►
From YouTube: Kubernetes SIG On-Prem Meeting 20170315
Description
Kubernetes Special Interest Group On-Prem Bi-Weekly meeting
A
All
right,
it's
like
two
minutes
after
five,
my
time,
nine
PST
I
think
so
I
think
we
should
start
what
we
have
for
today
is
another
demo.
This
time
it
will
be
stuck
by
job
kaiser
and
be
the
next
agenda
item.
I
think
it's
very
interesting
is
how
to
expose
your
workload.
They'll
fight
to
the
to
the
external
world
looks
like
there's,
there's
too
many
options,
probably
so
I
hope
it
would
be
interesting
discussion
discussion
what
people
are
doing
actually
so.
B
C
C
All
right
we'll
deal
with
it.
Ok,
so
I
have
very
few
slides
and
really
just
want
to
tell
you
what
I'm
about
to
show
you
and
then
we're
going
to
show
you
so
I'm
with
a
company
called
stack
IQ,
we
have
an
open
source.
Our
open
source
installer
is
called
stachy.
We
open
sourced
it
about
a
year
and
a
half
ago,
and
it
has
has
been
the
basis
for
the
installation
of
a
fair
number
of
enterprises.
It
was
based
on
the
rock
project.
C
If
any
of
you
are
familiar
out
of
of
UCSD,
we
have
two
of
the
original
developers
of
that
project
and
we
use
it
to
install
just
about
everything
at
this
point.
What
it
does
is
installs,
mostly
bare
metal.
We
do
VMs,
we
have
one
client
uses
it
to
install
not
only
their
bare
metal
machines,
but
also
a
fairly
large
number
of
their
VMware
machines
as
well
kickstart
based,
it's
not
image-based.
We
do
parallel
formatting
of
disk,
which
is
unique
in
the
which
is
unique
in
terms
of
the
installers
right.
C
So
this
is
just
your
basic
tftp,
GTP
pixie
stuff
grab
a
kickstart
and
install
a
machine.
You
know
the
partitions,
the
RPM
software
packages,
all
the
post
installation
scripts,
the
same
stuff
that
you
have
probably
done
at
some
point
in
your
career
with
regard
to
getting
a
group
of
machines
up
to
a
functional
level.
That
is
consistent
right
things
that
we
do
that
are
different.
C
Are
the
parallel
formatting
of
disk
and
I
need
to
qualify
this
simply
because
the
we
do
don't
do
parallel,
formatting
disk
and
stack
III
to
47
dot,
x
versions
of
some
to
us
and
redhat
that
will
change
with
the
next
release
which
hit
it
should
be
out
in
about
three
or
four
weeks,
and
it's
one
of
its
one
of
the
big
differences
between
us.
It
doesn't
matter
so
much
since
Lawrence
said
to
us
runs
to
XFS,
but
we
have
a
guy
on
our
blacklist
asked
about.
You
know.
Why
is
my
Y?
C
Is
the
formatting
of
a
48
dry
machine
that
has
six
terabyte
drives
for
each
of
those
48
drives
taking
so
long,
and
that
was
that
was
why
I
was
something
that
we
had
missed
in
that
particular
Rev.
So
we
also
do
parallel
sharing
of
our
pms
during
the
installation.
This
makes
it
so
that
it's
a
peer
peer
BitTorrent,
like,
although
we've
stripped
out
a
lot
of
stuff,
so
that
it
works
faster
on
a
closed
Network
and
all
that
stuff
getting
solved
in
same
time.
Right
now,
we're
since
OS
Red,
Hat
aced.
C
The
pro
version
has
ubuntu
and
slice
on
it,
but
I
am
pushing
hard
you
get
you
bunt
to
into
the
into
the
open
source
version
with
the
next
release,
and
if
I
can
do
that,
that
will
open
the
door
and
the
next
release
awesome
stack.
Acordo
makes
it
a
lot
easier
to
roll
in
additional
os's
and
use
their
native
language
for
an
installing
machine.
So
core
OS
and
rancher
OS
and
alpine
are
sorts
of
things
that
I'd
really
want
to
start.
C
Looking
at
we
do
networking
multiple
subnets
un
in
bonding
all
that
it's
definable
in
a
spreadsheet
before
you
laugh
about
spreadsheets
a
lot
of
the
people
that
we
deal
with,
think
it's
a
revelation
to
be
able
to
define
their
partition
schemes
in
their
control
schemes
by
spreadsheet,
so
I
know
it's
really
old
school
and
it's
not
yamo
or
any
of
the
other.
Egregious
sorts
of
configuration
language
is
not
there,
but
it
works
controller
config
as
well.
We
talked
to
the
control
of
configuration.
C
My
control
configuration
is
is
able
to
do
to
set
up
your
controller
configs
from
the
installation
so
that
you
never
have
to
touch
the
keyboard
and
monitor
ever
install
them
ever
again.
So
we
cover
the
lsi
in
HP
controllers
and
most
we
catch
most
of
them
with
with
that
support.
So
basically
what
stachy
it
allows
you
to
do,
whatever
you
can
do
with
linux,
except
on
a
whole
bunch
of
machines
all
at
once,
and
our
goal
is
to
make
it
fast.
C
So
the
difference
between
Don,
five
machines
and
500
machine
for
us
a
fairly
minimal.
In
terms
of
time
we
have
one
user
who
installs
their
infrastructure
with
you,
just
a
key
to
install
the
base
OS
and
partition
all
their
disk
and
drives
and
then
does
an
ansible
run
to
set
up
their
Hadoop
configuration
and
that
takes
them
14
minutes
on
280
machine.
So
it's
pretty
fast,
so
stachy
and
communities
and
dr.
really
it's
all
three,
because
I
haven't
explored
rocket.
C
Yet
so
one
of
the
things
about
being
the
director
of
open-source
engineering
means
that
I
get
to
play
with
real
cool
stuff
and
make
it
available
as
an
application
stack.
So
we
take
care
of
the
bare
metal,
but
you
can
layer
applications
on
top
of
that.
If
you
want
a
lot
of
people
will
simply
just
install
the
bare
metal,
and
then
they
run
their
DevOps
tools,
poppet
fall
anthem
or
whatever,
but
we
have
this
notion
of
a
palette,
but
I
inherited.
This
is
metaphor
for
forced
a
key
which
is
the
data
warehouse.
C
So
you
add
configuration
with
carts
and
with
pallets
and
we'll
get
into
a
little
bit
of
that
in
here
a
little
bit.
So
what
this
is
going
to
be
is
going
to
be
a
mayor
of
metal,
installation
of
commitment
to
burnet
ease
with
docker.
This
version
is
not
the
one
that
is
currently
available
on
the
github
site.
That
was
our
face
phase.
C
One
version,
which
means
it
just
gets
it
up,
gets
it
working,
it'll
it'll
function,
and
this
version
that
I'm
about
to
show
you
today
will
have
pls
sort
of
certificates
everywhere
and
a
little
more
lockdown
than
than
you
than
the
first
one
excuse
me:
I
CD
info
NLD,
we're
doing
three
node
for
SUV,
we're
not
doing
storage
back
end
on
that.
I
haven't
dealt
with
storage
in
any
of
this,
yet
Seth
OS
73
on
the
back
end,
except
there's
a
410
kernel
that
will
be
installed
and
I'll.
C
Probably,
why
did
that
was
because
I
wanted
to
use
the
overlay
as
best,
because
if
your
state
start
looking
at
the
documentation
%
OS,
they
say
I'd,
don't
use
the
the
device
mapper,
because
it's
horrible
performance,
ok,
cool
used
to
overlay
FS
and
in
73
and
72
that's
listed
as
a
you
know:
preview
software,
so
I
stole
a
curl
from
410
and
that
seems
to
be
working.
Fine
all
modes
are
on
a
private,
submit
and
I'm
accessing
the
public
subnet
through.
C
C
If
you
want
to
be
that,
but
I
don't
think
that's
advisable
as
soon
as
I
can
figure
out
how
to
mirror
a
doctor
registry
rationally,
it's
very
likely
that
will
make
it
so
that
there's
a
localized
docker
registry
or
you
can
call
these
things
from
prior
to
that
and
phase
one
I
was
pushing
hard
images
into
a
private
doctor
registry,
long-term
I,
don't
think
that's
sustainable,
and
so
some
of
the
caveats
I'm
still
pretty
new
at
Corrine
amis
I,
don't
understand
all
the
question.
They
certainly
don't
even
understand
all
the
answers.
C
Yet
so
they're
going
to
be
3
requires
house
and
we'll
talk
about
those
and
then
we're
going
to
do
the
customization
through
some
spreadsheets
and
then
we're
going
to
install
them
and
it'll
be
incredibly
incredibly
awesome
and
I
will
address
phase
1
phase,
2
and
phase
3
at
the
end
of
this.
So
we
are
going
to
start
off
with
a
how
much
I.
Don't
let's
talk
about
half
a
month.
A
C
C
Typically,
I
will
do
a
deuce
like
cooking
demonstration.
I'll
start
the
installation
of
management,
the
stacking
management
server
and
then
cut
to
the
running
server
so
pretend
that
I
did
the
actual
in
demo
of
the
start
of
the
statue
front
end
and
now
we're
cutting
to
a
running
server.
Let
me
just
say
two
things
about
the
Statue
front
end.
C
You
can
either
use
the
minimal
I
so
that
we
have
to
install
both
a
the
statue,
software
and
a
minimal
Center
West
70
s
and
then
do
everything
it
make
your
adaptations
from
there
or
you
can
start
with
a
sense
of
west
vanilla
server
and
you
can
add
the
stachy
palette
on
top
of
that
with
a
trip
that
comes
on
that
I.
So,
okay,
so
there
are
two
ways
to
do
that:
they're
in
greater
detail
on
the
deck
e.com
site
on
the
github
at
stack
at
qcom.
C
So
what
we
have
here
is
I
have
a
set
of
cops.
So
when
you
first
install
a
machine,
this
is
what
you
get.
You
get
the
stachy
talent
in
the
OS
pal.
The
statute
palette
has
all
the
software
that
makes
all
this
run.
So
you
need
that
the
OS
is
a
minimal
s.
So
you
need
that
as
well
and
we
have
a
front
end.
We
have
you
have
one
host,
we
have
the
front
end,
that's
it
and
everything
else.
Now
we
want
to
install.
There
are
a
couple
of
ways
to
install.
We
can
do
a.
C
C
C
C
You
should
be
able
to
hear
that
first
one
and
since
I'm
so
lazy
that
I
can't
send
getting
up
and
walking
12
feet
too
yeah
yeah
here
when
he
ought
to
hear
when
all
time
go
up
the
thing
that
I
can't
tan
is
walking
12
feet
to
actually
push
a
button.
So
I've
made
it
so
that
I
have
wired
ipmi
and
wake
on
land
into
those
machines.
C
What
I
tell
them
to
do
is
the
way
to
get
a
hold
of
your
infrastructure
is
to
start
with
the
very
base
right
make
it
so
that
you
don't
layer
anything
on
top
until
you
have
had
some
experience
with
with
with
doing
this,
so
you
install
one
back-end
machine
here.
We
go
we're
discovered
and
we're
going
to
wait
till
there's
a
little
star
here
which
will
tell
us
when
that
kick-started,
which
is
starting
to
kick
start.
C
So
you
install
one
backend
machine
and,
if
you're
happy
with
it,
then
you
install
all
them
so
that
you
can
get
them
under
your
control
once
they're
under
the
control
in
in
the
front
end.
Now,
you
can
start
adapting
them.
The
cost
of
reinstalling
is
fairly
minimal
in
terms
of
time,
so
it's
okay
to
kind
of
treat
them
as
disposable,
which
is
what
they
are,
should
put
them
up
beneath
them
going.
C
So
once
this
kick
starts,
we're
going
to
start
customizing
their
back-end,
we're
going
to
add
some
more
hosts,
we're
going
to
also
set
up
the
software
and
then
we're
going
to
reinstall
all
these
and
then,
while
we're
reinstalling.
All
them
takes
about
20
minutes
for
Cooper
natives.
Where
will
you
be,
will
be
some
other?
Take
some
questions.
Ok,.
B
C
My
interface,
my
interfaces,
so
this
machine
actually
has
two
interfaces:
I
had
one
on
the
private
man,
so
I
have
two
networks
here:
one
private
one,
public
I,
actually
don't
plumb
the
public
on
the
back
end
node,
I
plumbed
it
on
the
front
end
node,
so
I
can
get
out
and
and
but
it
will
discover
everything
that's
on
here.
These
machines
have
I
infiniband
on
them.
C
When
I
plug
that
in
all
the
infiniband
I
interfaces,
all
forget
it
getting
discovered,
and
what
that
allows
me
to
do
is
that
after
the
fact,
if
I
decide
that
I'm
going
to
add
a
network,
I
can
add
the
network
and
I
can
add
the
IP
addressing
scheme
to
the
database
and
resync
the
network
config
for
the
back
end
note
without
doing
a
reef
reinstall
and
will
automatically
get
that
network
punching.
Ok.
So
here
we
go
we're
we're
partitioning
here
on
the
left
on
back
end
0
and
now
we're
starting
to
install
the
RPM.
C
C
Dependencies
I
also
would
and
I
want
so
I
want
everything
there
if
I
need
to
pull
it
so
I
have
a
stack
e
as
Santa
OS
updates,
which
I
also
need
because
I
want
whatever
tree.
Should
this
actually
can
be
created
with
a
command
called
stachy
creates
that
create
mirror,
which,
if
you
give
it
a
repo
ID
or
a
URL,
will
pull
down
RPM
and
turn
them
into
an
iso.
Pallets
are
essentially
iso
files
that
contain
configuration,
or
software
or
close
so,
but
in
the
instance,
is
sent
to
us
and
sent
to
a
seven.
C
It's
just
a
bunch
of
RPM
for
stachy
be
awkward,
cooper,
Nettie's
and
Prometheus.
These
are
available
on
the
github.
The
most
current
version
should
be
available
in
the
next
week
on
its
promise
with
documentation,
which
is
actually
a
stumbling
point
right
now
right,
and
these
contain
both
RPM.
So,
like
second
clue,
bernetti
says
the
colonel
rpms
that
are
cold
for
the
more
advanced
colonel,
but
it
also
has
configuration
that
sets
it
up.
So
we're
going
to
do
is
we're
going
to
add
these.
C
The
next
version
of
stachy
is
actually
going
to
get
away
from
I
Sosa's
palette,
so
you
should
be
able
to
add
arbitrary
sistas
directory
with
packages
or
software,
so
it
should
be
a
little
easier.
I.
Don't
have
done
a
preview
of
that
myself
yet
so
it
should
be
HD
to
see
how
it
goes.
Okay,
so
where
we.
C
Through
viewing
pallets,
we've
got
an
installation
going
and
we
want
to
demo.
Oh
I
was
going
to
talk
about
I'm,
going
to
talk
about
another
way
to
customize,
so
the
typical
way
to
so
the
typical
way
to
customize.
Your
particular
set
of
things
is
with
a
cart,
so
pallets
are
kind
of
advanced
to
build
and
generally
they'll
come
from
us
or
go
come
from
people
in
the
communities
of
experience
building
with
them.
But
you
can
you
can
you
can
set
yours
with?
C
Ok,
we
have
a
site,
custom,
cart
and
it's
not
currently
enabled
anywhere-
and
there
are
two
directories
that
matter
so
the
RPMs
directory
here
that
matters,
because
if
there
are
particular
rpms
that
you
want
that
are
not
available
in
any
of
the
pallets
that
you
voted,
then
you
can
add
them
here
and
they'll
be
available
during
installation
to
all
the
machines.
So
the
other
one
is
the
nodes
directory.
So
this
is
going
to
freak
everybody
out.
B
C
It
always
does
because
it
says
XML
a
penis,
so
let
me
show
you
what
we're
going
to
do
here.
So
we've
created
a
kick
start
generator.
What
it
does
is
takes
a
bunch
of
snippets
of
configuration
turns
them
into
one
great
big
kickstart
file
that
kicks
our
file.
If
you
not
familiar
with
kick-start,
essentially
just
maps
to
the
installation
code,
when
you
make
a
request
for
reps
installous
into
it's,
a
red
hat
box,
is
this
a
big
shell
script
with
special
section,
our
special
section
get
map
in
HTML
tags?
C
D
C
Going
to
go
through
so
this
is
this
is
what
I
typically
customize
some
of
my
friends
to
do
on
my
back
in
there,
so
I
always
put
it
off
the
web
on,
because
I
can't
stand
not
knowing
who's
doing
what
so
cats
almost
always
on,
because
I
never
know
when
I
do
weird
networking
things
I
want
to
disable.
So
a
package
tag
will
essentially
just
take
this
package
and
put
it
in
the
back
of
second
room
kick-start.
C
If
I
disable
a
section
with
disabled
equals,
1
I
can
I
get
what
will
happen
if
you
get
the
minus
sign
from
removing
all
the
the
wireless
stuff.
I
mean
wireless
on
these
machines.
They
don't
have
wireless.
The
IV
TV
have
no
idea
why
guest
s
gets
installed
as
a
base
as
part
of
the
center
of
space,
but
it
does
and
then
I'm
making
sure
that
core,
which
is
part
of
a
group,
is
going
to
be
installed
as
well
as
well,
by
putting
a
tag
called
meta,
so
package
tags
can
have
other
things
within
them.
C
C
C
C
C
C
C
The
setup
of
the
attributes
in
the
storage
partition,
but
here
I'm,
going
to
just
go
ahead
and
load
all
these
you'll
be
amazed,
and
you
I'm
sure
some
of
your
like
CS
me
I,
actually
get
that
quite
a
lot,
and
until
we,
when
we
thought,
was
a
little
crazy
too,
and
so
we
met
a
significant
number
of
people
in
industry.
That's
the
way
to
manage
their
stuff.
They
set
it
out
to
get
all
the
data
that
they
need
from
it
before
they
install
anything
at
all.
So
this
is
so.
C
This
is
one
way
that
we
decided
it
was
the
best
way
to
do
it
exactly
so
so
now
I
have
a
bunch
of
hope
cool.
So
now,
let's
set
some
things
out.
So
let's
see
we
want
partition
by
default.
Partitions
will
be
a
/
16
gig,
flash
5,
20
gig
swap
of
8192,
and
everything
else
will
be
go
on
a
particular
partition
called
state
partition.
One,
that's
the
default,
so
I
actually
want
to
have
my
own
partition
trips
in
particular.
C
B
E
C
A
C
Cool
all
right,
ok,
so
we're
going
to
ask
them
after
we
filed,
because
there
are
some
in
order
for
the
queue
brunetti
files
to
work.
It
requires
that
some
attributes
get
set.
Attributes
are
key
value
pairs
that
are
used
in
the
kickstart
so
that
you
can
set
things
dynamically,
which
means
you
only
have
to
have
one
kickstart
file,
not
a
bunch
for
a
bunch
of
different
files.
So
we're
going
to
do.
E
C
And
Prometheus
does
also
stachy.
Docker
also
has
some
stuff
for
the
front
end,
but
you
only
need
it.
If
you're
going
to
run
the
statue
docker
palette
on
its
own
without
super
Nettie's
prometheus
is
going
to
set
up
on
this
particular
front
end.
Typically,
I.
Don't
do
that,
but
I
wanted
to
be
able
to
get
a
sense
of
what
kind
of
monetary
was
available
out
there.
Ok,
so
I
think
we
have
everything
added.
C
C
C
Ok,
so
the
reason
why
we
set
new
gifts
when
you
first
install
set
of
machines
that
you
added
with
partition,
is
we
don't
want
to
do
anything
to
your
machines
until
you're
ready
to
do
it?
New
fists
will
really
do
the
partitioning
there's.
Also
a
consonant
and
controller
attribute
called
new
controller
that
will
reset
your
controller
configuration
and
put
it
into
configuration.
You
expect
what
that
does,
except
you
to
a
known
state,
every
reinstall
after
that,
we'll
not
reform'
at
disk
unless
you
particularly
throw
the
new
discs
flag.
C
What
that
means
is
that
if
you
have
machines
with
with
large
numbers
of
discs-
and
you
want
to
hang
out
to
the
data-
that's
on
them-
we're
not
going
to
reinstall
it
on
a
rien
sup.
We
won't
reform
at
that
on
a
reinstall
unless
you
specifically
request
it.
So
in
this
instance,
we
make
it
so
that
you
have
to
physically
find
your
foot
to.
In
order
to
be
able
to.
You
know
sure
you
toe
off,
so
what
these
machines
are
going
to
do
is
they're
going
to
install.
C
We're
fauna
grafts
we're
going
to
install
firefox
and
mcdowell
08
here,
okay,
so
from
so
we're
going
to
let
this
go
and
I'll
answer
questions
in
between
and
we
can
actually
move
on
to
the
other
one
as
well,
and
you
just
keep
an
eye
on
it
when
this
comes
back.
So
let
me
finish
up
here,
so
there
are
phase
1
phase,
2
and
phase
3
phase.
One
is
currently
on
github
into
space
too,
so
that
it's
secure
with
a
minimal
needed
for
production
set
of
machines.
This
is
not
include
cube,
DNS,
I,
don't
know.
C
If
anybody
has
seen
my
questions
on
the
list
of
past
few
days.
I,
like
has
anybody,
said
of
cube
DNS
on
bare
metal
and
people
were
like.
Why
would
you
want
to
do
that?
And
after
thinking
about
that
I'm
not
really
sure
I?
Would
you
want
to
do
that?
Because
it's
not
currently
functioning
next
phase?
Three
is
going
to
be
core
OS
as
an
option.
You
can
do
that
from
the
same
trend
and
disk
integration.
I
haven't
done
anything
with
this
callicoon
we've
figure
out
the
private
registry
thing
and
rocket.
C
That
I
know
one
peep
one
company
here
that
use
dat,
cube
ABM
for
the
production
and
they're
havin
they're
havin
problems
with
it.
So
if
it's
alpha
I'm
not
going
to
touch
it,
if
its
data
I'll
consider
it
and
again
I,
don't
know
enough
to
be
able
to
make
the
best
decisions.
So
that's
a
big
promo
reason
why
drawing
sig
on
Prem
and
looking
at
the
flak
channel-
and
it's
been
really
helpful,
so
Tomas.
If
you
want
to
just
go
ahead
and
go
on
to
the
next
and
so.
C
A
A
Well,
we
can,
we
can
keep
it
there,
okay,
yeah,
so
thankful
them
again
and
let's
proceed
the
second
agenda
item,
which
is
essentially
how
people
expose
their
workloads
to
the
outside
world
and
pros
and
cons
on
different
methods,
so
I
think
again,
which
we
have
a
situation
where
there
are
some
standards
defined.
If
you
run
on
cloud
cloud
environments
like
gke
or
or
AWS,
but
on
pram
I.
A
D
G
A
G
E
A
D
When
we
were
building
the
tectonic
ingress
setups
for
AWS
and
for
bare
metal
being
explored
this
a
lot
and
I
guess
there
are
a
lot
of
pitfalls
actually
concerned
around
ingress
and
how
you
do
it
where
to
begin,
we
use
the
engine,
eggs,
ingress
controller.
We
explored
the
traffic
one
for
a
while.
We
settled
on
running
ingress
yourself
rather
than
like.
You
said,
there
is
no
load,
balancers
type,
the
equivalents
on
bare
metal
and
I.
D
Don't
know
this
one
even
really
makes
sense,
necessarily
among
the
when,
when
users
are
setting
up
that
culture,
they
have
two
options.
You
can
use
what
they
will
have
two
options
soon,
using
host
board
finding
you
can
run
an
ingress
controller,
bind
to
host
ports
on
particular
node
and
that's
sort
of
as
it
easy
task
for
set
up,
because
it
means
people
just
need
to
know
how
to
set
up
dns
records.
D
As
long
as
you
can
get
Packard's
to
resolve
to
that
particular
to
use
that
particularly
dns
name
and
resolve
to
a
cluster,
your
packets
will
arrive,
and
you
can
do
your
democracy,
multiplexing
and
terminations
there
at
the
culture.
The
other
approach
first
started
using
a
cluster
IP
route
should
sort
of
advanced.
D
D
D
It's
brilliant
on
their
part,
but
even
on
each
of
us
we
don't
do
this.
That
integration
is
enabled
for
people
who
want
to
throw
money
at
the
problem,
but
you
can
just
run
the
whole
point
of
a
virtualized
representations
is
to
have
run
as
many
replicas
of
ingress
as
you
need,
for
whatever
your
needs
are
and
then
do
that
all
through
one
point.
D
So
that's
where
that's
right
as
a
cost
in
there,
and
so
generally
for
my
personal
sculptures,
I
turn
those
off
and
then
deploy
an
ingress
controller
similar
to
what
tectonic
does
and
then
among
those
you
can
pick
host
port
node
for
it
closer
I,
repeat,
and
we
found
that
the
right
solution
depends
on
your
platform
a
lot
and
up
at
the
top
you
can
see
on
bare
metal.
You
can
do
all
three
of
those
and
they
have
their
own
closing
costs.
A
D
A
A
A
D
F
At
one
club,
we've
done
some
stuff
with
Calculon
vgp
and
I'm
doing
a
separate
reflector
that
looks
at
the
automatically
updates
the
bill.
Do
that
sodding
service.
Without
a
meeting
at
the
ingress
controller,
you
can
directly
to
the
quest
trapeze
with
a
rattle,
yes
for
the
rattle
ID
as
long
as
the
space
around
well
within
your
infrastructure,
yeah.
A
I
think
we
have
the
same
approach.
We
wrote
this
external
IP
controller,
which
works
I,
think
in
it.
Well,
when
it
when
it
was
written
it,
it
was
intended
to
be
used
with
Calico
and
essentially
does
the
same
right.
The
entry
or
are
you
had
the
code
to
get
to
give
some
more
details
about
the
college.
G
Yeah
and
yeah
I'm,
so
this
is
big,
no
less
the
same
approach
that
was
like
that
described
because
I
just
waited
the
link
to
the
chart
and
the
description
of
this
is
he
impeded
by
asking
stuff
that
we
did
with
this
calico
and
separate
those
reflectors.
Yes,
I'm
pleased
that
this
is
the
same
as
was
described
right
now,
oh
so
I'm
yeah
also
I
pick.
Add
our
external
etiquette
also
has
this
different
operation
mode
right,
but
this
is
probably
like
six
too
simple
for
football.
Michael,
probably.
A
A
D
Thing
you've
really
been
looking
into
a
chorus.
The
sort
of
approaches
been
thinking
that
get
packets
typically
related
cluster
and
the
multiplex
at
the
cluster,
rather
than
like
designs
that
have
a
load
down,
sir,
that
is
constantly
notified.
It
was
changing,
but
probably
other
people
have
other
opinions
and
might
be
more
interested
in
that.
A
F
E
E
E
A
A
C
C
C
B
C
C
A
A
Cleanup,
okay,
then,
thank
you
guys
see
you
in
two
weeks
well,
I
said:
actually
two
week
will
be
kicked
on
right,
then
we
need
to
decide.
You
are
going
to
call
the
meeting
or
not
yeah.
It
will
be
a
sitcom,
so
we
might
not
meet
into
exclusive
anyway,
we'll
see
we'll
see
each
other
life
much
better.
All
right!
Thank
you.
Everyone
so
see
you
whatever
the
next.
Whenever
the
next
meeting
will
be
will
be
fun.
Bye
answer.