►
From YouTube: TGI Kubernetes 036: Virtual Kubelet
Description
Come hang out with Kris Nova as she does a bit of hands on hacking of Kubernetes and related topics. Some of this will be Kris talking about the things she knows. Some of this will be Kris exploring something new with the audience. Come join the fun, ask questions, comment, and participate in the live chat!
A
Hey,
what's
up
everybody
how's
it
going
welcome
to
TGI
K
I.
Am
your
host
Chris
Nova
on
this
lovely
Friday
afternoon?
And
let's
do
everybody?
Well,
it's
certainly
my
favorite
part
of
the
week,
but
let's
go
ahead
and
do
hellos
from
everyone
see
who's
all
here
and
make
sure
everything
is
working.
Good.
Last
time
I
was
plagued
with
with
demo
was
trying
to
get
my
stream
up
and
running
so
I
had
to
reformat
my
entire
hard
drive,
reinstall
Mac
OSX
get
OBS
up
and
running
and,
like
poof,
everything
seems
to
be
magically
working
now.
A
A
Jason
Andy
good
to
see
you
Philip,
Joe,
Caleb,
Arnaud,
Sebastian
I,
think
do
already
have
a
Sebastian
there's
two
Sebastian's.
Okay,
we
have
Caleb,
we
have
Norman,
oh
my
gosh.
This
is
like
the
part
where
it's
like
really
hard
to
pronounce
people's
names,
dawn,
shot,
shot,
shot,
Aaron
good
to
see
you
Neil
Mitch
Stevo
Josh,
James,
Eric,
st.
Martin,
one
of
my
homeboys
from
Microsoft
RIA,
my
homegirl
from
Microsoft
how's.
A
A
Ria
gave
a
great
talk
on
the
virtual
cubelet,
which
of
course
we're
talking
about
today
that
everybody's
so
excited
about
seeing
so
I
got
in
last
night
and
came
in
today
and
today
has
pretty
much
been
all
TGI,
K
prep
all
day.
So
getting
ready
to
do
this
thing
so
first
things
first
I
got
my
sweet
virtual
cubelet
shirt
on
courtesy
to
the
folks
at
Microsoft,
who
gave
me
a
shirt
at
the
cube
con
booth.
It's
a
it's
a
dude
shirt
but
actually
looks
pretty
good.
A
Cuz
I
got
it
in
a
small,
so
rockin
my
VK
shirt
today,
oh
I
also
discovered
that's
like
the
thing
now
like
everybody
was
calling
it
VK
like
it's,
not
virtual
cubelet.
It's
totally
VK
so
for
the
rest
of
the
episode,
I'll
be
trying
to
remember
the
acronym.
So,
let's
start
off
with
live,
updates
kind
of
what's
been
going
on
in
the
week
and
I
have
some
exciting
announcements
for
everyone.
So,
let's
switch
over
to
my
screen.
Poof.
A
Okay,
thanks
thanks
Mitch,
it
is
an
iced
tea,
I
like
rocking
it.
So
this
is
exciting.
So
what
we
have
here
is
a
hefty
Oh
TGI
K
repository,
so
my
co-author
Justin
garrison
and
and
Joe
we're
talking
on
Twitter
about
how
folks
wanted
to
maybe
start
contributing
suggestions
and
ideas
and
start
documenting
TGI
K.
So
I
went
ahead
over
the
last
week
or
so,
and
I
created
this
new
private
repository
github.com
slash
kept,
do
/t,
GI
k
and
we've
added
our
first
episode,
which
is
episode
36.
The
episode!
A
Why
we're
on
today
in
this
this
episode's
directory
and
we're
gonna,
start
tracking
tracking
episode
information
here.
So
without
further
ado,
let's
take
a
look
at
this
new
repository,
so
we
go
here
to
the
front
page
and
we
have
this
fabulous
picture
of
little
Tahoma
peek
from
the
crest
of
Mount
Rainier,
looking
south
out
onto
the
mountain
range,
and
we
have
this
really
handy
watch
live
now
button.
A
lot
of
folks
have
been
asking
for
this,
so
this
is
a
great
example
of
a
pointer
pointer.
A
So
every
week
we
update
the
TGI
K
watch
live
now,
link
to
the
pointer
to
the
this
week's
episode.
So
this
should
always
work
and
go
to
the
most
recent
or
the
upcoming
TGI
K
episode
of
the
week.
So
you
can
just
bookmark
this
page
and
just
come
here
and
click.
The
watch
live
Now
button
and
then
we
just
have
a
really
quick,
quick
and
dirty
readme
folks
feel
free
to
contribute.
A
It's
really
easy
to
contribute,
it's
a
patchy
to
license
and
it's
a
great
way
to
get
some
github
brownie
points
and
start
working
with
myself
and
other
folks
here
at
FTO
in
open
source.
So
if
there's
anything
I,
probably
misspelled
a
bunch
of
stuff.
So
if
there's
anything,
you
see
feel
free
to
open
up
a
PR.
So
for
our
first
episode
we
have
a
readme
file
that
has
all
of
our
links.
We're
gonna
go
through
that
I
want
to
talk
about
today,
and
it's
got
stuff
on
the
Virtual
cubelet.
A
It's
got
stuff
that
happened
over
the
week
and
it's
got
everything
that
we're
going
to
to
bring
up
on
the
episode.
Hi
James
welcome
from
San,
Francisco
and
I
guess
before
we
do
this,
how
do
I
we're
gonna
open
source?
This
live
like
we're.
Gonna
do
it
live
so
we're
gonna
go
to
settings
and
we're
going
to
scroll
down
here
and
we're
gonna
make
this
repository
public,
so
we
type
T,
gik,
so
I
understand
the
risks.
Actually
I'm,
not
gonna
type.
This
on
the
screen.
Let
me
get
that
out
of
there.
A
The
kubernetes
github
org
has
like
some
crazy
security.
There
we
go
okay,
I
head
into
my
my
code
too,
so
moving
this
back
over
here
and
I
got
shortcuts
bam.
Bam
looks
like
now,
hep
D,
ot
gik
is
now
a
public
repo,
and
we
already
have
our
first
star
so
who
got
our
first
star
looks
like
we
have
Nick
doy
got
our
first
star
of
the
github
repo.
So
from
now,
until
the
end
of
t
gik,
he
will
be
number
one
on
the
list
for
a
first
person
to
start
e
di
k,
congratulations
Nick!
Oh!
A
We
already
have
three.
While
this
is
going
up
fast
anyway,
check
it
out,
and
what
we
have
here
is
the
issue
tracker
good
to
see
you
mark
welcome
for
Nell
born.
Yes,
my
password
is
one
two
three
four
five
six
I
have
no
idea
how
you
all
guessed
it
hi
from
Germany
good
to
see
you
Thurston
so
anyway
in
the
github
issue.
Tracker.
If
you
have
an
idea,
this
could
be
anything
from
hey.
A
A
It's
super
easy
just
put
whatever
you
want
in
there
we'll
figure
the
rest
out
we'll
start
in
posing
like
crazy
rules
and
unnecessary
laws
over
time
as
as
they
come,
and
you
actually
need
that,
and
then
we
have
a
couple
of
labels.
I
put
together.
We
have
updates,
we
have
shout
outs
and
we
have
episodes,
and
so
as
we
filter
through
these
we'll
start
adding
labels
and
stuff
and
I
guess.
This
is
a
accidental
new
update.
But
for
those
of
you
who
don't
know,
github
has
expanded
their
user
interface.
A
So
now
we
have
labels
for
new
contributors
and
github.
It
was
a
pretty
good
job
at
having
a
pretty
release
cycle
that
comes
out
every
so
often
I'm
on
the
maintainer
z'
list.
So
I
get
like
the
sneak
preview,
so
I
sometimes
don't
know
if
my
github
is
gonna
look
different
than
others
hi
septic
from
cloudy
Seattle.
Although
I
think
it's
it's
kind
of
sunny
today,
it's
pretty
sunny
out
there
so
anyway,
TGI
cam
repo
and
you
can
come
in
and
you
can
see
air
links.
A
What's
going
here,
I'm
gonna
open
up
the
portal
I'll
log
in
to
my
hefty
account
when
I
first
did
it
I
came
into
the
user
interface
here
and
I
went
to
kubernetes
services
and
I
tried
to
come
in
and,
like
add
a
cluster.
This
way,
and
it
kind
of
gave
me
some
trouble.
A
I
remembered
some
like
secret
ninja
moves
from
my
days
of
working
at
on
the
a
dirty
man
on
the
AKS
project,
which
is
what
we're
looking
at
here
and
I,
still
kind
of
like
ran
into
a
couple
of
errors,
getting
kubernetes
up
and
running
a
quick
google-
and
I
found
this
fantastic
article
and
for
folks
at
home,
who
are
interested
in
setting
up
kubernetes
on
Azure.
This
is
this
is
the
nugget.
This
is
what
you
want
to
do
here.
So
this
thing
worked
great
right
out
of
the
box.
A
I
ran
these
commands,
which
enabled
the
AKS
preview
so
that
we
were
able
to
actually
access
that
those
bits
of
software
behind
the
scenes,
I
came
in,
and
I
created
a
resource
group
and
I
gave
it
a
name
in
this
case.
Our
resource
group
is
named.
Where
did
it
go
here
should
be
Nova
TGI
Kate's,
which
is
yeah
right
here,
TGI
K
and
then
I
gave
it
location
east
us,
which
just
kept
the
same,
because
I
remembered
that
one
was
a
preferable
region
for
kubernetes.
A
It
spit
out
a
little
bit
of
Jason
and
then
this
is
like
the
magic
command
here
that
actually
worked
first
time
hole
in
one
poof,
kubernetes
up
and
running,
which
was
a
zas,
create
the
name
of
my
resource
group.
Their
name
of
my
cluster,
which
I
just
named
my
name
Nova
in
OVA
and
I,
gave
node
count.
I
set
that
to
three
and
I
went
ahead
and
did
generate
SSH
key,
so
I
can
SSH
into
those
nodes
in
case
anything
happens.
I
didn't
have
to
do
this.
A
Install
CLI
since
I
already
had
queue
Bechdel
installed
on
my
macbook
and
then
I
just
did
a
ze
Kaos
get
credentials
and
it
appended
my
cube
config
to
the
bottom
of
my
existing
cube
config,
just
as
I
wanted
it
to
do
and
that
took
the
whole
process
took
less
than
five
minutes
and
I
was
like
there's
no
way.
This
thing
is
working
and
I
pulled
up.
A
My
terminal
and
I
ran
my
infamous
ke
get
node
command,
which,
for
those
of
you
who
don't
know
I
a
Lea
s--
k4q
Bechdel
in
you
know,
is
short
for
nodes
and
poof
IKS
kubernetes.
In
like
five
minutes
it
was.
It
was
really
slick,
so
big
shout
out
to
all
my
homies
on
the
AKS
team
like
good
job.
This
actually
worked
really
well.
So
let's
go
back
and
go
back
to
our
repo
and
see
what
we
have
next.
A
So
the
next
thing
I
wanted
to
bring
up
was
some
other
Kubek
double
tips
and
tricks.
This
first
one
Jorge
tweeted
this
out
earlier
in
the
week.
George
works
for
us
here
to
have
do
and
he
tweeted
out
that
we
had
some
great
Kubek
tool,
tips
and
tricks
and
I
was
like
oh
I.
Remember
there
was
a
cheat
sheet
once
upon
a
time,
but
what's
this
we
have
here
and
I
started
reading
through
it,
and
it
was
like
documentation
and
I
was
like.
Oh
there's,
some
good
stuff
here
and
about
halfway
through
I.
A
Go
oh
wait
a
minute.
This
is
the
new
discuss,
kubernetes
that
I.
Oh
that
was
announced
just
recently
and
I
didn't
even
realize.
I
was
using
it
and
it's
actually
like
this
really
cool
interface
where
you
can
come
in
and
you
can
join
and
you
canyou
can
contribute
and
it
kind
of
becomes
like
real-time
documentation,
but
it's
also
like
a
chat
venue
as
well.
So
that's
really
exciting,
so
we're
gonna,
just
I
created
my
first
account
and
I
haven't
actually
like
it
contributed
this
to
this.
A
Yet
so
we're
gonna
see
if
we
can't
come
in
and
add
a
reply
thanks
for
all
the
hard
work,
we
are
loving
these
tips
and
tricks
on
GGI
kay.
So,
let's
see,
if
I
can
reply
and
poof
there
you
go.
I
am
now
officially
a
contributor
to
the
disgust
at
kubernetes,
the
I/o
project
for
the
first
time.
So
that
was
pretty
easy.
So
anyway,
there
actually
has
some
really
good
tips
and
tricks
in
here.
A
The
first
one
was,
if
you
guys
don't
know,
this
C
is
a
new
thing
for
a
lot
of
folks
who
don't
run
multiple
containers
in
a
single
pod.
So
that's
handy
to
know,
then
we
have
a
couple
of
these
rock
bands.
Oh,
my
gosh,
what
are
people
saying
so
plus
one
for
the
TGI
K
repo
eric,
says
cubic
dual?
No
more
cute
cuddle,
yes,
Eric!
It's
Quebec
till
now,
and
kubernetes
and
Marco
says:
keep
cuddle
furred
forever.
I
almost
wore
my
cube
cuddle
shirt
today.
A
Just
so
I
could
say:
Quebec
Thailand,
where
I
keep
cuddle
shirt.
Well.
Uz
keeps
ETL
hi
Bryan
from
Denmark.
No,
no
worries
we're
just
getting
started
so
yeah.
We
just
have
a
couple
of
really
great
examples
here
and
some
folks
coming
in,
like
here's,
my
alias
kay
for
a
Kubek
double
shortcut
here.
So
clearly,
not
the
only
one
who
does
that,
but
but
yeah
I
feel
free
to
come
in
you
can.
A
A
Floor
and
says
cubicle
tips
and
tricks.
Oh
my
god!
How
long
have
we
come
or
hello?
Have
we
fallen?
That's
pretty
funny
so
another
one
here
is.
This
is
like
a
little
bit
more
of
a
formal
cheat
sheet
that
this
looking
at
the
other
tips
and
tricks
kind
of
reminded
me
of
this
sheet
from
way
back
when
that
I
haven't
looked
at
in
forever
and
it's
it's
changed
a
lot
since
last
time,
I
looked
at
it,
so
I
figured
it'd,
be
good
like
a
refresher
for
folks.
A
If
you
don't
know
this
is
here
it's
in
the
official
kubernetes
documentation.
You
can
come
check
it
out.
It
talks
about
all
kinds
of
handy
things
to
make
you
like
a
kubernetes
wiz
and
look
awesome
on
live
demos,
particularly
the
québec
to
autocomplete.
So
you
can
get
some
really
awesome
tab
hinting
while
you're
doing
live
demos.
The
cube
context,
I
know,
is
a
big
one.
We're
going
to
talk
about
this
piece
of
software
next,
that
kind
of
solves
the
same
problem,
but
here's
the
official
Quebec
tool
way
of
managing
your
different
cube
context.
A
It
talks
about
creating
objects,
gives
you
different
examples
of
creating
it
from
a
director
year
for
a
single
llamo
file,
and
it
talks
about
how
you
can
explain:
multiple
objects
at
the
same
time
or
even
interact
with
multiple
objects
at
the
same
time.
So
it's
really
good
just
kind
of
to
run
through
this
every
once
in
a
while
and
usually
I
pick
up
one
or
two
tricks
that
I
just
kind
of
try
out
on
a
whim
and
they
they
work
so
yeah.
A
This
is
a
really
great
resource
if
you
guys
want
to
come
check
it
out.
So
the
next
one
is
cube.
Ctx,
so
I
usually
have
this
installed
on
my
Mac,
but,
like
I,
said
I
just
reformat
in
my
hard
drive,
so
I
don't
have
it
installed
yet
so
we're
gonna
just
install
it
really
quick
live,
so
you
guys
can
see
how
easy
it
easy.
It
is
to
to
get
this
thing
up
and
running
so
installation.
You
do
brew,
install
cube,
CTX
and
then
I
already
have
my
alias
built
for
it.
A
Beautiful
and
let's
see
if
my
alias
works,
it
does
so
right
now,
I
have
my
kubernetes
cluster
I
created
in
Microsoft,
Azure
called
Nova
and
I
can
rename.
Let's
say
this:
first
one
is
a
one
that
was
created
with
key
Badman
if
I
wanted
to
rename
that
I
could
do
Kate
TX
the
new
name.
So
what
if
it
was
called
this
old
cluster
and
then
the
the
existing
name
of
the
cluster
itself
would
I
do
it
wrong?
It's
a
new
name
equals
old
name.
So
there's
a
delimiter
there
sweet.
A
So
now,
if
I
run,
KTX
and
I
can
KTX
to
old
cluster
and
poof,
my
cute
config
is
switched
and
I
can
go
back
to
my
Microsoft
cluster
and
kagan
nodes
and
bam.
We
can
just
switch
cute
config
really
easily
that
way,
and
it
gives
you
a
kind
of
handy
user
interface
for
renaming
things
and
keeping
track
of
them.
A
I
use
this
on
stage
when
I
did
my
acute
flow,
an
arc
demonstration
where
I
I
migrated
some
components
over
from
Amazon
over
to
Google
fits
with
Q
flow
and
arc
Cuban
s,
yeah
Jason
brought
up
Cuban
s
is
another
good
one.
We
use
that
a
lot
in
the
movie
monolith.
It
calls
or
movie
monolithic
application
calls,
though
it
was
almost
done
every
piece
of
documentation.
A
A
We
have
the
virtual
cubelet
documentation,
which
I
just
added
this
here.
We're
gonna
look
a
lot
at
this
in
a
second,
so
I'm,
gonna
kind
of
breeze
past
it,
but
it's
there
as
a
pointer
for
folks
another
one
we
wanted
a
shout
out
with
some
more
great
work
that
can
save
contributes
folks
are
doing.
Is
me
are
contributors,
so
this
is
the
first
Wednesday
of
every
month.
It's
like
behind
the
scenes
meet
kubernetes
contributors
and
maintainer
x'.
It's
basically
a
call
just
like
this
one.
You
can,
you
can
tune
in
on
YouTube
and
you
ii.
A
Ask
me
anything.
You
can
contribute
in
slack
and
ask
contributors
anything.
I
did
it
a
couple
months
ago
and
somebody
asked
me
like
what
my
favorite
mountain
was
and
what
my
first
commit
was
like,
and
it
was
pretty
exciting
getting
into
to
interface
with
folks
that
way.
So
this
is
ran
by
a
by
george
here
and
have
to
go
in
paris
over
at
google
and
it's
a
really
great
program.
A
So,
if
you
guys
don't
know
about
this
or
if
you
would
like
to
be
started
or
featured
on
this
feel
free
to
reach
out
to
the
folks
in
the
slack
channel
there
next
up,
I
found
this
blog
that
a
lot
of
people
were
talking
about
at
glucan
earlier
this
week,
which
is
under
the
hood
of
the
operator
sdk,
which
totally
has
my
attention.
Right
now
I've
been
wondering
about
sort
of
a
generic
operator
framework,
and
this
looks
like
it's
a
really
good
sort
of
overview.
A
Oh
Jason
says:
Cuban
s
is
included
with
cube
CTX
awesome,
but
this
looks
like
it's
a
really
good
overview
of.
What's
going
on
and
kind
of
giving
you
the
the
play-by-play
of
the
new
operator
framework
that
was
announced
at
cube
con
from
our
folks
over
at
Red
Hat.
So
there's
a
pretty
handy
diagram
here.
A
Joe
gave
a
really
great
keynote,
and
there
was
this
one
thing
that
he
said
that
really
kind
of
struck
home
with
me
and
I
have
never
heard
heard
this
phrase
before,
but
I
figured
it'd
be
important
to
share
with
folks
which
Joe
was
talking
about
the
difference
between
controllers
and
operators,
and
he
said
he
was
able
to
describe
it.
An
operator
in
in
one
sentence,
which
was
an
operator,
is
just
a
domain-specific
controller
and
that
to
me
it
was
just
like
the
ball
like
well.
A
Somebody
finally
did
it
in
in
one
sentence
in
Lackey
over
at
Microsoft,
was
at
the
conference.
He
had
the
same
reaction
as
well,
so
it's
it's
a
very
unique
piece
of
code
that
behaves
like
a
controller,
but
it's
very
specific
to
a
single
thing.
You
know
it's
specific
to
my
sequel
or
to
Mungo,
or
it's
doing
some
concrete
tasks,
that's
important
to
humans,
so
the
operator
SDK
is
going
to
take
that
and
actually
make
that
easier
for
you
to
interact
with
and
make
it
easier
for
you
to
build
out
one
of
these
operators.
A
So
maybe
we
can
get
on
one
of
these
in
the
in
the
future.
But
anyway,
if
you
want
to
come
through
and
read
this
and
learn
more
about
it
via
the
link,
is
there
next
up?
We
have
an
update
on
arc
our
favorite,
backup
and
recovery
tool,
so
we're
here
live
on
my
Twitter
and
it
says
we're
starting
to
open
meetings
to
design
new
features.
The
first
is
on
replication
of
backup
data,
so
the
folks
at
arc
have
a
mailing
list
that
you
can
join
here
in
the
the
Google
Groups
and
you
can
start
getting.
A
Calendar
invites
and
email
notifications
about
design
meetings
that
are
going
on,
so
you
can
come
and
you
can
join.
You
can
contribute
to
our
can
help
steer
the
shape
of
this
open-source
tool
moving
forward.
So
we've
already
done
a
TGI
Kay
on
arc
I'm
sure
we'll
do
it
another
one,
because
the
program
is
growing
so
much,
but
it's
a
really
great
opportunity
if
you're
looking
at
getting
involved
in
heavy
f,
do
open
source
and
nailing
some
of
those
sweet
keys
that
we
were
passing
out
at
the
conference.
A
The
kubernetes
and
hefty
o
mechanical
keyboard
feel
free
to
to
join
it
on
this
project
and
get
some
commits
in
and
see
what
the
folks
said
and
rker
up
to
another
good
discuss
bit
of
documentation.
We
had
is
this
one
that
Josh
at
Red
Hat
put
together,
that
is
last
week
in
kubernetes,
and
this
is
sort
of
a
really
great
piece
of
documentation.
That's
also
a
form
of
communication
as
well.
That
goes
in
and
talks
about
everything
that
happened
in
this
last
week
in
kubernetes,
which
is
what
we're
talking
about
right
now.
A
He
brings
up
the
community
meeting.
What
went
on
in
the
community
calls
which
happened
every
Thursday
at
10
a.m.
Pacific?
You
can
get
the
link
from
github
comm,
slash,
kubernetes,
slash
community.
He
talks
about
the
release
schedule
and
the
different
various
pieces
of
future
work.
So
this
is
a
really
good
resource.
I
haven't
really
seen
anything
this
comprehensive
written
down
yet
in
the
kubernetes
ecosystem.
A
So
the
fact
that
we're
getting
this
every
week
is
a
pretty
big
milestone
for
the
project
and
pretty
exciting
for
you
to
come
in
on
a
busy
schedule
at
a
glance
see
what's
going
on
in
kubernetes,
and
if
you
have
any
questions
or
notes,
you
can
come
down
here
to
the
bottom
and
click
reply
and
throw
in
your
two
cents
cool.
So
we're
almost
done.
A
We're
gonna
start
on
the
virtual
cubelets,
soon
we're
about
20
minutes
in
so
I'm
trying
to
trying
to
keep
a
good
pace
here
and
the
last
one
I
wanted
to
bring
up.
Was
this
I
easy,
install
helm,
which
I
meant
a
couple
of
minutes
researching
this?
Our
friend
Paul,
who
brought
this
up
on
one
of
the
movie
monolithic,
calls
had
brought
this
up
when
we
were
trying
to
get
helm
installed
on
our
back
cluster
and
he
wrote
these
really
convenient
helm.
Installing
helm,
delete
shortcut
aliases
that
you
can
copy
into
your
bash
profile.
A
So,
what's
funny
is
I
spent
a
couple
minutes.
I
looked
these
up,
I
found
them
I
added
them
to
my
bash
profile.
I
deployed
a
azure
kubernetes
cluster
and
I
did
a
home
in
it
and
we
weren't
using
our
back
so
I
didn't
need
them
anyway,
but
it's
a
really
great
article
for
those
of
you
who
want
to
get
helm
up
and
running
with
with
our
back.
Ok
and
it
looks
like
we
have
some
stuff
going
on
in
the
chat.
So
let
me
just
take
a
quick
look
here.
A
Andy
said
thanks
for
the
plug
you're
welcome
Andy.
Is
this
on
github
newsletter,
mind
sharing
the
link
later
so
yeah
Florian,
it's
in
the
the
link
to
the
the
discuss
thing.
I
just
pulled
up
as
in
the
tea
tik
repo
github
calm,
so
I
should
have
to
go
/tj
gates,
I!
Think
I
can
share
links.
Let's
see
if
I
can
do
this
good.
A
Comm
so
I
have
to
use
/t
g
IDK
dam,
see
if
that
works
there
you
go.
So
that's
the
link
there,
thanks
for
helping
out
Marco
and
without
further
ado,
I
think
it's
time
to
start
hacking
on
the
virtual
cupid
and
eric
had
us
to
question
earlier
which
was
like
now.
Are
you
actually
going
to
be
hacking
on
the
virtual
cubelet
or
are
you
just
gonna
be
hacky?
That's
a
virtual
cubelet
and
I
think
the
word
hacking
is
a
bit
of
a
marketing
ploy
here,
we're
not
actually
exploiting
anything.
A
The
virtual
cubelet
as
far
as
I
know
it's
fine,
so
we're
not
like
exploiting
anything
we're
just
gonna
be
playing
with
it
and
doing
a
live
demo.
So
that's
what
we
were
implying
when
you
said
hacking
on
the
virtual
couplet
this
week.
So
let's
pull
up
the
virtual
cubelet
github
repository
I,
always
like
to
start
these
things
that
I
haven't
start
them
already,
because
I
get
handy
notifications,
but
it
looks
like
I
already
got
to
this
one
and
then
I
guess
real,
quick
Before.
A
I
think
I
was
in
the
boulder
office,
I,
don't
know
if
we
were
still
dais
or
we
had
come
over
to
Microsoft
yet
but
I'm
pretty
sure.
I
was
sitting
in
the
boulder
office
and
we
got
to
watch
Brendan.
Do
this
demo
and
like
he
did
it
kay
get
nodes
and
there
wasn't
it.
There
was
a
node
that
was
there
that
wasn't
a
real
node
and
we
were
all
like
it
didn't
make
sense,
and
then
like
have
this
big
aha
moment
where
was
like?
A
Oh
we're
we're
spoofing
the
cubelet,
it's
a
node,
but
it's
not
a
node,
but
kubernetes
thinks
it's
a
node,
but
it's
just
doing
everything
the
cubelet
does
and
it's
not
a
node
so
that
project
I.
You
know
what
it
vendored
it
everything
off
to
a
CI
and
so
that
project
I
think
later
turned
into
what
we
see
here,
which
is
the
virtual
cubelet.
So
there
used
to
be
as
let's
see
if
I
can
find
it,
there
was
the
what
was
it
called
at.
He
was
at
the
ACI.
A
Okay,
it
looks
like
it's
no
longer
on
github,
so
I,
don't
know
what
happened
to
that
repository,
but
there
used
to
be
a
repo
of
something
that
did
a
very
similar
thing
that
was
written
in
typescript.
That
then
was
later
ported
over
to
go,
which
is
this.
This
program
we
see
in
front
of
us
I
think
it
was
originally
presented
to
the
advocacy
team
which
I
was
on
at
the
time
and
I.
A
Think
we
remember,
we
were
gonna
call
it
the
fake
lit
for
a
while
and
then
virtual
cubelet
kind
of
came
and
took
over
okay,
so
Rio
says
it's
a
private
repo.
Now,
okay,
it
was
called
a
CI
connector,
but
yeah.
So
anyway,
it's
pretty
interesting,
like
I,
always
loved
a
good
piece
of
software
when
it
kind
of
goes
on
a
journey,
and
it
turns
into
this
new
thing,
because
you
know
that
it's
like
it's
been
tested
and
it's
kind
of
like
they
made
it
through
the
evolutionary
cycles.
A
So
to
see
something
that's
complete,
as
this
is,
is
always
really
exciting.
So
we
had
a
quick
question
which
is
virtual,
cubelet
cloud
agnostic.
The
short
answer
is
yes
and
we're
gonna
get
more
into
that
when
we
jump
into
the
source
code
here
in
a
second,
so
anyway,
virtual
cubelet
slash
virtual
cubelet.
It's
a
go
repo.
A
You
can
come
in
here
and
you
can
sort
of
see
that
the
the
first
bit
of
documentation
kind
of
just
echoes
what
I
said
earlier,
which
it's
a
kubernetes
cubelet
implementation
that
masquerades
that's
such
a
great
word
to
use
here
masquerades
as
a
cubelet
for
the
purpose
of
connecting
kubernetes
to
other
API
s.
So
anyway,
to
answer
your
question.
It
looks
like
right
now
we
have
a
CI
which
is
a
Microsoft
Azure
product.
We
have
AWS
Fargate,
which
is
of
course
a
tub
us.
We
have
hyper
SH
and
we
have
this
new.
A
Let's
pull
it
up
here:
the
azure
IOT
Edge
virtual
cubelet
provider,
which
is
a
virtual
queue
provider
for
azure
IOT,
which
I
think
Eric
n
mentioned
earlier.
This
is
a
pretty
new
thing,
so
it
is
cloud
agnostic
and
I.
Think
it's
pretty
easy
to
write.
One
of
these
implementations
and
we're
gonna
go
and
we're
gonna
kind
of
look
at
the
interface
and
we're
not
gonna.
A
Write
an
implementation
today,
but
we're
gonna
look
and
see
what
it
takes
so
to
explore
the
program
we
have,
our
good
ol
main
go
and
we
have
all
of
our
top-level
repository
things.
We've
got
our
go
package,
it
looks
like
they're
using
depth
good
job,
and
then
we
have
a
couple
of
packages
here
and
we
have
our
good
old,
CMD
directory,
which
I'm
sure
has
some
entry
points
for
the
program
as
well.
But
what
I
wanted
to
look
at
first.
Was
this
diagram
here
that
sort
of
explains
how
the
virtual
qiblah
works.
A
Running
in
the
wild,
we
have
a
kubernetes
control
plane
and
we
have
three
worker
nodes,
so
that
should
mimic
this
diagram.
If
we
had,
you
know
one
more
node
here
which
I
I
think
the
number
of
nodes
is
irrelevant,
and
then
we
have
this
like
new
virtual
qulet
that
we're
going
to
run
on
the
side.
I
think
the
way
this
is
going
to
work
in
this
demo
is
we're.
A
Gonna
run
the
virtual
cubelet
on
my
local
macbook
and
that's
actually
going
to
take
my
macbook
and
use
that,
as
like
intermediary
computer,
that's
gonna
revenger
off
to
a
CI
moving
forward
and
then
with
the
virtual
cubelet,
we
can
perform
a
couple
of
different
actions
like
we
can
you
do
node
conditions
we
can
get
pods,
we
can
get
status
of
a
pod.
We
can
update
it.
We
can
even
get
the
capacity
in
the
operating
system,
so
I
was
able
to
come
in
and
I'll
bring
you
guys
back
to
where
I
am
mentally.
A
With
this
whole
project,
cuz
I
I
started
to
jump
into
it
yesterday
and
then
was
like.
You
know
what
no
I'm
just
gonna
do.
This
whole
thing
live
on.
T
gik,
so
if
I
fumble
around
a
bit,
we
have
our
folks
at
Microsoft
here
to
help
me
and
we're
gonna
try
to
get
some
some
containers
deployed
into
a
CI
using
the
virtual
cubelet
with
Microsoft's
managed
kubernetes
service
today.
Well,
that's
a
lot.
Hopefully
we
can
get
that
all
in.
So
here's
where
we
are
so
we
have
our
kubernetes
cluster
up
and
running.
A
We
are
here
in
the
virtual
couplet
directory,
and
the
first
thing
I
did
when
I
came
in
here
was
I
took
a
look
at
the
Dominque
file,
so
we
can
come
in
and
we
can
notice.
We've
got
a
pretty
pretty
lengthy
make
file,
but
I
searched
for
the
string
go,
build
I
found.
We
have
this.
This
target
here
called
go
build.
It
calls
this
other
target
author.
A
A
Eric
says
he's
got
a
couple
of
additional
diagrams.
Okay,
maybe
you
can
open
up
a
PR
to
the
the
reap,
see
how
handy
it
is.
We
can
say
that
now
open
up
a
PR
to
the
repo
or
drop
an
issue
in
and
we
can
merge
those
back
into
the
the
readme
there.
Egg
anyway,
I
had
to
make
a
few
changes
to
get
this
thing
to
compile
earlier
today.
So
my
changes
are
here
in
my
local
directories.
So
I
think
what
I
want
to
do
is
commit
these
into
a
new
branch.
A
Come
back
to
the
master
branch.
Reset
everything
demonstrate
the
air
and
then
I
can
show
you
guys.
The
couple
of
one-liners
I
had
to
tweak
to
get
it
running
and
I
think
it's
just
an
issue
because
I'm
running
go
version
10.2,
but
anyway
we
have
some
go
experts
on
the
on
the
call
today.
So
maybe
they
can
give
us
a
little
more
insight
as
well.
A
So
the
first
thing
we
want
to
do
is
you
want
to
check
out
a
new
branch,
so
we're
just
going
to
call
it
t
gik
demo
I'll
call
it
go
ten,
because
I
have
a
feeling
that's
going
to
be
our
imposter
here.
That's
giving
us
so
much
problems
and
we
can
see
what
we
have
I've
got
a
couple
of
modifications
in
the
CR
I
go
file
and
I'm
gonna
go
ahead
and
add
that
to
our
index,
CRI
CRI,
dot
go
and
I'm
gonna.
A
Do
a
quick
commit,
adding
changes
to
get
program
to
compile
for
t
GI
K
when
KaBlam
and
now
we
can
get
check
out
master,
we
can
get
reset
hard.
Actually,
let's
look
at
it,
whatever
emotes
our
git,
remote,
V,
okay,
so
I'm
I
think
I
can
reset
from
that,
but
yeah.
That's
what
we
want
to
do
so.
Git
reset
hard
origin
master
beautiful.
So
if
I
wrote
a
make,
build
actually
before
do
that
go
version
now,
let's
do
a
make
build.
A
So
if
I
run
a
make,
build
it's
gonna
echo
out
building,
and
if
I
did
this
right,
we
should
get
a
couple
of
interesting
go
errors
here,
oh
and
it
looks
like
I-
was
buffering
just
a
little
bit.
If
folks
can?
Let
me
know
if
this
compile
is
thrown
off
my
CPU
and
if
I
just
started
buffering
I
can
not
try
to
compile
real
time
on
the
call.
A
Okay,
so
anyway,
it
looks
like
we.
We
hit
our
first
error
here,
which
is
we
have
some
undefined
sis
calls
which
last
I
checked.
This
is
called
sis
call
package
was
frozen,
so
if
this
is
an
with
with
syscall
package
changing
and
that's
kind
of
exciting,
because
that
means
we
finally
unfroze.
This
is
call
package,
and
if
we
go
back
to
our
TGI
K
demo
branch
get
checkout
BAM.
A
We
can
see
what
I
did
here
to
change.
This
I
haven't
seen
anything
in
the
chat
in
a
few
seconds.
So,
if
folks
want
to,
let
me
know
that
they're
still
seeing
me
that
would
be
helpful.
I
always
get
nervous
with
my
cpu,
taking
up
too
much
or
with
my
go
my
program
compiling
and
taking
up
too
much
CPU
anyway.
A
A
Video
looks
good
thanks
Jason,
you
always
got
my
back,
so
I
remember
telling
Joe
how
to
do
this.
We
can
zoom
in
here
a
little
bit
more
and
I.
Think
it
was
tool
windows
into
presentation
mode.
This
is
what
we
want
to
do
and
then
we
come
up
here.
We
close
that
we
zoom
in
there
and
then
I
can
go
view
tool,
windows,
project,
BAM,
I,
think
that's
what
we
want,
so
that
should
look
good
for
folks
at
home,
so
I
came
in
here
and
we
had
this
error
and
Craig
confirms
the
video.
A
It
looks
good,
so
we
had
this
error
down
here
and
you
can
see
what
I
did
I
really
have
to
this.
Basically,
all
we
were
doing
is
we
were
getting
some
information
on
this
call
package
and
remember
in
returning
a
UN
64
I,
just
hard
coded
1
0
to
4
and
returned
to
some
value,
we're
not
using
the
CRI
provider
in
this
call.
So
I
figured
that'd,
be
ok
just
to
get
things
compiling
and
to
give
it
a
non-zero
value
there,
and
then
the
other
thing
I
had
to
do
was
up
here
on
line
668.
A
We
had
pods
in
namespace
note
that
wasn't
it
was
one
of
these.
There
was
a
string
somewhere
in
here.
I
can
look
it
up
and
see
it
if
we
wanted.
If
we,
if
you
guys,
want
a
line
number
but
where
we
were
doing
a
%
s
and
referencing
a
pointer
to
a
struct,
so
we
go
compiler
yelling
at
us,
so
I
changed
that
to
a
percent
V
and
the
go
compiler
was
able
to
correctly
expand
that
into
the
string
and
the
program
compiled.
A
So
those
are
the
two
tweaks
I
had
to
make
in
a
CRI
provider.
I
can
push
that
branch
I
just
created
up,
so
that
folks
can
only
see
the
Delta
and
get
the
information
there.
But
I
was
able
to
make
those
two
changes
and
get
the
program
to
compile
so
we're
gonna
recompile
it
now
and
hopefully
my
CPU
doesn't
take
over
by
running
on,
make
build
and
that
should
compile,
and
we
can
actually
start
trying
to
run
the
program
and
and
see
what's
going
on
and
in
the
meantime,
I'm
gonna
open
up
that
PR.
A
So
we
can
see
the
Delta
here.
So
a
virtual
culet
virtual
cubelet
get
oh
I'm,
not
gonna.
Do
excited,
don't
have
my
branches
set
up
so
I'll
do
that
after
the
call
I
don't
have
my
remote
set
up
to
to
open
up
a
PR,
but
I
will
do
area
I
promise
just
trying
to
find
filler.
While
we
were
waiting
for
our
program
to
compile
anyway
we're
compiled
now,
and
if
we
look
here
in
the
my
IDE,
we
can
see
that
we
have
this
new
directory
called
called
been
in
in
the
bin
directory.
A
We
have
the
virtual
cube
late
binary,
so
we're
gonna
move
this
virtual
cubelet
binary
from
bin
virtual
cubelet
to
use
your
local
bin
virtual
cubelet
and
then,
like
I,
mentioned
earlier.
Everybody
calls
this
thing
VK,
so
I
want
to
be
the
first
part
I'm
sure
somebody
else
has
done
this,
but
I'm
gonna
claim
to
be
the
first
person
to
alias
VK
equal
to
I,
the
virtual
cubelet
I
guess
before
I.
Do
that,
let's
make
sure
VK
doesn't
do
anything
already,
okay,
yeah,
it
doesn't
I
want
to
break
anything
on
my
system.
A
Okay,
so
alias
GK
is
equal
to
virtual
cubelet
and
bam.
We're
running
the
virtual
cubelet,
it's
in
our
past
freshly
compiled
from
from
master
with
a
couple
of
small
changes,
and
we've
got
the
virtual
cubelet
program
running
and
let's
see
what
is
next
and
our
little
piece
of
documentation
here.
So
it
says
we
need
to
pass
in
a
virtual
cubelet
provider
and
then
equal
that
to
your
provider
and
I
guess.
A
This
is
probably
a
good
opportunity
for
me
to
talk
a
little
bit
about
various
providers
in
NGO
programs,
and
we
see
a
lot
of
this
in
kubernetes
and
especially
in
the
cluster
API
work
that
we're
doing
now,
and
it's
a
really
interesting
engineering
problem
to
solve
right
now.
If
you
look
at
how
the
cloud
providers
work
in
the
cluster
API
and
most
of
the
cloud
provider
implementations
that
you
see
in
and
around
the
kubernetes
ecosystem
cubic
Wan
cops,
these
are
all
no
different.
They
all
have
some
concept
of
a
cloud
provider.
A
Some
people
call
that
an
actuator
some
people
call
that
an
actual
cloud.
In
this
case,
we
call
it
a
provider
and,
if
I
think,
if
we
go
in
here,
we
can
find
probably
not
here
but
there's
going
to
be
an
interface
somewhere
in
the
source
code
that
describes
what
each
of
these
providers
need
to
satisfy
in
order
to
be
a
provider.
A
Now
the
problem
comes
in
is
when
there's
a
bug
in
the
CRI
provider
like
the
one
we
just
saw
and
we're
trying
to
use
the
azure
provider
and
we're
getting
the
conflicting
problems
and
the
the
azure
provider
isn't
compiling
because
there's
an
issue
in
somebody
else's
provider.
So
that's
one
of
many
great
examples
of
why
you
want
to
start
pulling
the
stuff
apart.
A
Instead
of
actually
trying
to
do
that
at
runtime
using
a
flag
or
a
switch,
so
there's
different
patterns
and
different
things,
we
can
explore
there's
a
couple
of
great
threads
on
the
topic
in
the
kubernetes
issue:
tracker
where
people
are
suggesting,
maybe
we
use
go
plugins,
maybe
we
can
use
grcc.
Maybe
we
have
some
sort
of
a
like
AP
like
rest.
Api
type
contract
like
we've
used
for
the
C
and
I
am
interface
and
implementation.
There's
a
couple
of
different
ways
to
skin
this
cat,
but
I
think
it's
you
know,
as
these
programs
are
evolving.
A
This
is
going
to
be
something
that's
really
important
for
the
kubernetes
community
to
look
at
and
say
how
do
we
start
having
a
cloud
specific
code
that
isn't
dependent
on
other
clubs
specific
code
to
to
work?
Okay?
So
that's
my
rant
on
cloud
providers.
So
anyway,
if
we
go
back
to
the
top
level,
we
can
top-level
directory
in
our
repository.
A
We
can
see
this
interface
here
that
we
were
talking
about
so
each
one
of
those
cloud
providers
is
going
to
be
implementing
one
of
these
provider
interfaces
and
each
of
these
interfaces
need
to
satisfy
each
of
these
methods
described
here.
So
we
need
to
be
able
to
create
pods.
We
need
to
be
able
to
update
pods
delete
so
on
and
so
forth.
A
A
It
goes
both
ways
and
so
you'll
need
connectivity
from
the
api
server
to
the
cubelet
and
back
from
the
cubelet
to
the
api
server,
and
actually
I'm
gonna
do
my
first
dot
cam
on
this
really
quick.
So,
let's
see
how
do
I
come
over
yeah,
okay,
so
I'm
going
to
use
Joe's
camera
here?
Okay,
so
you'll
actually
need
the
API
server
in
kubernetes,
which
is
going
to
have
TLS
in
front
of
it
and
you're
gonna
have
some
subset
of
cubelets,
and
this
is
a
little
known
fact
about
the
cubelet
here.
A
Each
cubelet
actually
has
its
own
TLS
encrypted
API
server
and
usually
they
have
their
own
group
of
certificates.
So
if
you
go
and
you
look
in
the
cube
admin
code,
there's
actually
different
phases
for
generating
these
certificates,
as
well
as
the
different
certificates,
each
of
the
cubelets
gonna
gonna
run
as
well.
So
the
API
server
is
going
to
need
to
be
able
to
speak
down
to
each
of
these
cubelets
to
call
the
methods
that
we
see
in
the
virtual
cube
lit
interface.
A
We
just
looked
at
in
github,
but
furthermore,
we're
actually
gonna
need
to
go
back
the
other
way
as
well,
because
the
API
server
is
actually
going
to
be
listening
to
the
cubelets
TLS
service
as
well,
so
that
we
can
do
things
like
exec
and
log.
So
it's
a
sort
of
like
unidirectional
transaction
that
you
can
prescribe
what
you
want
the
cubelet
to
do,
but
you
can
actually
request
information
back
from
the
cubelet
as
well.
A
A
really
great
example
would
be
if
we
took
our
API
server
and
we
said,
create
pod
and
it
came
over
to
the
API
server
that
the
cubelet
had
and
or
I'm.
Sorry,
not
the
API
server
to
the
qiblah
had
the
it
said.
You
know
the
cubelet
was
watching
this
API
server
and
said
we
need
to
create
a
pod.
The
scheduler
told
it
to
create
the
pod
pod
is
now
running
on
this,
this
host
here
and
then
somebody
said
you
know
what
I
want
to
grab
logs
from
that
pod.
So
what
happens
then?
A
A
So,
let's
scroll
back
up
and
let's
see
our
first
example
here,
it
says-
run
the
binary
with
the
chosen
provider,
so
I'm
gonna
guess
we
can
come
in
here
and
go
into
our
providers
directory
and
I
bet.
We
can
just
do
provider
azure
and
see
what
happens
so
we
can
do
VK
provider
and
I'm
just
gonna
guess
this
is
the
syntax
we'll
see
what
happens
here?
Oh
we
got
a
null
pointer
dereference
on
AC.
I
go
online
147,
let's
see
I
bet.
I
can
fix
this
really
quick.
A
Subscription
ID,
okay:
this
is
great
okay,
so
I
bet
there's
some
sort
of
looks
like
we
need
to
do
some
sort
of
authentication
setting
to
get
a
authenticated
with
the
azure
web
portal.
I
was
wondering
you're
gonna
hit
this,
so
it
looks
like
here
we're
doing
OS
good
environment
and
are
expecting
these
environmental
variables
to
be
set.
So
I
have
a
hunch
if
we
go
ahead
and
export
these
environmental
variables
with
working
credentials,
we'll
be
able
to
fix
our
null
pointer
dereference
and
get
the
virtual
cube
lit
up
and
running.
A
A
Let's
see,
there's
examples:
there's
charts,
there's
V
cubelet.
Is
there
like
an
asscher-cut
krupp
for
agar
container
instances
provider?
So
here
we
go
so
we
have
a
config
file.
Okay,
the
azure
container
instance.
This
provider
allows
you
to
use
utilize,
both
typical
posit,
VMs
and
ëcÃ.
You
can
find
detailed
instructions
on
how
to
set
it
up
using
the
ACI
instance
provider,
documentation.
A
I
think
this
is
what
we
want
to
go
through
here:
okay,
so
that
this
is
what
we
looked
at
earlier.
You
can
install
the
CLI
okay,
we
have
kubernetes
helm
installed,
I
mentioned
earlier.
How
I
had
set
that
up?
So
we
should
be
able
to
do.
Let's
just
make
sure
it's
running,
helm,
LS
and
yep.
Okay,
so
helm
is
up
and
running.
We
got
that
good
ol
curl
pipe
to
bash,
for
what
is
this,
our
friends
at
windows?
A
Maybe
no!
This
is
Lennox
curl
pipe
to
bash
okay,
Lennox
here
so
for
Windows.
You
can
follow
that
to
get
the
helm
dot
exe
and
if
not,
you
can
curl
pipe
to
bash
here.
But
this
is
what
I
think
we
want
to
do.
So
we
want
to
do
a
Z
account
list,
Oh
table
yep,
and
we
want
to
export
here's
our
environmental
variables.
We
need
to
export
I
haven't
seen
anybody
in
the
the
chat
in
a
minute,
so
my
anxiety
is
getting
the
best
of
me.
A
So
if
somebody
wants
to
drop
a
quick,
hey,
Chris,
you're
doing
great
and
the
video
is
working
well,
that
would
be
exciting.
So
yeah,
let's
get
our
subscription
ID
and
let's
export
that
first
and
I'm
gonna
try
to
do
this
like
a
clean
way
in
real
time,
so
that
you
guys
don't
see
my
full
batch
profile.
But
we
should
be
able
to.
A
Let's,
how
do
I
want
to
do
this?
I'll
just
export
it
in
the
shell,
and
let's
make
this
our
magic?
Shell,
that's
what
we
can
do?
Okay,
so
we
want
add
your
as
your
subscription
ID
good
thanks
everyone.
So
it's
just
weird
cuz,
like
you're
sitting
here
and
you're
talking
to
yourself
and
you're
like
man.
A
It
could
be
like
that
one
time,
a
couple
weeks
ago,
when
I'm
in
here
for
ten
minutes
talking
to
myself
and
then
I
get
a
phone
call
that
says
nobody
can
see
you
so
yeah
thanks
for
thanks
for
the
support
everybody.
It
means
a
lot
to
me
so
anyway,
as
your
subscription
ID.
So
we
want
to
export
that
and
we
can
grab
our
Azure
subscription
ID
from
here
in
the
portal.
Oh
sure,
calm.
A
So
I
went
through
so,
if
you
guys
don't
know
this
I,
this
is
like
from
you
know
a
couple
months
at
Microsoft
working
with
me
as
your
team.
You
can
actually
customize
these
favorites
here
on
the
left,
so
I
went
through
in
like
sort
of
created
mine
I'm,
get
rid
of
these
ugly
IKS
failures.
Here.
I
went
through
and
created
mine
that
is
sort
of
unique
to
today,
and
you
can
actually
have
like
a
custom
little
like
markdown
snippet
here
for
like
shortcuts
and
commonly
used
things
masher.
A
So
anyway,
you
can
customize
Azure
and
change
the
colors.
It's
pretty
cool,
but
I
come
in
here
and
you
can
pull
up
we're
subscriptions.
They
should
be
in
here
somewhere.
Maybe
not!
You
can
go
to
all
services,
it's
right
here
in
the
middle
subscriptions
yep
and
we
should
have
just
the
one
and
you
can
click
on
it
and
you
can
grab
your
subscription
ID
here
and
there's
even
a
little
copy
thing.
So
we
can
pull
this
up.
Oh.
A
A
A
Oh
urea
says
it
was
on
the
same
page:
I
was
just
done.
Okay,
I'll
go
back
and
check
in
just
a
second.
Let
me
finish
this
thought
on
this
piece
of
documentation.
Here,
the
addressee
Li
can
be
used
to
install
the
a
CI
provider,
we'd
like
to
say,
address
provider.
Implementation
for
virtual
Cuba
is
the
a
CI
connector.
For
this
section
of
VK
specific,
a
CI
provider
will
be
referenced
to
use
the
a
CI
connector.
A
If
you
continue
with
this
section,
you
can
skip
between
the
sections
schedule,
a
pod
and
a
CI,
so
I
think
what
we
want
to
do
is
we
want
to
run
the
Linux
a
CI
connector
here.
This
is
the
command
we
want.
So
let's
go
ahead
and
pull
up
this
terminal
and
let's
see
what
we
got
here:
okay,
so
install
ECI,
connector!
A
Oh,
she
says
I'm
doing
the
right
thing,
I'm
on
the
right
one,
thanks,
Riya,
okay,
so
resource
group,
we're
gonna,
put
this
in
the
same
resource
group
we
put
the
rest
of
our
kubernetes
cluster
in.
It
seems
like
a
good
idea
to
me.
So
the
name
the
AKS
cluster
name,
which
is
the
best
name
in
the
world,
Nova
we're
running
on
Linux
and
the
connector
name.
We're
gonna
call
this
recall:
Nova,
connector,
Nova,
I,
dunno
vac
on
cuz
I
just
sounds
like
a
cool
thing
plus.
A
Obviously
we
did
Chris
Nova
tech
conference
in
our
lives
and
the
ACI
resource
group,
auto-generated,
aka,
KS
cluster
resource
group,
now
I
wonder
if
this
can
be
the
same
resource
group
as
the
one
defined
above
or
if
I
need
a
second
resource
group
here,
we're
gonna
try
first
with
the
same
one
because
I
just
like
keeping
everything
in
the
same
resource
group,
but
if
this
doesn't
work,
hopefully
by
the
time
I
get
around
to
it.
Somebody
could've
have
suggested
a
better
alternative
approach.
A
So,
let's
run
this
command
resource
group
Nova
could
not
be
found
interesting,
I
thought
I
had
let's
pull
up
or
Azure
portal
here
over
here
on
the
Left
we
have
resource
groups,
Nova
t
GI
k.
That's
right!
That's
right!
That's
right,
nobodies,
the
name
of
the
cluster,
so
let's
do
Noah,
T
G
I
can
and
then,
if
you
guys
don't
know
this
trick,
you
can
do
Oh
might
not
work
as
I
have
a
weird
keyboard
right
here.
A
Yeah
it
doesn't
work,
you
can
do
all
its
left
and
right
in
a
proper
bash
shell
and
you
could
skip
a
whole
word,
but
because
I'm
using
a
larger
keyboard
right
now,
my
oath
isn't
working
anyway,
Nova
TGI,
K
and
I
think
I
misspelled
that
Nova
TGI
K
so
go
to
the
end
of
our
line
and
run
that
and
see
what
happens.
Deploying
the
ACI
connector
for
Lennox
using
helm
air
found
failed
to
download
hint
running
hell.
A
Repo
update
me
help
Helen
repo
update,
wham
KaBlam
I
just
installed
helm
today,
so
that's
usually
should
have
been
done,
but
since
this
is
effectively
a
brand
new
computer,
we're
still
kind
of
doing
some
of
the
setup
you
stuff
here.
So
let's
run
this
again
deploying
the
ACI
connector
still
failed
to
download
hint
helm
repo
update.
This
looks
the
smells
like
a
helm,
repository
issue
here.
A
A
Schedule
a
pod
in
the
ECI
section
install
the
Linux
at
UCI
connector.
Ok,
so
it
says
you
need
to
specify
the
a
CI
resource
group
to
a
due
to
a
bug
in
the
AZ
CLI.
The
resource
group
is
the
auto-generated
to
find
that
any
Mackay
to
the
ad
reported
resource
groups
scroll
down
and
find
the
name
that
matches
the
MCA
KS
cluster
name:
aks,
RG,
location.
Okay,
that's
a
bit
confusing,
not
really
sure
what
string
we're
looking
for
there.
But
let's
see
if
we
can't
find
any
documentation
here
on
helm.
A
Deployment
into
your
cluster,
okay,
I
think
we're
getting
off
in
the
weeds
here.
Let's
go
back
up
to
the
higher-level
documentation
and
see
where
we're
going
so
I'm
wondering
if
we
were
actually
trying
to
in
that
scenario
trying
to
deploy
the
a
CI
connector
into
the
actual
kubernetes
cluster
itself,
which
I'm
trying
to
run
it
locally.
A
So
let's
go
back
and
let's
get
back
on
that
track
and
then
hopefully,
in
the
meantime,
if
somebody
can
drop
a
link
or
two
in
the
the
chat
here,
that
might
might
help
us
or
not
a
link
but
a
pointer
in
the
right
direction.
So
anyway,
so
we
came
back
and
we
tried
to
run
our
VK
provider
Azure
and
we
got
this.
So
let's
go
back
and
start
exporting
these
environmental
variables.
A
Alright,
so
we
pulled
up
our
source
code
and
we
did
Azure
subscription
ID
already
so
now,
let's
do
as
you're
a
tenant
ID,
which
I've
already
got
just
from
the
months
of
working
at
Microsoft
when
I
created
my
service
principle,
I
also
catted
it
to
a
local
txt
file,
so
I'm
able
to
get
all
the
stuff
pretty
easily.
So
this
should
go
pretty
quick,
which
is
a
good
trick
for
folks
at
home
to
create,
as
your
service
principles,
always
cat
that
to
a
text
file
on
your
local
file
system.
So
you
can
use
it
later.
A
So
anyway,
as
your
tenant
ID,
we
can
save
that
command
and
we
can
cat.
What
is
it
service
principle?
Jason
I'm,
going
to
show
you
guys
my
secrets,
but
I'll
nuke
this
after
the
call
anyway.
So
it's
not
going
to
be
a
huge
deal,
so
we
can
do
as
your
tenant
ID
when
KaBlam
and
that's
going
to
be
this
string
here
and
then
we
also
want
to
get.
A
Oh,
it's
this
one,
here's
our
app
ID,
that's
the
one.
So
this
always
kind
of
was
like
frustrating
for
me,
which
was
the
service
principal
calls
all
these
directives,
one
thing
on
the
command
line,
but
every
piece
of
data
documentation
always
calls
it
a
different
thing.
So
in
this
case
our
tenant
ID
is
the
tenant.
The
client
secret
is
actually
the
password
and
the
client
ID
is
actually
the
app
ID
anyway.
That's
the
the
mapping
that
you
need
to
remember
and
I
have
a
hunch.
This
should
work
now.
A
If
we
just
check
everything
is
there
we
should
be
able
to
export
case-insensitive
grep
for
Azure
and
yeah.
We
have
all
kinds
of
has
your
secret
e
stuff
running
there.
So
now,
let's
get
our
VK
command
back
here
in
the
stack,
so
we
can
pull
it
up.
Vk
provider
Azure
and
we're
still
getting
some
no
pointer,
dereferences.
A
Okay,
so,
let's
before
we
look
at
debugging
this,
let's
see
what
is
going
on
in
the
chat:
okay,
so
recess,
no
you're
good!
That's
the
one!
You
found
it
different,
one,
okay,
so
so
this
is
kind
of
hard
to
grok
here,
because
the
timing
is
so
off
so
big
long,
complete
sentences
are
always
appreciated,
but
anyway,
so
yeah
ECI
resource
group
is
the
auto-generated
and
acronyms
are
also
frustrating
I.
Don't
know
what
I'm
assuming
RQ
is
our
cue
RG
RG
is
probably
resource
group
that
a
cast
creates
no
charts
in
the
repo.
A
It
was
what
Antoine
asks
and
Gabriele
says:
check
the
bash
script.
Andrea
says:
yep
yep.
We
have
charts
in
the
virtual
cubelet
repo
for
just
a
sure
which
you
can
use
to
okay,
so
I
think
what
we
need
to
do
is
we
need
to
figure
out
what
the
charts
are
actually
trying
to
do
and
then
what
we
would
be
trying
to
do
by
running
the
virtual
cubelet
here
on
our
local
file
system.
So
let's
go
and
listen
check
out
the
charts
and
see
what's
going
on
in
there.
A
A
Directory
here,
okay,
so
with
oh,
it
looks
like
helm,
was
trying
to
to
download
the
tar
balls
here
and
we've
got
two
charts:
we've
got
virtual
Qibla
and
we've
got
virtual
cubelet
for
for
aks.
So
let's
look
at
virtual
cubelet
for
aks,
so
the
first
thing
we
have
is:
we
have
Microsoft
virtual
cubelet
I
got
a
thumbs
up
from
Antoine,
so
it
looks
like
we're
probably
on
the
right
track.
A
Here
looks
like
we're
going
to
seeing
a
lot
of
directives
here,
which
is
I,
have
a
I'm
assuming
it's
going
to
be
okay
and
we
have
a
chart
here
and
in
our
templates
we
have
a
deployment
and
we
have
our
secrets.
So,
looking
at
our
yeah
mole
with
go
text
template
Ian
for
helm,
we
have
some
names.
We
have
our
image,
which
is
defined
and
not
other
values
that
UML
file
we
looked
at
and
we've
got
port
as
your
tenant.
A
Id
subscription,
ID,
okay,
so
I
think
what's
going
on
here
is
this
home
chart
is
attempting
to
deploy
the
virtual
cubelet
to
the
existing
kubernetes
cluster
to
run
in
the
cluster,
which
we
can.
We
can
fall
back
to
that.
If
we,
if
we
want
to
I,
was
really
hoping
we
were
going
to
be
able
to
run
it
on
my
local
file
system
and
actually
get
my
computer
to
basically
behave
as
a
broker
for
ECI.
A
So
we'll
try
to
hack
on
my
local
file
system
a
little
bit
more
and
see
if
we
can't
get
this
thing
up
and
running
and
if
not
will
default
back
to
trying
to
install
this
this
home
chart
so
to
go
back
to
our
original
plan
of
running
from
the
local
computer.
Here,
let's
see
what's
going
on
in
go
so
we
run
our
VK
command
dash
dash
provider,
let's
go
to
Asia
and
we
are
getting
no
pointer
dereference
on
135.
So
let's
see
what's
going
on
here.
A
A
A
If
anybody
on
the
virtual
cubelet
team
can,
let
me
know
if
I'm
on
the
right
track
with
thinking
that
the
helm
charts
are
actually
going
to
be
deploy
in
the
virtual
cubelet
to
run
in
kubernetes
as
a
pod,
and
if
there
are
any
caveats
or
reasons
why
running
it
on
my
local
file
system
might
not
be
the
best
approach
to
demo
in
the
virtual
keep
I'm
kind
of
new
to
the
whole.
You
know
the
new
paradigms
behind
the
virtual
cubelet
here,
so
any
any
help
with
the
higher-level
direction
would
be
greatly
appreciated
and
sure
enough.
A
Aha,
we
were
able
to
find
out
that
that
auth
struct
wasn't
being
initialized
in
the
go
code.
So
let's
go
and
see
why
I
have
a
hunch,
it
has
to
do
with
a
provider
config
file,
something
or
other
and
I
bet
we
can.
We
can
create
one
really
quick
if
we
just
go
and
look
and
see
what
the
code
is
doing
here.
A
So
we
want
to
get
rid
of
this
if
statement
and
we
want
to
go
up
and
make
sure
that
we're
not
importing
that
OS
package
that
we
don't
need
anymore
and
it
looks
like
my
IDE-
was
able
to
take
that
away
from
me.
So
we
call
new
a
CI
provider.
We
define
the
type
of
AZ
auth
and
we
say
if
auth
file
path.
Aha.
Here
we
go
so
we
have
as
your
auth
location,
easy
credential
location,
so
these
were
defined
in
the
helm.
A
Chart
as
well,
and
I
bet
they're
going
to
be
critical
to
define
in
order
to
get
the
virtual
couplet
up
and
running.
So
let's
look
at
this
as
your
location,
Erik,
st.
Martin
says
you
can
run
it
on
your
localhost.
That's
how
heidi
bug
it
and
work
on
it?
Okay,
so
we
still
got.
We've
only
been
working
on
this
thing
for
an
hour,
so
I'm
happy
to
continue
to
have
gotten
it
a
little
bit
more
and
show
off
some
of
my
go
skills
and
really
jump
into
the
source
code.
A
Here,
hopefully
encouraging
people
to
come
and
help
out
and
contribute
to
the
project
and
get
a
virtual
cubelet
running
on
my
local
file
system.
So
we
want
to
figure
out
what
types
of
values
at
ACS
credential
location
is
expecting
and
maybe
as
you're
off
location
as
well
so
reassess.
So,
yes,
VK
will
run
as
a
pod
with
the
charts
or
easy
IKS
install
connector
command.
Okay,
so
that's
starting
to
glue
the
bigger
pieces
together.
A
So
what
what
Ria's
saying
is
we
could
have
had
a
helm
one
option
on
running
the
virtual
qubilah
is
to
use
the
helm
charts
here
in
the
local
repository
which
I
have
it.
Feelings
got
to
be
pretty
easy
to
run,
and
that
would
actually
run
the
virtual
couplet
as
a
potting
kubernetes
and
then
could
the
kubernetes
api
server
would
detect
the
virtual
qiblah
as
a
node,
and
you
would
create
a
virtualized
cubelet
inside
the
cluster
itself.
A
The
other
way
of
getting
it
up
and
running
is
to
use
this
AZ
IKS
install
connector
command,
which
is
what
I
was
trying
to
do
earlier
and
now
that
seems
like
it's
just
another
way
to
go
about
installing
the
virtual
cubelet
connector
on
your
cluster,
but
using
the
azure
command
line
tool
instead
of
home
chard.
And
then
the
third
option
is
what
I
think
I'm
going
for
here,
which
you
guys
know
me:
I
always
have
to
pick.
A
The
hard
way
of
doing
things
is
to
to
actually
try
to
use
my
localhost
to
develop
on
to
actually
get
the
virtual
Kuebler
running
that
way.
So,
let's
finish,
reading
with
the
rest
of
RIA
had
to
say,
which
was
the
experience
you
were
going
to
towards
this
developer
on
vk1,
so
yeah
she's,
just
confirming
that
I'm
taking
the
the
developer
approach
here
and
not
necessarily
the
they
user
approach
here.
A
But
you
can
do
either
so
we'll
spend
a
few
more
minutes
here,
hacking
and
trying
to
figure
out
getting
the
up
and
running,
and
then
wolf
default
back
to
trying
to
get
the
the
helm
chart
deployed
and
getting
the
easy
as
your
command
line
command
working
as
well.
If
anybody
wants
to
vote
for
one
or
the
other
I'm
happy
to
prioritize
that
one,
if
we
think
that
that
one
would
be
a
little
more
interesting,
so
in
the
meantime,
let's
see
if
we
can
find
any
clues
and
documentation
here.
A
So
we
have
a
question:
can
the
virtual
cubelet
be
deployed
on
a
bear
VM,
or
is
it
always
in
a
container
resize
VK
is
a
pod
that
looks
like
a
node
in
kubernetes,
so
I
think
yeah
in
this
case
I'm
going
to
be
running
it
as
a
local
go
program,
but
then
we
can
later
rerun
that
as
a
pod,
if
we,
you
know,
push
it
up
to
a
registry
which
I
think
our
folks
at
Microsoft
has
already
done
and
then
deploy
that
into
kubernetes.
Somehow,
ok,
so,
let's
see
what's
going
on
here.
A
So
is
there
an
auth
file
we
have
off
there,
we
have
off
location,
so
testing
the
Azure
provider,
client,
the
unit
test
for
Azure
provide
require
credentials.
Json
and
the
file
exists
in
the
root
directory
that
you
have
at
your
off.
Ok,
ssin,
set
to
the
credentials
file.
You
can
generate
the
file
by
following
the
instructions
listed
in
the
readme
for
that
package.
A
So
I
think
we
want
to
do
yeah.
This
should
work
here
and
then
what's
the
name
of
the
JSON
file,
we
expect
it
to
to
be
in
okay,
so
in
credentials.
Jason,
okay!
So
let's
go
here
to
the
top
level:
yep
we're
there
and
let's
go
and
let's
create
this
JSON
file,
so
we're
going
to
call
it
my
credentials.
Jason.
A
Yeah,
an
arat
confirms
my
suspicions
that
you
can
run
the
virtual
couplet
anywhere
as
long
as
you
can
connect
back
up
to
the
kubernetes
api
server.
So
this
will
call
this
TGI
K
creds,
dot,
chasin,
wham,
KaBlam
I,
don't
know.
Somebody
said
that
this
last
conference
and
I've
been
saying
it
all
week.
Long
and
I
can't
stop
saying
it.
A
So,
while
we're
waiting
for
that
to
run,
we
can
then
export
our
Azeroth
location.
This
is
starting
to
make
me
feel
warm
and
fuzzy.
I
feel
like
we're
getting
close
here
very
good.
So
now
we
can
export
as
your
off
location
is
equal
to.
Oh
I,
spelled
it
wrong.
So
let's
do
this
move
T,
gig,
teacake,
creds,
dot,
JSON
to
TGI,
K,
creds,
dot,
Jason,
and
now
we
can
run
our
export
command.
A
And
that's
gonna
be
TGI
cake
right,
so
Jason
yep,
thanks
Mike,
you
got
it
so
now,
let's
run
our
make
well
recompile
it,
although
I
don't
think
we
need
to
this.
Is
we
already
had
the
command
in
our
history?
We
can
spend
a
few
seconds
at
compile
time
for
me
to
grab
some
diet
coke
and
see
if
this
thing
will
get
up
and
running
now.
A
A
Region
cannot
be
empty,
okay,
so
good
so
atleast.
Now
the
program
is
starting
to
kind
of
like
say,
like
okay,
you're,
getting
close
you
to
start
setting
these
other
things.
So,
let's
export
our
ACI
region
is
equal
to
I.
Think
it's
east!
Us
is
the
value
we
want.
Try
it
again,
I
think
we
just
got
really
close
everyone.
This
looks
like
exciting
thing
for
the
company.
Okay.
A
So
let's
look
at
our
logs
here.
So
the
virtual
cubelet
logs
start
about
right
here
looks
like
it
detected
our
cube
config
and
it's
creating
a
pod,
oh
cool,
it's
doing
all
kinds
of
stuff.
It
looks
like
we
are
getting
an
error,
though.
Let's
see
what
we
have.
Preview
got:
HTTP
response
status
code,
409
error
code
missing,
subscription
registration.
A
The
subscription
is
not
registered
using
a
namespace,
see
this
URL
for
how
to
register
subscriptions,
but
in
just
in
the
meantime,
let's
split
screen
here
and
let
me
zoom
in
a
little
bit
and
then
resize.
Okay,
that
looks
like
it
should
be.
Okay,
let's
just
see
if
we've
been
able
to
register
as
node
while
it's
still
running
yeah,
we
have
okay
cool,
so
virtual
computes
up
and
running
on
my
local
workstation,
which
is
pretty
exciting
and
I.
Think
it's
probably
for
now.
A
I'm
gonna
ignore
these
these
errors,
and
if
we
have
time
at
the
end,
we
can
actually
go
check
out
that
URL
and
see
if
we
can't
get
these
errors
to
go
away.
But
this
to
me
is
a
really
really
big
Stein,
so
working
in
working
in
Cuba,
Korn
working
in
cops
working
in
cube
admin
actually
getting
the
API
server
to
detect
a
node
is
kind
of
like
the
very
last
step
and
getting
your
node
up
and
running
a
lot
of
times.
You'll
have
a
cube
lit
running.
A
Maybe
it's
configured
close,
but
not
quite
all,
all
the
way
or
the
API
server
is
having
some
trouble
detecting
it.
So
it
looks
like
it's
running.
The
service
looks
happy,
but
then
still
the
API
server
says
no.
No,
it's
not
running,
but
it
looks
like
we
were
able
to
skip
past
all
that
and
get
straight
into
the
API
server
going
yeah.
The
the
virtual
cubelet
is
running
so
now,
let's
go
ahead
and
let's
see
if
we
can't
schedule
a
pod.
Oh
this
will
be
exciting.
A
A
Do
that
I'm
gonna,
just
kind
of
scoot
this
thing
over
here
on
the
left,
so
we
can
just
kind
of
see
if
it
fails
on
us
or
anything
goes
wrong,
but
I
just
want
to
kind
of
look
for
the
fact
that
it's
moving
there
on
the
left
side
as
I'm
going
through
the
rest
of
the
demo
here.
This
is
just
I'm
just
talking
about
how
I
think
about
Linux
but
yeah.
Let's
go
ahead
and
run
our
describe.
Node
here
and
let's
go
and
see
what
we
what
we
found
so
yeah.
A
The
question
is:
I
wonder
how
much
of
a
real
node
functionality.
This
is,
and
that's
a
really
great
point
to
bring
up
and
now
that
we
have
this
thing
up
and
running.
We
can
kind
of
start
to
inspect
than
the
virtual
node,
which
is
really
just
my
macbook
and
see
what
we
have
going
on.
So
we'll
start
all
the
way
up
here
at
the
top.
So
we
have
a
name.
It's
it's
running
as
an
agent.
It's
got
some
some
familiar
looking
labels
on
it,
so
this
always
bothered
me
the
unschedulable
directive.
A
I
I,
really
would
rather
see
schedule.
A
bowl
is
equal
to
true
here,
but
I
think
there's
a
whole
like
history.
Behind
that,
anyway,
I
we
can
schedule
to
it
which
looks
like
you
can't,
but
you
can.
It
says
the
cubelet
is
ready.
It
has
sufficient
disk
available.
So
these
are
the
traditional
cubelet
checks
that
you
see.
Usually
it's
the
people
that
fails
in
some
one
of
these.
Just
that's
just
been
my
experience.
Oh
so
reassess
issue.
A
Number
71
is
what
we're
hitting
with
the
air
you're
getting
okay
cool,
so
we've
already
got
that
issue
and
I
think
it's
being
worked
on,
which
makes
me
happy
so
yeah.
We
have
sufficient
memory
and
our
disk
pressure
looks
okay
and
we
were
able
to
create
a
route.
So
over
here
on,
the
left
looks
like
we're:
we're
missing
an
internal
IP.
We
were
able
to
get.
This
looks
interesting.
A
hundred
I,
don't
gie
I
hope,
that's,
not
gigabytes,
Giga
bits,
maybe
of
memory.
We
have
20
CPU
and
it
looks
like
it's.
A
It
thinks
it's
running
in
Linux,
AM,
D
64
and
we're
running
the
cubelet
version
1.8,
which
is
interesting
because
I
thought
we've
setup
at
1.9
kubernetes
cluster
earlier
today
looks
like
the
cubelet
does
have
a
pod
cider
here
so
I
wonder
a
quick
question
for
the
folks
at
Microsoft.
Is
this
pi
pod
cider
address
advertised
by
the
API
server,
or
is
this
built
into
the
virtual
cubelet
program
and
it
created
it
for
itself
I'm
wondering
where
I
got
this
pipe
hub.
A
Cider
from
seems
like
it
might
be
something
that
the
cloud
provider
would
have
to
figure
out
somehow
but
wondering
where
that
is
coming
from,
and
it
looks
like
we're
in
the
cube
system.
Namespace
and
we've
got
a
couple
of
warnings
here
could
not
create
route
for
node
virtual
cubelet
after
some
amount
of
seconds.
A
So
it
looks
like
a
handful
of
warnings,
but
but
nothing's
really
jumping
off
the
the
page
here
at
me
and
then
I
think,
following
up
on
on
Ria's
talk
from
glue
Kon
earlier
this
week,
I
think
one
of
the
things
I
wanted
to
kind
of
point
out
about
just
the
virtual
couplet
in
general
is
because
it
is
running
virtually
like
it's
gonna.
Have
it's
not
going
to
be
able
to
do
everything?
A
A
node
traditionally
has
done
in
kubernetes,
I
think
some
of
the
backwards
transactions
that
we
talked
about
earlier
with
the
the
exec
eat
and
the
logs
are
going
to
be
there
that
we'll
be
able
to
pull
those
out
of
a
vendor,
but
because
every
vendor
implementation
is
going
to
be
different.
You're.
Now,
at
the
the
risk
of
you
know,
whatever
the
vendor
supports
or
whatever
implementation
you
support,
that
being
able
to
satisfy
different
features
of
the
cubelet
itself.
A
It
says
that
the
pod
cider
is
a
virtual
arranged.
Everything
is
virtual,
it's
virtual
cubelet
and
let's
see
what
Eric
said
so
he
says
right
now.
It
does
a
lot
of
the
reconciliation
status
logs
from
pods
etc,
but
still
has
a
bunch
of
things
to
be
added
for
full
cubelet
compatibility,
so
I
think
yeah,
I,
think
Eric
and
I
were
just
almost
saying.
A
Remember
hearing
a
lot
about
this
at
Microsoft.
This
concept
of,
like
oh
we're,
running
some
workloads
on
kubernetes
and
oh
no,
it's
it's.
You
know
it's
Black,
Friday
and
all
of
a
sudden,
our
workload
like
spiked
through
the
roof,
and
we
can't
create
enough
virtual
machines
or
we
don't
have
enough
hardware
in
our
data
center
to
support
the
spike
so
like
having
the
virtual
cubelet
can
be
there
to
be
really
nice
to
kind
of
act
as
like
a
spillover
right
to
kind
of
help.
A
A
So
anyway,
let's
go
and
let's
try
to
schedule
a
pod
on
the
virtual
cubelet
running
on
my
local
macbook
and
then
I
have
a
feeling
in
my
brain.
It's
just
gonna
start
to
kind
of
explode
with
all
kinds
of
different
things.
I
want
to
do
now.
Cuz
I've
always
dreamed
of
my
my
macbook
being
a
cubelet
and
a
kubernetes
cluster
and
opening
up
my
laptop
and
then
joining
a
kubernetes
cluster
and
then
the
the
API
server
scheduling
containers
on
my
macbook,
so
that
I
can
use
them
for
various
things.
A
But
this
is
me
and
my
weird,
my
weird
way
of
wanting
to
kubernetes
anyway
I,
remember
seeing
an
examples
directory.
So,
let's
go
see
if
we
see
anything
up
here
in
the
examples
directory.
A
So
we
looks
like
we
have
an
engine
X
pod.
We
have
a
nano
server
and
we
have
an.
I
is
pod,
yeah
moles
yeah,
no
files
here
and
let's
see
what
we're
doing
here.
So
we
do
to
find
toleration
and
we
give
it
the
key,
a
sure,
calm,
slash
a
CI
and
it
says
no
schedule
and
it
looks
like
it's:
a
pretty
simple
engine,
X
server,
so
I
was
I'm
wondering
if
we
can
vendor
up
to
a
CI
from
my
macbook
using
this.
This
ya
know
manifest
here.
A
So
let's
take
this
and
let's
cube
cut
a
little
Q
Bechdel
apply
it.
So
where
do
we
want
to
go?
We
want
to
go
to
go
source,
github,
calm,
virtual
cubelet,
virtual
cubelet,
and
we
want
to
k,
apply
minus
f
examples,
engine
xpod,
dot,
yeah
mo
and
I.
Don't
think
we
needed
to
edit
anything,
but
let's
go
check
one
more
time
here
in
our
examples
directory
okay.
So
we
have
it's
a
pod.
It's
in
genetics.
A
It's
gonna
pull
the
official
docker
engine
X
container
image,
we're
gonna
poke
holes
in
84,
43,
DNS
policies,
cluster
first
and
we
have
a
node
selector,
which
is
the
virtual
cubelet,
my
ACI
connector
Lennox.
So,
let's
see
if
we
can
run
this,
we
create
a
pod.
What
namespace
was
it
in
okay?
So
it's
in
the
default
namespace.
Let's
describe
this
pod.
A
A
A
A
Let's
pull
up
our
chrome
again,
we'll
go
to
our
virtual
cubelet
and
let's
go
to
the
issue
tracker
some
diet,
coke
on
the
way
cuz.
That's
just
how
I
roll
pull
up
any
old
issue
and
we'll
come
in
to
issue
71
here,
Oh
Rio
says:
I
need
one
more
cred
I
have
to
add
in
any
idea
which
credential
I
would
need
to
specify.
It
would
be
helpful
as
well.
This
looks
pretty
wordy.
I
can
try
to
read
through
this
really
quick.
A
To
add
this
service
principle,
I've
already
done
that
can't
seem
to
create
the
assignment
created
the
ion
contributor
on
resource
group
in
getting
wow.
This
is
SP
boy
or,
if
not
SD
Boyer,
who
is
a
friend
of
mine
from
open-source.
So
it's
like
well,
Sam
changes,
github
name,
but
it's
just
a
different
person.
A
The
PIM
air
is
different
from
the
authorization
error.
I
get
the
error
from
the
AC
command
line
now
copy.
The
number
after
service
principle
in
your
aks
resource
group
add
an
iamb
contributor
role
to
the
service
principal
number.
So
I
think
this
is
what
we
want
to
do
here.
So
I
think
this
is
the
the
solution.
A
We
just
got
to
kind
of
figure
out
what
the
right
combination
of
steps
we
need
to
do
are
in
order
to
make
this
happen
here
so
to
add
the
service
principle,
so
you
cat
the
JSON
file
which
we
have
on
our
local
filesystem
and
you
copy
the
number
after
service
principle.
So,
let's
see
what
this
looks
like,
so
let's
get
some
more
wiggle
room
here,
so
it
was
service
principle.
A
Interesting,
an
interesting,
interesting,
okay
Riya,
if
you
could
do
me
a
favor
and
be
a
little
more
verbose
on
a
good
next
step.
If
you
have
anything
in
mind,
that
would
be
helpful
just
turn.
Some
of
this
is
a
little
bit
challenging
because
I
haven't
really
looked
at
any
of
it
before.
A
A
A
So
yeah
so
Rio
says
add
that
role
on
the
right
track
and
then
we
have
somebody
else
say:
is
it
somehow
related
to
our
back,
which
I
don't
think
we're
running
our
back
on
this
cluster,
so
I'm,
not
really
sure
or
I,
don't
really
know.
If
it's
going
to
be
an
Arabic
issue,
it
looks
like
we
had
to
go
through
and
we
had
to
create.
We
had
to
define
these
environmental
variables
so
I'm
just
wondering
we
have.
A
A
A
A
A
So
this
is
the
arrow
we're
getting.
The
next
error
is
a
service
principle
issue.
Looking
for
a
workaround
to
add
the
service
principle,
you
cat
this
file
copy
the
number
after
service
principle,
so
I
guess
when
I
created
my
service
principle,
my
file
might
have
been
created
slightly
different
and
that's
why
I'm
not
able
to
find
this
number
after
the
service
principle
in
your
IKS
resource
group,
add
the
Ayhan
contributor
rule
to
the
service
principle
number
and
delete
the
pod.
A
Operation
failed
with
bad
requests.
I
was
seeing
what
this
easy
command
was
here,
created
the
INF
contributor
and
resource
group
and
is
now
getting
failed
to
find
any
PEM
data.
You
should
be
good
to
go.
That
error
is
because
the
cert
is,
he
need
pass
through
four
key
Bechtel
logs,
but
I,
don't
know
what
Oh
W
means
the
connector
will
work.
Fine.
The
PIM
error
is
different
from
the
authorization,
air
tri
ACS
service
principle,
tri.
What
with
ACS
surface
principle
I,
don't
really
know
what
that
means
anyway.
A
That's
going
to
be
running
in
kubernetes
when
we
get
done
so.
Let's
go
back
up
here
to
a
virtual,
cubelet
and
I
think
we
can
go
to
charts
and
we
can
do
vert
cubelet
for
aks,
so
I
think
the
helm.
Syntax
here
is
gonna,
be
helm,
install
charts
and
then
we
want
virtual
cubelet
aks.
Let's
see
what
that
does
felt
download.
A
A
A
Helm,
install
T,
GI
k,
VK
minus
F
charts
and
failed
to
download
so
I
think
we
had
it
right
there,
maybe
I
just
add
it
to
the
end.
Without
the
name.
Let's
try
that
BAM
go
to
the
end
of
our
command.
T
GI
k,
VK
error,
felted,
l,
no
t
j,
VK,
o
so
Oh
Nikolas
says
no
F,
but
you
have
to
reference
it
relationally,
so
I
just
doesn't
assume
the
prefixing.
A
Period
/
there
bam
thanks
for
the
clue.
Nikolas
greatly
appreciated.
Okay.
So
now
let's
do
a
helm
LS
and
we
have
ideal
sasquatch
and
let's
see
what
pods
are
running,
so
it
should
be
in
the
default
namespace
since
we
didn't
specify
otherwise
so
okay,
so
this
is
running
okay
cool.
So
let's
get
our
logs
here.
A
I
didn't
specify
any
off
anything,
so
I'm
wondering
if
it
pulled
it
out
of
my
local
file
system,
I,
don't
I,
don't
my
local
environmental
variables
that
were
set
on
my
local
file
system
or
if
we
need
to
go
in
to
find
out
somehow
and
I
think
that's
what
the
documentation
we
were
looking
at
earlier
was
doing.
So,
let's
get
our
Pio.
A
A
A
So
now
we
can
go
in
and
we
can
try
to
reschedule
this
pod
and
see
if
it'll
get
rescheduled
on
the
the
newly
created
virtual
cubelet
here.
So
let's
kill
nginx
k,
delete,
P,
oh
and
because
it's
a
deployment,
the
schedule
should
reschedule
it
for
us
automatically
and
let's
get
our
pods
and
see
what's
going
on
now,
Oh
Oh
cuz.
Was
it
not
a
deployment?
It
was
a
pod.
Wasn't
it
so,
let's
apply
it
again,
so
it
was
examples.
A
A
A
Reading
ACS
credentials
file,
so
this
is
at
the
ACS
Asherah
Jason
and
it
said
loaded,
ACS
credential
file
successfully.
So
it's
getting
creds
somehow
and
it
says
node
virtual
cubelet
with
the
OSI
type
Lennox
has
been
registered
and
it
says
TLS
failed
to
find
any
PEM
data
in
certificate
input.
So
let's
go
back
to
this
issue
number
71
and
see
what
we
got
going
on
here.
A
A
And
we
had
some
him
you'll
get
a
pen
de
terre.
If
you
don't
feed
a
sir,
but
that's
fine.
If
you
don't
want
to
use
Kubek
two
logs
on
you
see
I
contain
it,
so
this
thing
should
be
working
and
then
the
only
other
pen
created
I
am
contributor
and
resource
group
and
getting
failed
to
find
PEM
data.
So
let's
look
in
the
law
and
the
documentation
here
and
see
if
there's
any
clues
on
installing
this
hound
charge.
A
So
we
have
these
directives
that
are
not
said,
but
they
should
it.
They
should
be
being
pulled.
From
that
add
your
credentials
file
yeah.
It
looks
good
to
me
to
RIA.
So
it's
it's
pretty
interesting.
Why
it's
not
wanting
to
schedule!
Maybe
it's
a
nut
dreaded
case
of
eventual
consistency
and
we're
just
waiting
for
some
things
behind
the
scenes
to
to
take
over
no
we're
still
pending
alright
engine
next
pod.
A
Here
interesting,
interesting,
stuff,
cool
well,
I
mean
at
this
case
I
think
you
know,
we've
demonstrated
that
the
the
virtual
cubelet
kind
of
how
it's
supposed
to
be
running
the
concept
of
ending
out
to
a
party
like
far
gate
or
like
Microsoft
Azure,
is
a
CI
and
we've
learned
about
the
the
different
interfaces.
You
can
satisfy
two
to
create
your
own
virtual
cubelet
and
to
come
and
create
your
own
provider
in
the
program
and
we've
learned
that
you
can.
A
You
can
run
it
locally
and
use
your
your
local
workstation
as
a
broker
to
one
of
these
container
services
like
far
gate
or
a
CI,
and
you
can
also
run
it
in
a
pod
in
the
cluster
itself
which
to
me
really
that's
the
real
virtual
nosov.
It
is.
We
have
some
set
of
nodes
running
in
the
cluster
and
then
we
deploy
a
pod
and
the
virtual
cubelet
kind
of
comes
from
than
air
and
it's
virtualized.
And
now
we
have
this
magical
new
node.
A
That's
actually
going
to
go
and
and
vendor
out
to
to
these
services
outside
of
the
kubernetes
cluster
itself
and
then
just
by
defining
a
simple
line
in
your
your
pod
spec.
Here
this
node
selector
kubernetes,
io
hostname,
is
equal
to
the
name
of
the
the
vendor
you
want
to
describe.
You
can
actually
control
which
pods
are
going
to
get
scheduled
on
which
nodes.
A
So
this
is
a
pretty
powerful
construct
and
I'm
really
excited
to
see
this
project
moving
forward
and
I'm
going
to
tinker
on
it
a
little
bit
afterwards
to
try
to
try
to
get
this
up
and
running
with
Microsoft
Azure
ACI
and
actually
get
one
of
these.
These
examples
deployed
and
the
beauty
of
the
new
tgia
repository
is
after
I
get
all
this
done.
I
can
go
and
back
port
and
document
all
this
and
it
will
become
a
resource
resource
moving
forward.
A
So
looking
forward
to
seeing
questions
and
answers,
I'm
gonna
open
up
an
issue
or
follow
up
on
H
271.
With
my
experience
here,
and
hopefully
we
can
figure
this
out
moving
forward
and
then
hopefully
I
you
know
folks
can
come
and
contribute
to
the
TTI
K
repository
with
any
questions
they
had
or
suggestions
or
ideas
for
future
episodes.
Moving
forward
for
t,
GI
k
and
without
further
ado
I
think
you
know,
I'm
gonna
call
it
here.
It's
been
about
an
hour
to
30
minutes
and
yeah
I'm
gonna.
Let
everybody
get
back
to
their
weekend.