►
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 to Cloud Provider Azure - Rita Zhang, Microsoft & Craig Peters, Microsoft
In this session, you'll learn about how Kubernetes runs on the Azure infrastructure. We will cover development in the cloud provider over recent Kubernetes releases with support for new features in Azure compute like VMSS, networking like Standard Load Balancer, and storage. We'll also cover how all of this is tested and developed, and help you get involved if you would like to contribute.
https://sched.co/Uaic
A
We're
gonna
get
started.
Welcome
to
the
maintainer
session
for
cloud
providers.
U're
today
is
the
introduction
session
tomorrow.
There's
a
session
on
the
deep
dive
and
you'll
hear
me
talk
about
that
a
couple
more
times
today,
real
quickly.
We
want
to
deduce
ourselves
I'm,
Craig,
Peters
I'm,
a
p.m.
in
Azure
working
on
open
source
containers
stuff.
The
azure
cloud
provider
is
a
big
part
of
what
we
do
as
a
team
hi.
A
Doesn't
like
the
title
well,
we'll
fix
that
so
real
quickly.
What
we're
going
to
go
through
is
kind
of
introductions
to
the
team.
Happily,
a
lot
of
the
team
is
actually
here,
so
you
can
actually
meet
face
to
face
with,
like
the
people
who
do
this
stuff
and
make
it
real
we're
gonna
talk
about.
What's
happened
recently,
we're
gonna,
look
quickly
what
we're
planning
to
do
next,
what
our
focus
is
is,
then
we're
going
to
help
introduce
you
to
different
ways.
A
You
can
get
involved
and
contribute
as
much
as
you
want
to
contribute
or
understand,
what's
going
on
and
have
visibility
into
that
and
then
hopefully,
we'll
have
some
time
at
the
end.
For
some
questions.
The
first
thing,
though,
is
that
I
want
to
get
a
sense
of
who's
here.
It's
really
actually
pretty
interesting
to
me.
I
love
these
conferences,
because
it's
a
great
chance
to
see
people
I
haven't
seen
in
a
long
time
to
make
new
friends
to
understand
what
kinds
of
problems
people
are
facing
and
how
we
can
we
can
solve
them
together.
A
The
thing
I
observe
is
that
there
are
more
people
in
this
room
than
probably
all
of
the
as
your
maintainer
sessions
combined
through
kubernetes
history
and
so
to
me.
That's
a
really
exciting
sign
that
what
we're
getting
is
a
combination
of
people
who
are
using
it
and
people
who
care
or
people
who
are
running
into
problems
and
need
to
solve
those.
So
what
I'd
like
to
do
real
quickly
is
ask
a
few
questions
and
see
if
folks
can
raise
their
hand,
so
how
many
people
here
are
running
kubernetes
Ana's
your
today.
C
A
A
A
A
B
A
A
A
Thank
you
guys,
absolutely
and
and
that's
kind
of
a
segue
into
the
mission
right.
So
what
what
is
cloud
provider
as
you're
all
about
the
mission
of
cloud
provider
Azure,
is
to
make
as
you're
the
best
place
that
you
can
operate
kubernetes,
let's
just
say,
right
up
front.
We
know
that
we
have
some
work
to
do
right.
There's
there's
a
lot
of
things
to
accomplish.
A
We,
we
kind
of
highlight
that
this
team
sitting
up
here
and
some
over
there
are
contributors
to
cloud
provider
as
you're
their
core
Microsoft
team
members,
and
we
want
you
to
kind
of
know
who
they
are.
If
you
go
and
look
at
the
approvers
and
reviewers
in
the
owners
file
you'll
see
the
those
github
handles,
I
put
the
names
against
the
folks.
Quite
a
few
of
them
are
in
the
room,
feel
free
to
reach
out
and
talk
to
us.
So
one
question
peep.
A
My
task
is
what
happened
to
see
Gazoo
there
used
to
be
a
whole
separate
sig
for
each
cloud
provider.
So,
just
a
little
bit
of
history
that,
if
you
don't
know
already,
you
may
sorry
if
this
is
a
repeat,
but
the
cloud
providers
are
now
all
sub
projects
under
the
sig
cloud
provider
special
interest
group
each
each
one
of
these.
We
actually
have
three
projects
under
that
we
have
the
cloud
provider
as
er.
A
We
have
the
as
your
CSI
files
and,
as
you
see
a
CSI
disks
sub
projects
to
handle
all
the
different
nuances
of
how
you
deal
with
storage
as
well
we're
easy
to
get
a
hold
of.
So
please
reach
out
to
us.
Our
slack
channel
is
listed
here
cloud
provider
as
your
the
mailing
lists
are
certainly
available
there,
honestly,
not
very
active
right
now.
I
would
love
to
see
more
activity
there.
A
B
B
Another
really
awesome
feature
that
we've
added
bring
your
own
key
feature
capability
to
manage
disk,
to
allow
you
to
bring
your
own
key
for
disk
encryption
for
vmss
and
also
upgrading
the
SDK
to
35
and
then
again
update.
One
of
the
improvement
was
to
update
this
lock
per
VM
so
that
we
can
actually
do
concurrent
updates
for
attach
and
detach,
which
is
a
pretty
big
deal
and,
as
you
can
see,
here's
some
test
results
that
came
out
of
that
fix.
You
know
going
scaling
from
tendus
to
660
disk
used
to
take
45
minutes.
B
So
we
we
thought
the
API
throttling
issue
deserved
its
own
slide.
Specifically,
a
lot
of
PRS
went
into
this
specifically
switching
to
a
list
for
vmss.
Instead
of
get
for
each
instance
and
changing
the
cache
TTL
from
one
minute
to
10
minute
for
vmss
and
then
one
minute
to
15
minute
for
vm,
AAS
and
and
one
major
change
was
to
read
from
sale,
cache
for
certain
lists
and
get
operations
and,
as
you
can
see,
here's
just
a
graphical.
B
We
recognize
that
let's
go
to
155
now
it
takes
5600
requests
for
disk
crea
and
an
attached
one.
It's
idle!
It's
only
40
a
request.
Now
it
array
some
more
15
6.
Now
we
get
way
better
number
of
calls.
When
you
create
just
an
attach
now
you
get
34
52
number
requests
and
then,
when
the
clusters
idle
is
only
6
requests
holy
crap,
so
kudos
guys
yeah.
B
B
Currently
we
use
a
KS
engine
to
provision
a
cluster
on
Azure
and
run
all
the
eateries
across
those
clusters
and
if
you're
curious,
how
our
team
actually
maintains
a
lot
of
these
test
jobs
is
we
make
sure
that
the
person
who's
on
call
that
week
actually
looks
at
these
jobs
and
when
there
are,
the
jobs
are
failing.
Someone
is
gonna,
go
and
look
at
why
they're
failing
and
it
go
actively
fix
it
I'm.
So,
specifically,
we
have
e
to
e
pure
periodic
jobs
that
run
on
periodically
and
for
PRS
for
cloud
provider
a
sure.
B
So
that's
the
outer
tree
one
and
then
for
either
yi,
and
then
we
also
run
ete
on
PR
sr4,
kubernetes
kubernetes
entry
provider,
and
then
we
also
have
periodic
jobs
and
for
PRS
for
kubernetes
cavities
with
Windows
nodes,
and
then
we
also
have
jobs
for
dual
stack,
which
is
pretty
awesome
and
I.
Think
you're
gonna
hear
all
about
it
tomorrow,
at
the
keynote
and
last
but
not
least,
we
also
have
PR
and
periodic
jobs
that
actually
tested
disc
after
disc
CSI
drivers,
as
well
as
the
entry
drivers.
A
So
all
that's
great
we're
making
progress
we're
doing
a
bunch
of
stuff
to
improve
all
of
our
lives
to
you
know
basically
make
all
the
codebase
much
more
efficient
and
much
more
effective
with
the
underlying
infrastructure.
The
you
know.
The
key
thing
is
that
we're
moving
towards
this
ability
to
have
a
different
cadence
and
a
faster
cadence
about
the
cloud
provider.
You
know
right
now.
We
have
to
wait
for
a
patch
release
of
kubernetes
to
make
improvements
in
the
endo
in
the
cloud
provider.
A
The
next
is
that
we
are
undergoing
kind
of
a
renaissance
with
the
rest
of
the
teams
within
Azure,
so
the
growth
of
kubernetes
has
been
just
incredibly
fast
on
Azure
I.
Think
the
presence
of
all
these
people
in
this
room
is
a
testament
to
that,
and
one
of
the
things
we've
observed
is
that
not
all
of
the
azure
services
have
the
same
kind
of
prioritization
of
the
kinds
of
use
cases
that
kubernetes
has
kubernetes
as
an
API.
It's
this
closed
loop
control
system.
A
At
the
same
time,
we
need
to
improve
the
the
cloud
provider
itself
so
that
it
doesn't
make
unrealistic
demands
on
the
underlying
infrastructure,
so
we're
coming
together
and
we're
doing
that
by
essentially
working
first
with
the
vmss
team.
You
know
this
is
new
capabilities
on
Azure
and
we're
introducing
our
end-to-end
tests
into
their
engineering
processes,
and
so
you
know
in
the
past,
they've
done
a
release.
We've
discovered
something's,
wrong,
they've
gone
and
fixed
it.
Now.
A
What
we
want
to
do
is
we
want
to
discover
something's
wrong
before
the
vmss
does
a
release,
I
think
that
would
be
revolutionary
for
a
lot
of
us
and
that's
that's
coming
next.
At
the
same
time,
what
we're
doing
is
we're
working
with
them
as
they're
doing
design
for
new
features
so
that
we
can
understand
what
those
features
are
and
how
they
should
be
exposed
in
the
kubernetes
context,
what
value,
if
any
they
provide,
and
then
how
do
we?
How
do
we
represent
those?
A
A
We
have
increasingly
invested
in
cluster
API
and
the
provider
for
azor,
and
one
of
the
things
that
is
just
immediately
true
is
that
you
got
a
both
need
interface
with
VMs
s.
Both
need
load
balancers.
Both
you
know
both
need
to
deal
with
the
underlying
infrastructure.
So
we're
going
to
consolidate
a
lot
of
that
common
code
into
a
set
of
modules
that
so
that
we
have
one
place
to
make
these
improvements
and
make
these
fixes.
And
it's
been
reflected
in
this
out
of
treat
cloud
provider,
the
cluster
API
provider
and
everybody
gets
the
benefit.
A
So
this
is
a
just
sort
of
an
optimization
that
we're
gonna
do
we're.
We've
been
continually
investing
improved
and
then
testing.
We
have
some
pretty
rich
testing
already,
but
we're
now
gonna
do
a
lot
more
focus
on
performance
and
scalability
and,
most
importantly,
representing
real
workload
scenarios.
A
So
that's
kind
of
a
high
level
picture
of
what
things
we
have
on
the
plate
and
the
direction
we're
headed.
I,
don't
know.
One
of
the
things
I'd
love
to
get
out
of.
This
is
a
conversation
to
understand.
If,
if
these
are
the
right
things-
or
these
are
the
things
that
will
meet
your
needs
in
the
audience,
I'd
like
to
turn
this
into
more
of
a
discussion,
we'll
come
back
to
that
in
a
second
I,
just
want
to
make
sure
that
everybody
knows
how
to
participate
in
this
community.
A
One
of
the
things
that
we
strive
to
do
is
to
be
as
transparent
and
consistent
with
the
rest
of
the
open-source
community
as
possible.
You
know
we
conduct
our
meetings
in
the
open.
We,
you
know
post
everything
on
YouTube.
The
slack
channel
is
completely
open.
We
keep
our,
probably
all
our
issues
in
the
upstream
project
board.
You
know
it's,
it's
it's
a
standard
project.
The
one
thing
that
we've
seen
is
that
we
haven't
had
a
lot
of
active
participation,
part
in
mm-hmm
participation
in
the
meetings.
A
The
recently
we
changed
our
schedule
with
the
hopes
of
driving
up
some
participation
there.
You
know
there
used
to
be
bi-weekly.
At
the
same
time,
all
the
time-
and
it
wasn't
at
a
time
that
was
terribly
friendly
to
some
of
our
colleagues
in
the
asia-pacific
area.
So
what
we
did
is
we.
We
instituted
this
alternating
schedule
where
it's
on
alternating
Monday's,
either
early
in
the
morning
for
those
of
us
on
the
west
coast
of
the
US
or
in
the
later
in
the
day,
and
so
I'd
love
to
get
some
feedback.
A
Also
on
whether
or
not
this
this
is
working
out
for
folks,
so
I
think
we'll
come
back
to
that.
So
what
I?
Let's
have
a
little
discussion
about
the
plans?
You
know
the
the
last
thing
is
QA
right,
but
I
think
the
the
I'd
love
to
get
some
feedback
on.
What
of
these
things
are
important
to
people
here
or
if
there
are
things
that
are
missing
from
here
that
are
important
to
people
in
the
room.
A
B
D
I'll
ask
about
the
azure
stock
because
we're
gonna,
we
have
government
cloud
and
then
on-premise
the
azure
stack.
How
do
you
plan
on
in
those
plans
managing
that
if
you
have
a
team
that
may
say
work
on
both?
You
know,
knowing
that
the
stacks
gonna
lag
behind?
What
are
your
thoughts
on
sorry,
I
didn't
pay
the
services
that
would
go
into
the
the
cloud
provided
first
and.
E
F
So
if
you
are
deploying
any
a
specific
application
and
using
specific
features
which
may
not
be
present
in
Azure
SEC,
just
like
the
simplest
example,
is
as
your
files.
So
that
may
not
work
but
say
you
are
using
only
know
as
your
disk
a
not
will
exactly
similar
work.
So,
if
you're
deploying
that
application
on
Azure
and
you
try
to
deploy
and
I
just
Jackie
should
both.
A
C
Yeah,
I
can
speak
a
bit
to
that,
so
we've
been
investing
a
little
bit
in
the
cluster
api
as
your
provider.
Recently
I
don't
know
if
you've
noticed
activity
on
slack
and
on
the
repo,
but
I,
if
you
haven't
checked
that
out,
I
highly
encourage
you
to
go
hang
out
with
us
on
the
slack
channel
cluster
api
azure.
As
far
as
the
roadmap
right
now,
we're
finishing
up,
b,
1,
alpha
2
and
starting
to
think
about
v1,
alpha
3
there's,
so
we're
we're
sharing
that
room
met
with
the
overall
sync
cluster
lifecycle
project.
C
So
that's
probably
the
best
place
to
check
that
out.
I
think
their
sessions
there's
a
deep
dive
happening
tomorrow,
but
on
the
menu
I
guess
we
have
control,
plane,
management,
vmss
support
and
to
end
testing
framework
cluster
cuddle
v2,
so
improve
UX
in
general,
just
name
a
few.
But
yeah
we'd
love
some
feedback
on
that
and
if
you
want
to
help
us
and
contribute
that'd,
be
awesome.
A
lot.
B
C
A
B
I
also
want
to
mention
you,
if
you're
using
like
say,
pot,
identity,
project
or
virtual
cubelet
project
or
Kewell
flex,
volume
project
or
secret
or
CSI
project.
A
lot
of
the
maintainer
czar
here.
So
we're
happy
to
take
any
questions
about
roadmap
or
feature
or
feedback.
So
did
I
forget
any
projects.
Yeah.