►
From YouTube: SIG Azure Community Meeting 20180613
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).
B
C
B
B
Is
there
so
we're
slowly,
adding
them
as
D
being
added
to
that
cloud,
controller
manager
and
then
we'll
begin
to
see
them
we'll
begin
to
see
them
in
the
car
provider,
hand
that
a
cloud
controller
manager
I,
don't
think
it's
in
a
state
where
you
can
even
play
play
with
it
now
as
a
test
cluster
or
you
can
you
can
just
kick
the
tires,
create
the
cluster
use
it
and
just
delete
it
right.
I
do
believe
that
next
release
things
will
be
clearer.
B
C
B
B
While
this
worked
really
well
for
a
while,
it
has
proven
that
it's
hindering
the
progress
of
the
cloud
providers,
such
as
a
juror
in
emulation,
Google
and
OpenStack,
and
they
want
to
add
features.
But
the
release
cycle
of
three
months
is
is
quite
too
long
for
some
features
to
actually
make
it
to
the
cloud
provider.
So
there
has
been
some
sort
of
a
unanimous
agreement
long
ago
on
moving
this
piece
of
code
outside
right
to
run
as
it.
C
B
I'm
gonna
send
you
the
link
later
on
a
tracer.
What
we
did
is
we
moved
all
the
cloud,
the
volume
drivers
into
our
ebook
under
agile
username,
that
has
all
the
volume
drivers
that
we
support
all
the
Flex
folds,
all
the
CSI
stuff
in
one
repo.
The
rationale
has
one
place
for
all
everybody
to
actually
look
at
comment,
sent
pr's,
have
comments
and
so
on.
If
one
of
the
CSI
providers
grow
too
big,
which
is
gonna,
create
it
into
its
own
repo
right,
okay,.
B
B
This
will
also
the
the
this
will
also
include
that
list
of
everything,
any
volume
that
specific
title
will
be
on
this
people,
all
right,
I
link
it
to
the
dock,
once
once
I,
once
I'm
opening
the
seaside
windows,
no
progress
yet
reason
behind
that
is
we've.
Some
of
our
team
members
or
windows
is
specific,
just
up
move
to
a
different
place
and
things
are
slow,
but
we
will
recover
this.
B
B
The
agile,
the
agile
exporting
host
package-
I,
don't
know
if
you
guys
following
what's
happening
right
now.
We
have
so
many
many
many
components
that
are
dependent
on
the
configuration
of
agile,
5,
all
right,
some
controllers,
the
core
provider
itself.
The
hazard
of
Jason,
has
the
specs
that
tells
anybody
who
wants
to
use
the
cluster
identity.
B
What
is
it
so
I
think
it's
all
original.
Also
this
approach
was
called
so
it
has
a
servant
like
is
this
customizing
service
principle
or
MSI
or
emsi?
Alright,
and
the
cloud
provider
has
code
that
entry
gates
these
settings
and
accordingly
acquires
the
token
now
that
means,
if
I
decided
to
go
tomorrow
and
had
a
new
feature,
took
the
clock
providers
such
as
supporting
EMS
I
anybody
who's
depending
on
this
header
file,
is
gonna
break.
B
A
C
B
B
A
B
C
B
Availabilities
skill
sets
and
believe
it
is
on
so
the
idea
is
vm
skill
sets.
If
you
go
to
the
documentation
today,
gave
you
99.95
availability
for
your
v
ends
and
when
you
do
like
three
or
more
two
or
more
alright,
you
have
to
go
through
the
docks
and
just
make
sure
that
what
you
choosing
and
they
will
meet
your
targeting.
The
idea
is
very
simple:
every
vm
skill
sets
the
data
centers
in
Azure
is
they
have
region
and
then
a
data
center,
multiple
buildings?
B
If
you
will
and
then
you
have
buildings,
each
building
or
each
unit
right,
it's
continued
I,
don't
own
to
this
world,
as
if
you
cooling
power
and
an
that
alright,
when
you
go
today
and
create
availability
of,
am
skill
sets
your
all
your
VMs
are
actually
in
one
building
of
those
solar
wall
zone.
It's
the
same
power
supply.
It
relies
on
the
same
cooling,
the
source
and
so
on.
If
a
failure
happened
it
you
are
not
in
a
good
position.
What
availability
zones
does
as
the
they
spread
out
your
VMs,
the
VM
skill,
sets.
A
A
B
That's
that
that's
the
long
discussion
I'm
having
with
the
VM
and
VM
skill
sets
people
over
okay.
Now,
if
you
go
to
the
VM
skill
sets
availability
zone
and
the
abilities
on
documentation,
one
of
the
things
you
will
clearly
see
as
storage,
IP,
external
IP
load,
balancers
and
and
most
of
the
core
services
and
I'm
talking
about
the
actual
fabric
of
your
of
our
life
staff.
Computer.
C
B
I'm
gonna
say
something
very
clear
here:
the
storage
folks
starting
manage
disks.
They
did
the
copy
moon
right
to
that
was
the
ability
to
switch
zones
and
switch
fault
domains
as
the
VM.
Let's
say
you
have
two
VMs
right
on
twofold
fault
domains,
and
then
you
attach
this.
The
disk
em
manage
just
em
into
the
first
machine,
which
happened
to
be
on
one
for
me.
If
I
hold
the
mean
domain,
think
of
it
as
a
rack,
if
you
were
all
right,
it
takes
a
rack.
B
Actually,
the
fundamental
answer
now
Copernicus
move
the
pot
from
machine
to
machine
B,
which
happened
to
be
in
a
different
fall
to
me.
Now
the
disk
leads
to
move
as
well.
Storage
handles
the
movement
between
full
domains
for
us,
so
we
take
care
of
that.
My
understanding
is
this.
Magic
happens
also
for
zones.
B
A
Seen
lots
of
bugs
in
the
past,
with
AWS
in
terms
of
whatever
their
equivalent
functionalities,
are
called
in
terms
of
disk
scheduling,
I've
just
seen
a
lot
of
plate
places
where
misconfigurations
or
bugs
in
the
cloud
provider
was
causing.
You
know,
discs
from
a
different
region
or
a
different
whatever
to
be
to
be
attached
and
I
guess
aw
iced
wasn't
handling
that,
but
it
sounds
like
maybe
Asher
is
hiding
some
of
that
complexity
from
us.
B
We
have
a
guy
in
Philly,
actually
thinking
about
the
most
pin
table
now
driving
the
ability
zone,
and
then
we
have
different
people
working
on
different
pieces
of
the
zone.
So
we
have
one
person
on
the
networking
one
person
on
the
desk
to
make
sure
that
all
the
feature
as
all
the
behaviors
do
they
work
awesome.
B
B
C
B
C
B
Note
failure
or
a
note
deletion.
Alright,
that's
really
what
it
is
right,
so
for
note,
failure
and
notary
stores
they
complete
the
knobs
around
affection
pot
eviction
and
all
of
that
taken.
The
controls.
Allow
you
to
decide
rate.
If
the
note
stayed
in
failure,
because
once
the
notes
risk
to
restores
it
goes
into
not
ready
mode
according
to
a
heartbeat.
So
if
using
five
stock
and
heartbeat-
and
you
miss
five-
that's
25
seconds-
then
it
will
be
flagged
as
not
ready.
B
Then
the
clock
for
eviction
starts,
so
you
may
have
average
setting
for
a
if
two
minutes
then
start
affecting
the
pots
over
there
and
move
them
to
different
different
note.
That's
like
in
a
typical
scenario,
what
happened
in
1:10
and
we
want
to
support
in
111.
It's
a
very
simple
feature.
You
feature
like
shut
down
teens.
So,
as
the
note
goes
to
the
shutdown
which
we
tainted
by
the
cloud
providers,
so
the
schedule
doesn't
place
things
over
there
because
it's
going
shot
down
and
then,
as
it
comes
up,
remove
that
taint.
B
Then
the
last
phase
of
note
completely
go
away
once
the
note
coat
goes
away,
then.
After
all
the
clock
and
so
on,
then
it
will
remain
as
not
ready.
Alright,
my
understanding
is
and
conky
you
can
correct
me
if
I'm
wrong.
If
you
restart
the
controller
manager
sort
of
like
reset
all
the
cash,
then
any
note
that
doesn't
really
register
itself
again
will
not
be
included
as
not
ready,
not
like
it
will
just
disappear.
So.
C
And
then,
why
am
I
asking
is
mostly
because
with
scale
said,
with
a
s
before
everybody
said,
the
note
even
though
goes
away,
it
can
come
back,
would
have
the
same
name,
but
now,
with
the
scale
set,
we
sort
of
lose
that
control
and
the
nodes
will
in
fact
get
a
new
name
right,
and
so
this
is
where
I'm
trying
to
see
are
we
gonna
end
up
with
over
a
period
of
six
months
with
like
a
thousand
node
in
a
naturally
state?
If
we
don't,
we
start
the
controller
manager
or
you
know,
yeah.
B
C
B
C
B
Basically,
the
idea
is,
if
you
have
a
ship
and
this
ship
left
for
250
years
and
every
day
you
move,
you
change
one
of
the
port
parts
of
the
ship,
all
right,
at
which
point,
would
you
consider
this
ship?
No
longer
the
auto
ship,
very
philosophical
discussion,
but
it's
exactly
what
you?
What
you're
referring
to
right
not
comes
with
the
same
name,
is
not
the
same
note
by
the
way
yeah.
B
C
One
of
the
reason
I
was
asking
is
because
we
work
we're
starting
to
work
on
sort
of
maybe
leveraging
the
scheduled
event
metadata
to.
You
know:
work
we're
working
on
vmss
migration,
we're
running
vmss
and
now
we're
trying
to
see.
How
can
we
work
on
having
a
cleanup
process
and
I'm
wondering
if
kubernetes
will
do
that
cleanup
for
us
I
know
right
now
it
doesn't
seem
to
so
just
trying
to
see
over
time.
B
C
Right
now
we
were
just
looking
at
Asia
scheduled
events
right,
the
the
metadata
service
that
will
allow
you
to
detect.
Maybe
if
a
node
gets
removed
or
the
move,
the
nodes
is
getting
the
allocated
or
I
know
that
the
the
low
priority
VMs
will
be
sending
those
events,
but
I
haven't
been
able
to
see
if
and
I'm
all
scale
down
and
we'll
be
sending
those
events
right.
So
this
is
where
I
was
trying
to
see
what
the
best
approach
to
you.
B
Want
to
get
you
want
to
sponsor
this
project.
I'm
whole
game
actually
work
with
you
on
this.
The
idea
is,
you
want
to
drive
the
scheduled
events
by
scheduled
events,
just
for
people
who
don't
know
as
your
fabric
sends
events,
you
can
actually
get
them
from
Anniston
metadata,
like
any
your
notice
about
to
shut
down
hey
your
notice
about
we
upgraded
the
host
is
about
to
go
down
the
other
other.
What
you
can
do
is
you
can
drive
these
events
into
teens
and
labels
which
can
be
used
by
the
schedule.
From
that
perspective,
I
agree.
A
Interested
in
this
as
well
I
think
what,
if
I'm
understanding
Sylvain
what
you're
asking
is,
if
the
scaling
events
and
the
vmss
and
the
same
event,
so
you
could
just
build
a
single
daemon
set
that
lives
on
the
node
and
listens
for
events
that
are
either
maintenance
events
or
scale
down
events,
but
you're
asking.
But
it's
unclear
if
scheduled.
If
those
events
are
used
for
these
scale
operations
as
well,
no
I,
don't
Indy.
C
B
C
B
Think
might
have
this
scale
because
skill
event
is
not
a
fabric.
Even
it's
a
a
norm,
Evan
yeah.
Okay,
what
somebody
click
scale
either
called
an
API
via
library
or
just
call
that
many
by
the
portal
or
something
but
that's
an
armament,
and
that's
where
I
would
look
and
that's
exactly
what
I
was
trying
to
I
I
wanted
one
of
the
things
as
well.
Is
this
hidden
feature,
but
it's
there
we
can
actually
ask.
Let's
say
you.
B
B
A
D
B
B
Things
are
moving
like
cow,
like
forward
with
other
stuff.
The
autoscaler
is,
the
work
is
done.
We
have
it's
deployed
for
its
wrongs
with
VMs
skill
sets
and
ability
sets
and
I.
Guess
it's
already
there.
There
are
a
couple
of
bugs
Austrian
tracking
there,
but
it's
overall
in
a
good
state
supporting
2500
note:
I
actually
have
clusters
that
supports
4000
modes
and
plus
plus
on
Azure.
B
The
trick
is
they're
talking
location
for
Howard
city
would
work,
that's
really
good
majority.
The
majority
majority
of
the
work,
if
you
ever
noticed
that
110
comes
with
a
significance.
Caching
for
all
calls,
so
is
that
significant
caching
to
aren't
cool.
We
passed
that
shock
point
choke
point
of
arm.
So
now
we
can
support
a
very
large
number
of
nodes
in
a
same
cluster.
I
am
it's
like
my
10%
project
every
now
and
then
I
update
the
toxic
scripts
and
I,
then
I
will
merge
them
into
a
CSS.
B
B
B
D
C
B
It
was
multiple
to
the
stop
net.
This
has
told
all
right.
No
progress
really
made
over
there
I've
been
asking
like
everybody
was
talking
with
of
customers.
Do
we
want
to
support
like
cluster
spanning
multiple
subnets,
multiple
units
and
so
on?
And
the
answer
so
far
is
yeah.
There's
a
couple
of
customers
really
looking
into
this,
but
nothing
concrete.
Yet
so
as
things
change
and
as
our
recovery
requirements
becomes
more
clear,
you
may
need
to
revisit
this.
B
B
Pr
is
there
I
think
be
like
there's
a
web
B
or
anybody
wants
to
pick
it
up.
It's
a
cool
for
help.
Okay,
what
identity,
I
think
everybody
is
keeping
an
eye
on
this.
We
moved
it
to
beta
yesterday,
actually
should
have
been
moved
last
week,
but
then
people
forgot
that
I
did
the
docs.
But
now
it's
moved
to
beta
I
am
as
I
was
scrolling
through
the
box.
There
were
some
questions,
so
we
have
two
minutes
to
go
through
these
questions.
Anybody
has
any
questions
on
them.
A
A
B
Well,
we're
in
the
process
of
discussing
who
would
own
it
all
right,
because
it
means
that
means
somebody's
body
to
babysit
it
and
right
now.
It's
me
plus
plus
a
couple
of
guys
here
in
Microsoft
and
something
at
Microsoft
are
actually
helping,
but
we
want
this
to
become
a
thing
so
we're
in
the
process
of
discussing
who
alright,
let.
A
B
Just
questioned
about
Jojo
who's
asking:
can
it
be
used
with
keyboard?
Yes,
can
it
be
used
to
keep
secret
assets
of
the
VM?
There
is
another
projects
happening
around
bringing
key
vault,
strong
type
secrets
and
two
pots
alright
still
being
discussed,
should
have.
We
have
a
like
a
basic
concept
document
alright
and
I'll
share
it.
I'll
share
I'll
open
up
the
repo,
probably
early
next
week,
Kukoc
just
Comfort
document
of
what
we're
trying
to
do.
B
Okay,
our
D
haps
can
write
agonist
I
am
yes
same
as
iron,
then
haps
will
just
work
what
identity
proxy?
Yes,
that's!
Basically,
that's
what
we're
trying
to
get
we
are
in.
The
discussion
was
perfect
team,
the
dude,
the
dude
from
CIO
and
we're
trying
to
get
stiffy
as
face
to
for
this.
For
this
for
identities.
That's
it!
That's
super
duper,
update,
I'm,
sorry,
I
haven't
been
able
to
provide
it
all
the
loss
to
information
of
this
meeting,
but
hey.
B
A
B
Mostly
fixed
we're
still
tracking
random
I/o
timeout
q
penis
against
hazard,
minute
dns,
where
we
actually
captured
one
of
our
massive
clusters:
internal
massive
clusters,
very
busy
cluster.
We
capture,
24
hours,
captured
just
to
understand
exactly
what's
happening,
and
that's
this
is
really
the
last
issue
were
dealing
was
because,
to
be
honest,.