►
Description
Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io
Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects
Intro: Kubernetes (AWS) SIG - Justin Santa Barbara, Google
sig-aws is responsible for the interfaces between kubernetes and AWS. The kubernetes project is itself increasingly moving momentum outside of the main kubernetes/kubernetes repository into subprojects. sig-aws is one of the leaders here, with 5 subprojects at the start of the year (broadly: cloudprovider, ALB, IAM, encryption, EBS) and likely at least one new one by Barcelona (EFS, FSx). We'll also discuss the cloudprovider extraction and our collaboration with sig-cloudprovider. We'll provide an update on the progress of those subprojects and discuss the future plans and roadmap together. And if you want to, we can describe how to get involved in these projects!
https://sched.co/MPhs
A
All
right
should
we
should
we
get
started:
okay,
hello,
everyone.
This
is
an
introduction
to
cicada
press
I
am
just
in
Santa
Barbara
I'm,
one
of
the
co-chairs
of
cicada
booths.
We
Mishi
is
another
co-chair.
She
is
not
able
she's,
not
a
coupon
this
year
or
Cavani
new
this
year,
Chris
may
or
may
not
show
up.
I
see.
You
said
she
had
a
conflict,
but
would
try
so
if
she
makes
a
dramatic
entrance.
You'll
know
why
this
is
an
introduction.
I
want
to
do
what
people
want
to
do
so.
A
What's
a
duress
does
some
of
the
where
the
activity
is
and
then
how
you
can
get
involved
in
the
sake,
I,
don't
know
if
that's
of
interest
to
people
or
where
the
people
would
rather
like
our
people
interested
in
contributing
or
our
people
more
interested
in
questions
about
AWS
should
I
more
on
the
contribution
side
more
on
the
questions,
side,
who's,
more
interesting,
contributions,
put
your
hand
up,
hey
and
who's.
One
sit
in
questions.
A
Put
your
hand
up
you
voted
twice:
okay,
I
will
go
through
very
quickly,
the
slides
then,
and
we
will
save,
will
go
through
and
who's
interested
in
what
the
sink
has
been
up
to.
I
should
say:
okay,
okay,
no
clear
winner.
There
I'll
go
through
reason
very
quickly,
so
goals
understand
a
six
structure,
the
projects,
the
sub
projects
that
plans
to
evolve
into
a
user
group
and
how
you
can
get
involved
in
sig,
a
device
or
actually,
in
general
this
most.
This
applies
to
any
sig
or
kubernetes
in
general.
A
Sig's
kubernetes
is
a
project
that
grew
very
rapidly.
Sig's
are
the
way
that
the
kubernetes
project
decided
to
sort
of
break
up
within
the
organizational
structure
so
that
it
wouldn't
be
a
handful
of
people
that
had
to
approve
everything.
Sig
AWS
is
therefore
responsible
for
AWS
specific
kubernetes
functionality.
There
are
many
other
SIG's,
SIG's
storage
does
storage
and
sig
networking
does
Network
sig
a
device
is
relatively
uniquely
a
user
group
there's
a
very
active
community
of
end
users
that
are
using
kubernetes
on
AWS
and
having
varying
experiences
and
sharing
their
experiences.
Talking
about
it.
A
The
primary
mechanism
for
doing
that
is
on
slack.
The
Szigeti
bus
channel
is
a
bit
of
a
fire
hose
at
this
point,
but
is
still
a
very
welcoming
place
there.
There
are
three
co-chairs
myself
initially
and
Chris,
and
we
organized
by
weekly
meetings.
Those
happen
at
Friday
on
Fridays
at
9:00
a.m.
Pacific.
Please
do
attend
they're
held
on
zoom'.
A
The
time
is
not
great
for
everyone,
but
it
is
not
as
bad
as
sometimes
there's
a
link
there
to
the
home
page,
which
you
can
go
to
their
PDF
version,
should
have
the
links
and
other
ways
you
just
search
and
there's
a
mailing
list
as
well.
But
honestly,
most
of
the
activity
happens
on
on
sack
and
in
the
bi-weekly
office
hours.
A
There
are
SIG's,
do
own
sub
projects,
and
so
there
are
a
number
of
sub
projects
which
are
under
the
security
West
purview
I
will
they
are
listed
here?
I
am
sure
you
couldn't
see
them
and
probably
observe
them
better
than
I
can
read
them,
but
essentially
ingress
controller.
I'd,
see
the
encryption
using
I
am
for
end
user
authentication,
a
bunch
of
CSI
drivers
and
an
interesting
one,
which
will
talk
a
little
bit
more
about
any
a
little
bit,
which
is
around
the
extraction
of
code
from
the
main
project.
But.
A
You'll
be
with
service
they'll,
be
yes,
sir
yeah,
so
that
so
the
the
lobe
and
you
can
create
a
service
of
type
load,
balancer
and
there's
some
code
which
currently
lives
in
the
main
repo,
which
will
create
that
that
was
sort
of
done
for
historical
reasons
before
the
kubernetes
project
got
too
big
and
so
that
what
that
will
end
up
in
this
bottom,
one
in
the
cloud
provider
extraction.
So
we'll
talk
a
little
bit
more
about
that.
But
that's
a
good!
That's
a
good
point!
There's
another
class
of
code
which
actually
you're
right.
A
We
see
a
dos
does
own
which
lives
in
the
the
kubernetes
repo
itself
for
now.
Okay,
so
in
one
month
in
115,
there's
some
bins
and
progress.
Promotion
of
NLB
support
the
entry
and
I'll
be
support
to
beta
some
alpha
features
on
these
external
projects.
Some
beta
and
some
alpha
projects
on
those
and
two
of
those
new
CSI
drivers
to
those
sub
projects
to
those
CSI
driver
projects
are
actually
new.
A
This
time,
that's
a
fairly
large
surface
area
for
any
sig,
and
so
one
of
the
things
we're
trying
to
do
is
be
more
sort
of
user.
Driven
and
make
sure
that
we're
listening
to
what
people
are
actually
using
or
how
people
are
using
kubernetes
on
AWS
and
so
initially
did
send
out
a
survey,
and
we
did
receive
thirty-one
responses,
which
is
good
but
I,
hope
everyone.
There
are
more
than
31
people
in
this
room
and
I
hope
in
future.
A
You
will
respond
so
we
can
have
a
more
representative
sample
at
the
next
at
the
next
survey,
but
there's
some
interesting,
interesting
things:
I'm,
not
gonna,
read
through
all
of
this,
but
essentially
the
idea
was
ingress.
Controller
has
done
a
bunch
of
stuff
particular
on
e
to
e,
which
is
super
important.
The
way
that
all
kubernetes
projects
maintain
quality
or
the
primary
way
by
which
they
maintain
quality
is
to
ensure
that
there
are
good
ete
tests.
A
A
Thank
you
thanks
and
initially
all
the
volume
drivers
were
entry
in
the
kubernetes
communities,
repo
that
proved
to
be
unmaintainable,
and
so
the
CSI
project
is
aiming
to
create
a
common
interface
by
which
anyone
can
create
a
driver
so
that
you
can
write
a
driver
for
EFS
or
EBS
or
FS,
X
or
any
or
NFS
or
any
other
file
system
that
you
want
and
it
doesn't.
You
don't
have
to
go
through
the
whole
kubernetes
integration
process.
You
can
develop
it
at
your
own
pace
outside
of
kubernetes
and
so
there's.
A
Some
progress
in
terms
of
that
that
moved
from
entry
code
to
all
the
drivers
being
in
external
in
via
the
CSI
interface
is
progressing
and
part
of
that
progress
is
to
make
sure
that
we
have
all
these
drivers
in
place
for
for
the
drivers
that
we
for
the
file
systems
that
we
want
and
progress
is
being
made.
That
is
a
great
area
of
contribution
for
anyone
that
wants
to
get
involved,
even
if
it's
as
simple
as
testing
some
of
these
out
of
free
providers.
A
A
So
the
github
flows
and
mechanics
are
wonderful
but
sort
of
start
to
break
down
at
a
certain
scale
and
the
people
that
were
trying
to
keep
up
with
it
simply
no
longer
keep
up
with
it
and
the
way
that
they
decided
to
scale.
The
project
was
essentially
to
split
that
big
mono
repo
into
a
whole
set
of
smaller
repos.
A
So
new
code,
new
cloud
providers,
for
example,
go
into
sub
projects
or
separate
projects
ideally
owned
by
a
cig,
and
that's
how
we
have
seven
sub
projects
of
our
own
and
in
addition,
we
wanted
to
try
to
take
some
of
that
existing
code,
which,
of
course,
is
still
growing
out
of
the
KK
out
of
that
main
repo
in
two
sub
projects.
So
that
is
the
CSI
drivers
and
there's
also
the
actual
extraction
of
the
core
cloud
provider
code.
A
A
A
Yes,
it's
a
great
question,
so
the
the
intention
is
that
there
will
be
no
AWS
specific
code
in
KK
and
no
GCE
specific
code,
no
clubber
code
at
all.
Instead,
that
code
will
so
kubernetes
works
with
this
idea
of
controllers.
So
a
controller
observes
the
state
of,
for
example,
a
service,
objective
type
load
balancer
and
will
go
and
create
an
EOB.
A
A
A
You
could
install
the
GCE
cloud
controller
manager,
I,
don't
think
it
would
do
you
any
good,
but
and
then
in
theory,
you
can
update
the
ADA
Boesky
Cloud
Controller
manager,
totally
separately,
or
if
you
have
a
private
cloud,
you
can
write
your
own
private
cloud
cloud,
controller
manager
and
never
share
with
anyone
else
in
the
world.
So
you
don't
need
to
go
through,
like,
though
the
flipside
of
older
maintainer
is
having
to
having
to
deal
with
all
this
coding
in
KK
is
that
everyone
had
to
push
all
their
code
up
into
this
central
place.
A
A
Sorry,
the
question
of
people
didn't
hear,
which
is
another
excellent
excellent
question,
is
how
I
get
installed
in
kubernetes.
These
are.
These
are
the
gracious?
Yes,
this
is
another.
One
of
these
question
mark
question.
Mark
question
marks
the
there
are
efforts
to
figure
it
out
plus,
if
there's
some
work
going
on
to
figure
out
Takota,
that's
around
add-ons
or
is
this
an
add-on,
but
essentially
creating
a
package
that
is
more
than
just
the
core
components.
A
Essentially,
there
will
likely
be
an
ongoing
piece
of
tooling.
There
will
be
a
piece
of
tubing
that
will
will
do
that
for
you,
whether
it's
eks,
whether
it's
cops,
whether
it's
when
you
write
yourself,
but
essentially
that
will
be
the
role
of
that
piece
of
tooling
and
we
have
to
produce
artifacts
such
that,
particularly
if
you're
writing
it
yourself,
which
a
lot
of
people
still
do
that
it
is.
It
is
impossible
for
you
to
do
that
and
doesn't
become.
You
know
exponentially
harder,
as
we
take
the
number
of
components
from
four
to
ten.
A
Okay,
the
this
project
says
the
club
products.
Traction
project
has
been
going
on
for
multiple
years
and
it
is
not
done
yet,
but
I
do
think
it
is
at
the
end
of
the
beginning.
In
that
now
the
code
is
in
staging.
The
hope
is
that
now
temporarily
other
AWS,
the
cloud
provider
Davis,
will
vendor
that
code
temporarily,
and
then
we
will
just
stop
ven
during
and
move
it
all
into
the
other
place.
A
There
there
are
mixtures,
there
are
some
places
where
it's
where
it's
intricate,
like
the
leader
election,
tough
stuff,
is
very
difficult.
There
are
other
places
where
it
it's
not
nearly
as
bad,
and
yes,
if
you
are
interested
in
that,
I
think
the
right
way
to
go
through.
That
is
to
go
to
the
signal
provider
meetings
which
are
Wednesdays
at
around
lunchtime.
Sometimes
in
some
time
zone
there
are,
they
are
sort
of
driving
that
effort
and
yeah.
A
A
That
comparable
chair
experience
on
vSphere
or
OpenStack
or
wherever
it
is,
and
so
that
is
one
of
the
big
per
views
of
signal
provider
is
to
try
to
establish
this
sort
of
common
behavior
across
platforms
and
as
such,
sig
AWS
as
owner
of
code
is
going
to
fold
into
the
sig
cloud
provider
sake.
So
I
don't
know
if
in
a
year
we
will
have
a,
however,
so
that
that
group
is
going
to
do.
A
You
know
common
behaviors
lots
of
testing
and
documentation
stuff
and
trying
to
figure
out
a
lot
of
these
release,
issues
that
are
indeed
challenging
as
a
end
user
group
Sogeti
bus
is
gonna
continue
to
exist,
although
it
may
have
a
different
name.
I
want
to
call
it
hug,
AWS
and
I'm,
going
to
keep
saying
kubernetes
user
group
as
many
times
as
I
can
and
hope
that
it
sticks
I'm
told
it's
gonna
be
AG
AWS,
your
GA
degress,
which
is
disappointing
to
me
personally.
A
Okay,
so
I
think
this
is
like
my
drawing
down
of
like
very
many
ways
you
can
get
involved
in
the
project
and
there
you'll
see
that
a
lot
of
them
are
not
at
all
about
code
and
the
slides
are
available.
But
to
take
yes,
I
think
one
of
the
sort
of
the
meta
points
that
I
would
do.
Is
you
know
it's
not
all
about
code?
Even
asking
questions
is
super
valuable,
trying
things
out,
debugging
things
like
joining
the
sig
meetings.
Sharing
your
experiences
as
a
actual
user
of
kubernetes
on
AWS
is
super
important.
A
It's
a
little
bit
in
the
cups
one
earlier
about
how
a
lot
of
the
people
that
join
cups
in
like
the
second
wave
of
contributors
had
a
much
better
understanding
of
cops
than
I
did
even
though
I
sort
of
was
one
the
people
that
started
that
project
because
cops,
evolved
and
I
had
not
evolved
with
it
and
they
has
new
contribute.
New
people
were
using
it
in
a
way
in
ways
that
I
hadn't
anticipated,
and
that
was
really
good.
A
So
you,
as
newer
people
that
are
coming
in
and
seeing
this
with
new
minds,
are
probably
more
like
better
reflective
of
where
the
project
is
today
and
the
other.
The
other
meta
thing
I
would
say
here
is
the
cake.
A
repo
is
a
terrible
place
to
start
contributing.
I
would
start
in
the
sub
projects.
So,
like
the
great
thing
about
all
these
sub
projects
is
that
they
all
probably
are
much
more
much
much
smaller
ponds
that
are
much
friendlier
and
easier,
and
it's
much
less
likely
that
your
PR
will
get
lost.
A
Where
is,
despite
all
our
best
breaking
up
KK
into
a
whole
bunch
of
other
repos,
it
is
still
far
too
easy
for
the
arts
to
get
lost,
and
unless
and
that
sort
of
people
know
you
or
you
know
who
to
poke
or
the
PR
is
basically
perfect
from
the
start.
It's
pretty
hard
to
get
code
merge
into
there
and
there's
also
a
ton
of
history
in
there
like
why
things
I've
done
the
way
they
are
so
I
would
definitely
encourage
you
to
look
in
the
sub
projects.
A
We
have
a
lot
of
coders
at
this
point,
there's
a
pretty
big
conference.
There
are
plenty
of
people
that
can
code
things
the
hard
way
of
sort
of
figuring
out
what
to
code
or
like
what
is
broken.
But
if
you
do
want
to
code,
then
you
know
start
simple:
everyone
starts
with
typos
or
Docs,
it's
okay,
some
people
don't
like
doing
typos,
PRS
I,
actually
don't
mind.
It
I
think
it's
an
easy,
easy
all
rounds,
a
good
place
to
get
started,
started
and
familiar
with
the
there's.
A
A
lot
of
workflow
the
when
you
submit
a
PR
or
like
all
these
tests,
fire
off
and
like
some
of
them,
will
fail
randomly
even
on
your
typos
PR
and
you'll,
be
like
oh,
why
did
I
break
something
with
my
typos
and
probably
is
just
a
test
flake,
and
so
it's
a
nice
way
to
get
going.
Can
we
doing
cops?
You
can
always
I'm
sure
some
PR
some
projects
are
welcoming
to
it.
A
I
some
of
them
are
less
happy
about
it,
but
if
you,
if
you
do
a
bunch
of
type
of
fixes
in
one
PR,
everyone's
gonna
be
pretty
happy
with
that
again
starting.
The
extremities,
like
the
newer
repos
that
are
that
are
not
a
they're
out
of
tree
and
then
be
willing
to
have.
A
girl
like
go,
is
a
relatively
easy
language
to
pick
up.
There
are
a
couple
of
gotchas,
but
everyone
hits
them.
You
also
will
incorrectly
close
over
a
loop
variable.
I,
don't
know
if
anyone
has
any
go
barometer
has
not
hit
accidentally
done.
A
That
I
do
it
regularly,
even
now
and
and
I.
Just
just
try
and
people
will
probably
help
you
help
correct
here.
You're
go
code
and
and
you'll
learn
the
both
go
and
the
kubernetes
style,
and
then,
even
if
your
PR
isn't
merged
you,
you
have
achieved
something
you
have
helped
and
if
you
are
going
to
tackle
something
complicated,
which
you
know
we're
sort
of
like
moving
down
the
tree
or
the
they're
moving
up
the
levels
then
do
talk
about
it
with
people.
A
First
like
talk
about
it
on
slack
or
in
the
github
issue
that
you're
looking
at,
and
then
this
is
my
personal
favorite,
which
is
to
run
multiple
PRS
in
parallel,
like
I,
always
treat
my
PRS
as
cattle,
not
pets,
I
apologize
that
offends
anyone.
Sorry,
but
like
it's
easy
to
get
personally
invested
in
a
PR
and
then
like.
Why
isn't
this
person
reviewing
my
PR?
A
They
hate
me
I
hate
this
person
what's
going
on,
but
if
you
have
ten
PRS,
then
it
doesn't
really
matter
you're,
like
oh
cool,
that
PR,
that
someone
got
respect
me
on
that
one,
so
runt
NPR's
in
parallel
and
see
what
what
flies
and
if
your
type
of
PR
gets
ignored
in
one
repo
but
happily
accepted
another
than
that's
where
you
should.
Probably
you
know
focus
your
efforts.
I
think
yeah
I
wanted
to
save
a
lot
of
time
for
questions.
I.
A
Think
we've
done
that
we
can
also
go
into
more
detail
about
any
of
the
things
that
we
glossed
over,
but
you
know
if
we
have
time
we
have
time
for
Q&A
deliberately.
Please
here
are
some
things
I'd
love
to
hear
like.
Where
do
you
wanna
get
involved?
Do
you
have
any
immediate
questions
or
feedback
from
using
communities
on
AWS,
and
maybe
are
we
even
missing
a
whole
sub
project?
Is
there
like?
We
have
seven
should
we
have?
We
have
yeah,
we
have
seven.
Should
we
have
eight
all
right
who
wants
to
go?
First.
A
B
C
C
So
you
could
one
of
the
projects
that
I'm
an
authorizer
or
an
approver
for
is
the
ATO
Siam
Authenticator
that,
for
example,
you
could
install
on
your
kind,
cluster
or
docker
from
well
now
you
could
might
be
able
to
do
it
doctor
for
Mac
I,
don't
know,
but
on,
like
mini
cube,
any
of
those
kind
of
things
it's
just
a
web
hook
from
the
API
server.
It
just
sends
our
token
across
and
does
authorization
you
could
technically
contribute
to
that
from
nothing.
B
A
Technically,
you
can
send
up
PRS
without
having
tested
them.
First
particular
you
write
work
in
progress
on
them
and
they
will
run
a
suite
of
ete
tests
on
them,
which
I
may
or
may
not
have
done
frequently
in
the
past.
So
you
can
do
that
without
any.
We
can
do
that.
Docs
yeah,
Docs
Doc's
better
past,
but
here
we
are
authenticated.
It
would
love
Doc's
contributions
that
good.
D
This
is
actually
cops,
feature
question,
but
it's
tightly
connected
to
what
you
doing
right.
Okay
and
I
haven't
seen
this
in
the
rollout,
so
I
wanted
to
ask.
We
have
really
hard
times
with
security
people
because
they
say
listen.
There
was
a
hole
in
docker
and
I
want
to
have
it
tomorrow
and
with
cops.
It
takes
up
two
months
to
get
this
done.
Is
there
anything
that
we
could
do
to
try
to
I
know,
maybe
automate
the
process
and
have
cops
support
every
doctor
release,
just
as
it's
coming,
yeah.
A
It's
a
it's
a
great
question,
I
think
that's
it's
sort
of
like
we
tried
to
rotation
between
eight
of
us
and
cops,
but
anyway
the
in
the
cops
project.
We
do
explicitly
list
some
darker
versions.
We
could
definitely
support
allowing
you
allow
you
to
put
a
URL
in
there,
maybe
or
assume
guess
a
the
URL
for
a
docker
version,
because
I
think
they
follow
a
pattern.
It's
it's
not
as
simple
as
that.
There
is
also
there's
also
container
D,
coming,
which
I'm
a
big
fan
of
and
we'll
be
a
little
bit
more
regular,
I
hope.
A
E
A
C
So
we've
sort
of
done
this
a
little
bit
with
CSI,
but
it's
being
done
at
the
docs
level.
It's
not
really
coordinated
at
all
yeah,
which
is
which
goes
to
exact
what
you're
saying
about
Sigrid
about
sig
release,
but
in,
for
example,
all
three
CSI
drivers
there's
a
table
of
what
version
supports,
what
from
the
CSI
perspective
and
what
kubernetes
version
and
that
installs
in.
A
Figuring
out
its,
it
is
a
real
challenge:
I'm
figuring
out
how
we're
going
to
test
those
things
to
the
same
degree
that
we
were
testing
kubernetes
previously
is
a
great
problem.
I
think
it
would
be
nice
if
we
had
this
this
bat
bundle
or
llaman
file,
which
listed
all
the
things
and
that
we
ran
the
tests
based
on
that
I,
don't
see
how
we
could
I,
don't
see
how
we're
gonna
get
the
test
to
pass
without
having
that.
So
then
at
least
give
me
some
hope,
but
we
might
split
the
testers
up.
A
C
There
was
talk
at
one
point
about
SIG's
having
multiple
roles
being
added
to
them.
They
came
up
in
the
community
call
earlier
this
year.
Do
you
know
anything
about
that?
Multiple
roles
is
in
was
in
like
so
we
there's
there's.
Obviously,
some
cigs
has
that
have
like
a
tech
lead
for
him,
like
Micah's,
for
a
cloud
provider,
but
there
was.
There
was
talk
that
they
were
trying
to
add
in
things
like
release,
managers
to
individual
SIG's,
to
kind
of
offload
that.
A
The
last
I
heard
they
were
like
there
are
people
that
are
better
at
technical
things
that
are
then
they
are
at
organizational
things,
naming
no
names,
and
there
was
talk
of
having
like
some
of
the
streams
like
a
tech
lead
and
a
chair
distinction,
I
haven't
I
can
imagine
more
of
those
distinctions
would
be
valuable,
but
it's
not
clear
to
me
that
a
sig
can
do
so.
If
anybody
wants.
A
B
A
This
is
a
good
question.
We
don't
have
that
as
a
feature.
I
think
I
think
that
in
general
in
the
Cobras
ecosystem,
we
try
not
to
build
something
into
the
tool
if
we
could
build
it.
As
say
Damon
said:
oh
so
like
could
this
be
a
daemon
set
yeah
and
therefore
there
was
a
daemon
set.
Could
it
work
identically
on
e
KS
as
it
works
on
cops
or
identically
on
any
tooling
that
you're
using
and.
F
I
want
to
actually
that
so
that
is
the
daemon
set.
That
is
CSI.
So
that's
what
the
CSI
driver
will
do
for
you.
It
will
provision
the
three
CSI
drivers,
the
EBS,
CFS
and
FS
x
for
lustre
will
provision
the
EBS
or
the
file
system.
But
it's
not
it's
not
when
the
node
comes
up
per
se,
its
separate
through
that
daemon
set
so.
A
B
A
B
B
The
root
guys
are
planning
to
do
something
like
that.
They
want
to
run
the
hoses
on
top
of
Peavy
claims,
but
still
it's
a
feature
request
and
they're
planning
to
do
that
till
that
time,
I
think
only
having
some
kind
of,
but
having
a
demons
at
how
they
we
can
do
something
like
we
can
have
a
process
running
as
demon
said
and
call
it.
B
A
Guess
I,
so
perhaps
I
think
there
was
like
a
bunch
of
approaches,
but
yeah
the
the
short
answer
is
we're
not
planning
anything
on
cops.
The
longer
answer
is
we
would
love
to
find
some
that
isn't
if
it's
not
if
it's
something
that
we
don't
have
to
build
into
cops
or
if
we
can
leverage
suicide,
then
that's
the
way
that
we
try
to
do
things
in
the
community
like
we
try
to
say
like.
Can
we
not
make
it
specific
to
a
tool
and
and
be
applicable
everywhere?
G
Potentially
just
lightly
loaded
question,
given
the
number
of
projects
you
already
are
responsible
for
so
for
reference,
I've,
I've
sort
of
been
semi
involved
with
sig,
auto
scaling
for
quite
a
while.
We
run
all
of
our
workloads
apart
from
control
plane
on
spot
instances.
Cluster
also
scaler
does
not
like
that,
and-
and
it
is
one
of
those
tools
where
there's
a
loader
vendor
or
there's
a
lot
of
cloud
provider
specific
code
in
there
and
that
and
there's
a
lot
of
functionality.
G
A
A
If
we
can
expose
those
things
through
the
kubernetes
api,
we
we
can
hopefully
in
the
same
theme
we
can,
as
we
talked
about
with
you,
know,
getting
it
across
all
the
tools
we
can
have
cluster
autoscaler
work
across
all
those
cloud
providers.
Api
is
ongoing
and
is
not
at
there
yet
I.
Don't
there
is
a
proposal
for
an
integration
between
autoscaler
and
cluster
api?
I
don't
think
it's
been
merged.
I
think
there
are
some
challenges
with
races
around
some
edge
scenarios.
A
A
D
Merging
those
groups
doesn't
also
mean
that
the
feature
sets
among
different
cloud
providers
will
be
aligned
or
does
it
leave,
for
example,
cops
as
a
tool
which
is
like
purely
mate
with
Amazon.
You
can
do
whatever
you
want.
You
don't
have
to
wait
to
get
an
azure
support,
GOI
cloud
support,
Alibaba
crowd,
support,
and
so
on.
So.
A
I
mean
cops,
does
support
multiple
clouds
and,
and
the
with
that,
that's
sort
of
separate
I.
Think
to
your
core
question,
which
is
yes,
the
goal
of
sig
cloud
provider
is
to
say
we
I
think
one
of
the
things
that
prompted
it
was
there's
this
big
debate
about
what
happens
when
you
stop
an
instance
on
AWS,
and
it
turns
out
that
when
you
stop
an
instance
on
other
cloud
providers,
we
in
criminal
design,
we
do
one
thing
and
when
we
stop
it
on
AWS
people,
we
do
something
else.
A
A
A
The
sakes
is
that
we
have
too
many
sinks
for
people
to
keep
track
of,
so
I
can
keep
track
of
the
the
4
or
5
that
I
am
involved
in,
but
there
are
a
lot
more
than
4
or
5
I'm
told
and
the
so
that
is
part
of
sync
cluster
life's,
like
which
really
deals
with
management
of
clusters
and
like
how
you
bring
them
up
how
you
update
them,
how
you
destroyed
them,
but
isn't
really,
as
concerned
with
per
cloud
abstractions.
Oh
yeah,
but
there
is
a.