definitely
think
that's
a
you
know,
I
think
out
of
everything
you
know,
like
I've,
seen
some
a
lot
of
a
lot
more
movement
like
the
AWS
as
and
I've
seen
some
movement
on
the
Google
Cloud
one.
That's
what
kind
of
shocked
me
like
that
you
said
it
was
it
was
it
wasn't
working
but
but
yeah
yeah,
definitely
I,
think
I
think
we'll
I'll
go
I
can
create
that
issue
and
we
could
just
say
Hey.
Listen.
These
are
the
things
that
we
see
that
people
are
actively
working
on
and
are
maintained.
B
A
You
could
also
see
these
the
Cadence
that
we
have
with
the
sigs
we
can
see
if
we
can
combine
them.
I
think
that
was
one
of
the
proposals
saying
today
is
even
kubernetes.
Is
it's
regularly
done,
but
other
things
they're
not
necessarily
regularly?
Iran
participation
also
matters
right,
so
what
we
could
do
is
we
can
combine
those
and
the
issues
come
up
based
on
that,
we
can
create
a
project
if
needed,
we
but
use
a
combined
Sig.
So
we
have
a
better
coverage
on.
B
Those
yeah
since
I
don't
have
a
lot
of
experience
like
on
the
Google
Cloud
side
of
things,
especially
with
the
cloud
driver
portion.
Is
there
a
GitHub
issue
that
highlights
all
the
things
that
you're
talking
about
that?
Doesn't
work,
because
if
there's
not
then
yeah,
if
you
could
create
one,
that
would
be
fantastic.
C
C
E
C
D
E
A
D
F
Sorry,
the
fix
for
Cloud
driver,
not
starting
with
the
Google
cloud
provider
enabled
is,
is
fixed.
There's
been
a
PR
raised,
it's
merged
to
master,
it's
back
ported
to
the
129
Branch,
but
it's
true,
there's
there's
another
problem:
I
think
it.
If
it's
actually
asking
this
question,
we're
still
trying
to
fix
the
the
variety
of
things
that
were
happening
in
that
clone
dialog
box
and
the
AWS
provider,
and
some
of
them
are
fixed
and
some
more
of
them
are.
F
B
Yeah
I
think
for
me,
with
with
AWS,
because
I
know
David's
put
out
some
PRS
that
I
saw
but
I
didn't
know
about
the
Clone
issue
with
the
front
end
of
it.
That's
yeah,
that's
that's,
definitely
an
issue
that
we've
had
well.
We
had
one
engineer:
I
forgot
who
it
was,
but
she's
she's
done
an
excellent
job
with
I
think
she
works
at
AWS.
She's
done
an
excellent
job
with
both
the
UI
and
and
the
the
back
end.
F
Maybe
we're
all
talking
about
the
same
person
but
there's
or
someone
named
pratibha
who
works
for
AWS
who,
whose
job
at
various
times
has
been
to
make
improvements
of
Spinnaker
and
she's,
made
a
bunch
and
yes,
a
lot
of
it
has
been
on
the
Java
side
and
then
there's
also
been
some
on
the
on
the
UI
side.
She's,
definitely
much
more
of
a
of
a
Java
person,
and
you
know
we
can.
F
We
can
all
sort
of
do
the
thing
where
we
like
crawl
around
in
a
dark
room
and
try
to
find
our
way,
which
is
I.
Don't
I,
don't
want
to
say
that's
what
she
did,
but
that's
what
I
would
have
done
and
you
know
she
got
it
95
right-
and
this
is
some
like
sort
of
gnarly
Corner
case
about
field,
invalidation
and
other
validation
logic.
And
you
know
when
you
touch
a
field
and
you
dirty
it,
and
you
display
the
message
at
the
right
time
and
anyway
so
she's.
F
You
know,
like
all
of
us,
juggling
priorities
about
internal
things
and
external
things
and
she's
totally
committed
to
fixing
this
and
we've
Tracked
Down
people
that
are
sort
of
all
in
the
most
difficult
time
zones
possible
to
collaborate
with
her
she's
I.
Think
on
the
east
coast
of
the
US,
so
we're
getting
there
and
I
guess
we
could
release
1.29.1
with
with
that
stuff,
partially
fixed.
The
Google
cloud
provider
thing
like
I
said,
is
fixed
and
and
there's
another
deck
issue.
F
It's
also
fixed,
so
it
might
be
worth
getting
it
out
there.
It
just
seems
like
we're
so
close,
and
everybody
was
here
anyway,
so
it
doesn't
really
seem
like
we're.
Gonna
do
it
today
or
yesterday
so
I'm
I
I,
know
it's
really
bad
to
have
a
DOT
zero
version
out
there,
that's
busted
or
any
version
out
there,
that's
busted,
but
it
feels
like
it's
better
to
wait.
It
feels
like
we're
so
close
that
I'd
rather
I'd
rather
wait.
F
C
Are
there
at
least
like
test
builds
that
we
can
run
like?
Is
that
like
do
we
have
like
a
process
like
do
an
RC
and
like
run
it
in
eager
candidates,
maybe
not
exposing
to
the
community
like
1.29.1
right,
but
maybe
we
say
like
1.29.rc1
and
like
if
people
want
to
try
it
give
some
feedback.
Maybe
we
can
even
find
things
before
we
merge
and
look
again.
We
are
in
different
environments,
so.
F
F
It
would
be
just
as
much
work
to
publish
an
RC
as
it
is
to
publish
a
real
thing,
and
it
would
actually
be
more
because
I'm
not
sure
we
have
a
way
for
to
publish
a
bomb
that
hired
would
consume,
but
but
on
every
single
PR
merge
to
master
Docker
images,
do
get
published
on
Debian
packages
do
get
published
so
in
at
least
in
one
way
they
are
out
there.
It
takes
a
little
more
Ingenuity
to
use
them.
It's
not
as
easy,
as
you
know
how
deploy
apply.
F
B
A
E
Yep
yep
yep
exactly
so
I
I
just
want
to
say
that
before
the
previous
build
process
broke,
we
did
publish,
nightly
unvalidated
heads
of
Master,
so
there
should
be
a
way
for
Howard
to
consume
the
unvalidated
bombs.
So
we
could
create
release
candidates
Etc
in
order
to
give
them
testing
back
to
David's
point,
though
it's
just
as
much
work
to
release
a
release
candidate
as
a
final
build.
So
it's
yeah.
B
F
No,
no
I,
I,
think
I
know
at
least
I
have
something
constructive
to
say
if
it's
like
a
regular
deployment
like
a
target,
a
Target
deployment.
So
the
fact
that
it's
AWS
eks
I
think
is
maybe
a
little
bit
of
a
distractoid.
It's
a
kubernetes
thing.
It
probably
doesn't
matter
whether
it's
running
Amazon's,
kubernetes
or
anybody
else's
I,
don't
know
if
it's
fair
to
inject
this
into
the
conversation,
but
I
have
gathered
over
the
course
of
this
conference
that
there
are
I.
F
What
I
think
our
fixes
in
the
kubernetes
provider
to
make
them
make
it
possible
for
them
to
be
better
and
and
have
a
real
Argo
rollouts
thing
and
then
yes,
it
is
just
a
kubernetes
problem
to
have
Argo
rollouts
through
the
blue
green
deployment,
but
first
for
Spinnaker
to
say,
like
Cube
control
apply
manifest
and
to
get
the
Argo
robots
thing
down.
There
is
one
thing
and
then
to
have
you
know
Spinnaker
to
understand.
F
I
mean
there
are
lots
of
ways
to
do
blue
Greens
in
in
kubernetes
Argo
rollouts
is
just
one
of
them,
but
it'd
be
nice
to
have
at
least
that
answer,
because,
like
I
said
I
think
three
companies
have
done
it
and
other
people
are
asking
so
we'd
at
least
have
some
a
story
to
tell
which
we
don't
have
today.
I.
C
The
time
like
I
saw
today,
we
spun
up
like
three
different
pipelines
in
the
contributor
Summit
right.
So
like
me,
if
we
take
a
lot
of
things
for
granted,
because
we
we
live
this
world,
but
when
someone's
first
seeing
this
they're
just
trying
to
figure
out
how
these
dots
connect
like
in
that
world
yeah.
F
F
A
F
F
B
So
I
guess
what
I'm
getting
out
of
this
is
that
I
don't
know
if
that's
what
they
were
asking
because,
like
I
said,
it
was
a
little
vague,
but
from
what
I'm
gathering
from
what
you're.
Saying,
though,
is
that
Spinnaker
needs
a
better
way
of
of
being
able
to
provide
blue
green
deployments
to
to
the
end
user.
Is
that
is
that
correct?
Sorry
for
just
yet
kubernetes
yeah
yeah.
B
D
B
D
B
How
do
you?
How
do
you
manage
deployments
in
how
did
you
manage
the
deployments
in
Spinnaker?
So
it's
asking
about
deployments
that
Spinnaker
does
so
so
yeah
for
that
yeah
for
the
eks
question,
so
so
yeah
I
think
to
answer
the
first
question
which,
if
I,
think
if
I
understand
it
correctly,
it's
what
does
Spinnaker
support
for
deployment
I.
Think
that's
what
they're
asking.
B
We've
we
listed
quite
a
few
of
them
already
AWS,
being
one
of
them,
Google
Cloud
being
another
major
one
and
then
the
less
used
ones
that
were
mentioned
like
Oracle
and
whatnot
and
there's
various
different
ways
that
you
can
deploy,
but
there's
also
the
whole
kubernetes,
which
you
know
is
supported
by
both
these
major
Cloud
providers
that
we
just
listed
and
the
ways
of
going
about
to
deploy
to
these
kubernetes
environments
could
be
via
customize
or
Helm
charts.
So
yeah,
there's
there's
a
lot
there.
B
B
B
Oh,
how
do
you
manage
the
deployments
in
Spinnaker?
For
the
most
part,
we
don't
really
manage
the
so
once
the
deployment
is
done
like
we
don't
really
well
I
I,
don't
really
just
manage
the
point.
I
usually
go
through
the
kubernetes
because,
like
we
deploy
due
to
points
through
kubernetes,
so
I
just
use
kubernetes
to
like
investigate,
look
at
and
see
like
if
the
pods
are
running
or
or
whatnot,
or
you
know,
look
at
the
log
game
there.
So
I
don't
really
use
Spinnaker
as
a
management.
B
A
E
A
E
B
Cool
so
yeah,
so
that's
pretty
much
like
you
know,
like
I
hope
that
answers
his
or
her
question.
So,
if
it
doesn't,
please
please
let
us
know
so
something
that
I
wanted
to
talk
about
which
actually
David
brought
to
my
attention.
Yesterday.
I
didn't
know
this,
but
I
didn't
know
that
the
AWS
Cloud
driver
was
I.
I
didn't
know
that
it
was
using
actually
V1
of
the
SDK
and
I.
B
Think
it's
a
good
shout
out
that
if
we
can
move
to
V2,
so
if
you
know
someone's
willing
to
make
that
one,
because
it's
going
to
be
a
little
bit
of
work,
there's
a
lot
of
code
there
and
V2
is
pretty
different
and
how
it
manages
like
configuration
like
how
it
does.
The
lake
injects
with
so
clients
in
the
V2
of
the
Java
AWS
Java
SDK,
for
instance,
is
region
bound
as
before,
like
so.
B
D
F
I
I
don't
know
how
to
do
this
off
off
the
top
of
my
head,
but
the
bombs
are
all
published
in
Google
Cloud
Storage.
Maybe
Cameron
can
dig
up
a
link
and
send
it
around
sorry.
I
don't
have
well.
So
the
question
is
like
how
to
how
to
try
out
I
think
the
question
is
about
how
to
try
out
new
unreleased,
stuff
and
I.
Think
I
said
before
that.
F
The
way
to
do
that
is
to
is
to
make
your
own
bomb
and
so
I,
don't
actually
I
know
exactly
what
to
type
with
my
own
two
hands.
But
what
I
would
do
is
download
an
existing
bomb
and
then
start
adjusting
it
I
I'm,
pretty
sure.
What's
in
the
bomb,
are
you
know,
Docker
tags
and
and
other
version
numbers
and
I
I
just
haven't
done
it
with
Debian
packages,
so
I'm
I'm
really
shooting
in
the
dark
but
I
I.
Think
if
you
do
some
pattern
matching,
it
will
all
come
together.
F
I
mean
there's,
you
know
the
the
there's,
a
git
repo
called
build
tools,
github.com
spinner
card,
slash,
build
tool
and,
if
there's
even
a
readme
for
it-
and
this
is
what
we
use
to
construct
the
releases,
it
has
like
a
a
build
bomb
command
and-
and
you
could
potentially
try
that
out.
But
those
are
the
same
bombs
that
are
uploaded
to
GCS.
So
I
would
ignore
what
I
just
said
and
and
download
a
bomb
from
from
GCS
and
start
editing.
B
B
Everyone,
but
that's
where,
like
a
lot
of
our
knowledge,
resides
but
with
Debian
like
I'm,
a
little
hesitant
like
even
still
how
that
would
work
just
depend
because
I'm
worried
that
some
dependencies
might
require
other
dependencies,
because,
like
Debian
packages
are
just,
you
know,
they're
actually
going
to
be
linking
to
one
another.
So
I'm,
just
kind
of
curious
like
how
that's
going
to
work
like
being
able
to
just
change
things
they
might
not
line
up
properly,
but
I
could
be
wrong.
B
B
B
Well,
yeah
I
think
I
think
that's
it's
pretty
much
like
every
cloud
like
every
cloud
provider
lives
in
Cloud
driver
from
what
I
know
so
yeah
so
I
mean
there's
Google
Cloud,
there's
Oracle,
there's,
there's
Cloud
driver
there's
and
then
yeah,
there's,
basically
everything
so
and
if
it's
not
there
I'd
be
surprised.
To
be
honest.
B
B
Mostly
support,
oh
sorry,
sorry,
okay,
yeah,
yeah,
sorry,
all
right
all
right
backwards,
all
right
so
so
I
know.
Eks
has
no
issues
running
Spinnaker
like
AWS
eks,
which
I
imagine.
If
the
cloud
provider
supports
kubernetes
you
can
you
can
definitely
run
Spinnaker
and
then
what
even
with
that
said,
I
would
say.
E
A
B
And
it's
obviously
going
to
be
very
different
from
company
to
company
too,
like
you're,
going
to
probably
have
different
network
policies
that
you
might
need
different
IM
roles.
You
know,
depending
on
what
your
security
team
wants
so
but
like
with
all
that
said
and
done,
you
know
like
AWS-
is
a
very
powerful
cloud
provider,
so
is
Google
Cloud.
C
B
B
There
would
only
be
me-
and
you
know
sometimes
one
other
person
in
there
as
well
so
I
think
having
getting
more
people
and
then
also
collaborating
with
them,
because,
like
AWS
problems
are
going
to
also
exist
in
Google,
like
they're,
you
know
like
they're,
basically,
the
same
same
deals
just
different
apis
right
so
and
then
I'm
sure,
like
the
problems
that
AWS
is.
That
has,
like
you
know,
I'm
sure,
like
like
the
less
technical
ones,
but
like
wanting
to
support
something
is
probably
going
to
be
true.
You
know,
like
oh,
we'll,
be
like.
B
Oh,
we
just
supported
this
new
AWS
feature
and
then
Google
Cloud
might
be
like.
Oh,
we
also
have
that
feature
we
might
want
to
implement
it.
So
I
definitely
want
to
consolidate
into
a
single
seg
as
as
Cloud,
but
maybe
keeping
kubernetes
separate
like
I.
Don't
know
that's
up
to
Cameron
but
like,
but
we
can
definitely
include
kubernetes
as
well
because,
but
we
might
have
to
call
like
cloud
cloud
Netties
there
we
go
yeah.
E
A
B
B
C
B
Yeah
no
worries
all
right,
so
I
think
that
pretty
much
wraps
it
up
right,
all
right,
I
think
we're
good,
so
yeah.
Thank
you,
everyone
for
joining
and
then
we'll
go
ahead
and
update
everything
to
now
point
to
like
this
new
Cloud
Sig
make
sure
we
update
the
calendars
and
then
update.
You
know
like
whether
or
not
we're
using
Zoom
or
Google
cloud
and
all
that
jet
or
Google
meets
and
then
be
sure
to
get
everyone
invited.
That's
already
on
those
lists
cool
all
right
thanks.
Everyone.