►
From YouTube: Kubernetes SIG-Azure meeting ~ 06-14-2017
Description
http://bit.ly/sig-azure
https://groups.google.com/forum/#!forum/kubernetes-sig-azure
Meeting Time: Every other Wednesday at 9AM PST / 12PM ET
Zoom Link: https://deis.zoom.us/j/2018742972
A
All
right
welcome
everybody
to
the
first
sig
as
your
community
meeting,
which
I
am
extremely
happy
to
be
participating
in
and
leading
at
this
moment.
So
what
I'd
like
to
do
is
go
through
the
meeting
minutes
and
agenda
document
which
is
bitly,
slash,
sig
asher,
if
you
want
to
retroactively
Lee,
look
at
that.
If
you're
watching
this
recording,
this
recording
will
be
made
available
on
YouTube
under
the
sig
or
actually
I.
Think
it's
the
communities
umbrella
in
YouTube.
A
So
without
further
ado,
I
want
to
start
with
a
little
bit
of
introduction,
and
that
is
what
this
sig
is
really
trying
to
accomplish
and
to
make
sure
that
you're
in
the
right
place
and
also
after
that,
I
want
to
take
a
moment
and
just
have
everybody
just
briefly
introduce
themselves
where
they
work.
And
if
you
have
any
specific
thing
that
you'd
like
to
see
come
out
of
this
ache.
A
We
have
to
be
a
little
bit
judicious
with
time,
because
we
have
a
big
agenda
and
we
only
have
about
26
minutes
left
so
from
a
organizational
standpoint.
I
want
to
really
clarify
that
this
sig
is
to
discuss
the
evolution
of
kubernetes
running
in
Azure,
with
a
focus
on
Azure
container
service,
which
is
emerging
as
one
of
the
premier
places
to
stand
up
and
run
communities
in
Beco
systems
and
also
in
Azure.
So
what
we're
not
is
we're
not
a
place
here
to
really
hash
out
technical,
operational
or
implementation
issues?
A
So
it's
great
if
people
want
to
share
things
that
we
want
to
create
an
issue
for
or
maybe
search
for,
an
issue.
If
you
don't
know,
if
there's
one
out
there
already
for
what
you're
experiencing,
but
we
basically
don't
want
to
co-op
this
time
and
everybody's
valuable
a
time
for
AI
tactical
issues
or
I
really
want
to
focus
on
strategic
issues,
and
if
we
could
just
run
around
real
quick
and
have
people
introduce
themselves.
A
B
A
C
My
name
is
Seth
Cohen
I'm,
a
operating
manager
housed
in
E
ApS
group
also
came
from
the
the
day.
Attacks
ition,
as
you
can
tell,
by
the
suite
of
colored
squares,
that
almost
look
like
they're
they're,
Microsoft
group,
but
they're,
not
quite
bright
colored.
So
close,
yes,
I'm
interested
in
see
red
code,
not
Tigers
or
software
they're
here
and
then,
as
your
bull
Barajas,.
F
A
D
Well,
serve
I
am
some
working
from
the
nuance
communication
speech
and
AI
company
we've
been
using
to
binaries
for
an
hour
over
two
years
on
cam
and
we're
migrating
to
Asia
by
the
end
of
the
year.
We
should
have
everything
running
in
Asia,
so
really
trying
to
keep
the
focus
on
and
trying
to
align
with
the
best
practices
and
stay
up-to-date.
Great
thanks.
G
H
I
Hi
everybody:
this
is
Michael
Michael
from
aprende
I'm,
the
fig
leaf
for
sig
windows.
There's
a
lot
of
synergy
between
the
two
different
SIG's,
so
I'll
try
to
attend
here
as
much
as
possible
responsible
for
a
wide
variety
of
things
around
kubernetes,
including
the
big
windows
I
just
mentioned
some
of
the
Akismet
enterprise
toolkit
dependent,
delivering
and
and
then
a
few
other
features
that
we'll
have
folks
in
different
seats
across
the
board
of
the
kubernetes
project.
Ok,.
K
L
A
M
A
It
all
right
moving
right
along
let's
begin
to
the
agenda,
a
quick
set
of
working
agreements.
This
should
be
pretty
straightforward,
respectful
communication
in
all
matters.
So
if
you
have
a
disagreement
or
you
have
a
concern
or
something
that
you
need
to
raise
with
another
community
member,
even
if
it
gets
heated,
please
remember
to
be
respectful
and
listen
with
an
intent
to
understand,
as
opposed
to
respond.
A
A
There's
a
demo
slot
here,
I
think
what
I'd
like
to
do
is
come
back
to
that.
If
we
want,
if
we
have
time
but
there's
a
lot
of
other
things,
I
want
to
cover,
and
if,
if
you
want
to
be
on
a
future
demo,
please
at
the
top
I'm
going
to
put
it
in
the
agenda
right
now
put
in
a
date
and
what
you
would
like
to
do
for
your
demo
and
who
you
are.
A
And
that
will
be
see
a
lot
of
demos.
Some
people
are
doing
some
really
amazing
things
out
in
the
community
right
now.
Let's
get
some
of
those
here
amongst
this
good,
so
the
next
section
is
work
in
progress
and
work
in
progress
is
really
to
talk
about
initiatives
that
we
are
going
to
be
facing
as
a
community,
and
these
are.
This
is
really
forward-looking
too
right
now
and
be
version
1.8
1.9.
A
A
The
first
one
is
the
cloud
provider
controller
breakout
for
Azure,
and
this
is
a
fairly
big
chunk
of
work
and
that's
because
the
current
cloud
provider
implementation
is
officially
deprecated
and
the
impetus
is
really
to
provide
a
pluggable
instantiation
of
cloud
providers
as
opposed
to
having
it
in
core
and
there's
a
lot
of
benefits
to
that
it
does
post
some
really
interesting
challenges
that
we're
going
to
run
into
in
this
process.
I
put
a
link
in
here
for
ranchers
cloud
controller
manager
and
the
developer,
who
was
doing
this
work?
A
It
also
allows
us
to
prototype
and
test
other
approaches
of
the
implementation
which
are
pretty
cool.
So
the
details
are
here:
I,
don't
necessarily
want
to
go
into
all
the
history
and
back
story
behind
all
this.
But
what
I
would
like
to
do
is
just
open
the
floor
for
a
moment
and
if
anybody
has
any
specific
details
on
this
or
ideas,
concerns
questions.
Let's
get
those
documentaries.
A
Cool
all
right
moving
the
next
major
feature
that
we're
going
to
be
working
on
for
1.8
and
beyond
is,
as
your
manage
disks
there's
some
work.
That's
already
been
done
on
this.
It's
and
we
basically
have.
This
was
a
lot
of
this
work
too
was
done,
pre
cig
a
sure.
So
now
that
we
have
the
ability
to
manage
the
the
code
in
the
cloud
provider,
it's
giving
us
a
little
more
flexibility
in
terms
of
the
rate
of
velocity
on
getting
this
done.
A
So
this
work
is
being
broken
down
into
smaller
stories
at
the
moment
and
right
now
the
work
is
being
handled
inside
of
Microsoft
by
a
team,
and
we
are
definitely
going
to
be
having
some
PRS
for
review
soon.
There
is
a
PR
links
below
for
seven
one,
five,
eight,
which
is
Brendan's
PR
and
I'll.
Let
him
speak
to
that
if
he
needs
any
help
on
that,
but
I
wanted
to
raise
awareness
of
that,
and
that
is
specifically
the
implementation
of
the
mount
details
in
communities
itself.
J
M
Separate
but
related
news
that
may
not
may
not
mean
that
sternum
or
derivative
we're
breaking
the
cloud
provider
out.
I
got
the
update
of
the
address
decay,
PR,
all
green
and
ready
to
go
so
much.
The
window
opens
up
after
code
freeze
that
should
go
in.
Although
we've
made
very
immediately
like
rip
everything
out
in
favor
of
the
external
cloud
proof
right,
but
it's
it's
there,
so
that
adds
API
support
in
the
SDK
for
managed
disks.
Okay,.
M
I
mean
not
not
if
they
not
the
implementation,
but
just
the
API,
okay
and
also
was
needed
for
clustered
Federation.
So
one
thing
I
don't
actually
know
if
you're
retracting
Jason
but
I
should
have
hooked
up
with
Christopher
Shh
she
should
go.
She
should
go
something
like
that.
A
business
case
I
was
doing
collective
Federation
DNS
work
at
your
DNS
support
for
kubernetes
culture,
Federation.
L
M
M
A
A
F
The
controller
manager
to
market
is
not
available,
so
there,
the
BZ
that
the
PR
I
should
say
that
I
have
linked
there.
Someone
else
had
gone
through
and
added.
Some
logging
calls
around
some
get
as
your
API
calls.
I
was
hoping
to
get
some
eyes
on
that,
and
hopefully,
movement
and
I
was
actually
like
to
have
that
it
be
expanded
to
be
logging
around.
F
A
There
is
something
already
in
play:
called
a
cloud
provider
API
metrics
and
it's
implemented
for
basically
all
the
providers
except
Asher
and
we're
going
to
be
working
on
Asher
for
1.8
essentially
does
all
the
things
that
you've
got.
It
might
be
better
to
invest
in
actually
the
work
associated
with
getting
that
cloud
provider
metrics
for
for
Azure
in
place,
because
there's
already
the
implementation
so.
F
This
API
call
is
done,
was
logging
that
that
and
that
PR
for
Frazer
and
what
I've
suggested
for
AWS
and
GCE
is
literally
logging
around
the
API
calls
so
that
you
know
a
log
before
a
log
immediately
after
so
you
know
exactly
how
long
the
call
itself
took
usually
the
metric
stuff
that
I
saw.
You
would
see
an
API
call,
you
do
error,
checking
you
know,
plays
play
with
a
value,
and
then
you
do
a
metrics
call
to
say
this.
You
know
to
do
your
metrics
call
and
that's
not.
F
A
F
M
M
That
tells
you
like,
hey,
you
know
what
your
99th
percentile
latency
is
really
bad,
but
then
in
the
logs
you
have
specific
information
that
tells
you
about
the
specific
calls
at
which
low
so
that
you
can
actually
debug
them.
If
the
aggregate
stats
will
tell
you,
you
have
a
problem,
but
they
don't
really
help
you
debug
like
where
and
when
the
problem
happened,
which
is
I,
think
kind
of
the
analogy
or
try
to
get
at
yeah.
F
M
So
I
think
it's
definitely
worth
saying
something
like
hey.
You
know
what,
if
the
API
call
takes
longer
event,
I,
don't
know
if
it
needs
a
log
of
beginning
and
end
of
every
single
API
call.
But
if
we,
if
you
know
an
API
call,
it
takes
longer
than
five
seconds
or
something
like
that,
we
log
as
much
as
we
can
about
that
API
call.
M
D
Just
amount
of
Campinas,
if
you
need
someone
to
test
or
test
out
GPR
we're
getting
it.
Maybe
every
three
days
with
this
issue
on
randomly
between
10
to
20
nodes
they
go
down
for
15
minutes
could
let
just
as
stopped
responding.
So
we
do
see
this
quite
often.
So
maybe
if
you
need
us
to
or
me
to
try
your
new
logging
or
try
to
investigate
this
is
on
our
side,
something
that
we
tracked
as
a
top
priority
right
now,
because
we
do
lose
all
the
pods
on
those
nodes
due
to
the
dial
liveliness
probe.
N
M
J
M
J
J
M
A
F
That
one
is
me
as
well:
that's
another
kind
of
informational.
We
have
there's
inconsistencies
in
cloud
providers
around
what
happens
if
a
cloud
provider
node
gets
stopped.
Most
of
the
cloud
providers
end
up
with
with
the
node
being
deleted
from
kubernetes
GCE
is
a
notable
exception
and
that
it
makes
it
just
marks.
A
node
is
not
ready
so
that
the
issue
that's
linked.
F
So
that
the
current
there
is
current
issues
with
there's
issues
with
those
what
is
currently
implementing,
and
one
of
them
is
disc
right.
If
the
if
the
node
gets
stopped
and
then
it
gets
started
again,
it
goes
away
from
criminales
and
get
started.
It's
all
sorts
of
issues
with
persistent
disks,
and
you
know-
and
things
like
that,
so
there's
a
lot
to
kind
of
figure
out.
I
kind
of
thought
on.
That
is
a
good
place.
F
To
start,
we
kind
of
come
up
with
what
are
the
kind
of
attributes
of
an
instance
that
we
say
that
if
these
enter
a
tribute
you're
the
same
that
this
note
is
the
same
and
then
you
can
then
take
what
you
have
the
kind
of
attributes
you
could
apply
it
to
each
one
of
the
individual
cloud
providers
to
say
you
know
when
we
do
this
action.
All
of
these
metrics
or
all
these
attributes
state
is
saying
so
there,
for
we
should
treat
it
as
the
same
node.