►
Description
Charo Gruver and Bruce Link demonstrate OKD single node cluster deployments.
A
A
I
have
several
tutorials
that
I've
worked
up
and
a
few
more
that
are
coming
this
one
right
here
that
I
have
pinned
to
the
top
right
is
an
installation
for
a
single
node
cluster.
Now
what
I'm
going
to
show
you
guys
here
is
going
to
be
running
on
a
nook,
an
intel,
nook,
8i3
bek.
It's
got
32
gigabytes
of
ram
in
it.
A
So
so
this
little
guy
right
here
is
not
a
high-powered
piece
of
equipment
by
any
means
and
they're
actually
pretty
affordable.
This
one
is
getting
as
computing
goes.
It's
getting
a
little
long
in
the
tooth.
The
newer
i3
fnk
will
actually
allow
you
to
put
64
gigabytes
of
ram
in
it,
has
faster
cpus
and
and
if
you
really
want
to
spend
a
little
bit
more
of
the
i7-
and
I
should
have
brought
one
of
them
down
here,
so
I
could
hold
it
up
to
the
camera.
A
A
A
Yeah
there
you
go
bruce
is
holding
one
up
for
you.
If
you
can
see
it
on
the
on
the
image
there,
they
are
they're,
tiny,
little
things
and
I'm
ashamed
to
admit
it,
and
it's
probably
embarrassing
to
my
wife,
but
I
actually
travel
with
four
of
those
when
we
go
on
vacation,
because
I
like
to
take
my
openshift
cluster
with
me,
because
it
keeps
me
happy
so
without
further
ado,
let's
deploy
a
single
open
cluster.
A
I'm
going
to
move
this
link
back
over
here
out
of
the
way
I'm
going
to
minimize
this
browser
because
we
won't
need
it
for
a
while.
We
are
going
to
be
working
from
the
command
line
for
a
bit
okay.
So
if
you,
if
you
look
at
the
github
pages,
the
first
section
is
installing
and
setting
up
your
host
now
in
in
my
lab,
I'm
actually
running
a
a
pixie
environment
with
with
ipixi
that
I
boot
and
install
my
machines
from
so
I'm
skipping
that
point
for
you.
A
It
does
need
to
be
an
rpm
based
if
you're,
if
you're
following
this
tutorial,
because
I
make
some
assumptions
about
the
the
commands
that
are
available
for
managing
the
machine.
So
the
first
thing
I'm
going
to
do
is
install
my
vert
environment
and
you'll
see.
This
is
not
going
to
do
anything
because
my
pixie
boot
already
knew
that
this
was
going
to
be
a
kvm
host.
So
it
did
that
for
me.
So
the
next
thing
I'm
going
to
do
is
install
all
of
the
tooling
that
we're
going
to
need
for
this
lab.
A
A
C
C
C
A
C
A
A
Okay,
so
we're
involving
installing
the
epel
release,
because
we're
going
to
need
some
of
the
libraries
from
there
libvert
development,
the
httpd
tools
that
are
going
to
give
us
some
of
the
things
that
might
some
of
the
tooling
that
my
scripts
use
and
finally
nginx
and
all
nginx
is
going
to
do
here-
is
host
the
ignition
configs
for
a
newly
booted
machine
to
grab
its
ignition
from.
So
I'm
going
to
go
ahead
and
let
those
install.
A
And
now
what
I'm
going
to
do
is
because
I
just
like
to
control
things,
I'm
going
to
tell
libert
where
to
store
its
virtual
machines
when
I
create
them,
so
I'm
creating
I'm
deleting
the
default
pool
if
it
exists.
Now
you
can
see
here
in
this
case
there
was
no
default
pool
and
then
I'm
going
to
create
this
default
pool
and
I'm
going
to
point
it
at
slash
virtual
machines.
So
all
of
the
vms
that
I
build
with
libvert
they're
going
to
go
into
that
location.
A
And
finally,
I'm
going
to
configure
the
firewall
all
right.
So
I'm
enabling
http
and
https
from
the
firewall
I'm
enabling
dns
because
we
are
running
bind
and
the
rest
of
the
firewall
I
will
leave
in
place
next
step.
Is
we're
going
to
need
an
ssh
key,
so
we
can
get
to
our
openshift
cluster
if
we
need
to
directly
connect
to
any
of
the
nodes.
A
So
I'm
going
to
create
myself
an
ed25519
key
pair,
the
next
part
of
the
tutorial
here,
I'm
not
going
to
actually
execute
I'll
put
it
on
the
screen
here
for
just
a
minute
so
that
you
guys
can
see
it.
This
is
creating
the
network
bridge
and
what
I've
included
for
you
here
is
the
command
line
way
using
nmcli
to
create
your
network
bridge
now
there
are.
There
are
graphical
ways
to
do
this
as
well.
You
can
even
do
it
during
the
install.
A
A
B
A
A
A
There
are
several
environment
variables
that
you're
going
to
set
that
I've
provided
an
out
of
the
box
in
in
the
set
snc
env
script,
but
you
will
probably
want
to
modify
them
for
your
own
environment.
The
10.11.11.0
is
what
I'm
currently
using
in
my
home
lab
network,
so
we'll
set
a
a
domain
that
will
be
our
dns
domain
for
this
lab.
A
This
snc
host
is
going
to
be
this
server.
That
is
rebooting
right
now.
The
name
server
actually
is
going
to
be
the
same
one,
but
I've
allowed
by
using
variables
for
you
to
use
different
hosts.
So
you
can
create
your
own
name
server
host.
If
you
want
to
you,
can
have
your
own
bastion
server.
You
can
run
the
single
node
cluster
on
its
own
machine.
A
A
A
A
A
A
To
modify
this
script
here,
to
ensure
that
everything
is
set
up
for
your
particular
network,
I'm
not
going
to
change
anything
here.
I
believe,
because
I'm
going
to
use
the
domain
snc
test
206
is
this
host
that
we're
currently
on.
I
am
using
a
slash,
24
network,
so
I
don't
need
to
change
either
of
those.
A
A
I'll
put
this
in
the
chat
for
you
guys
there
you
go!
So
if
you
go
there,
you
will
see
the
current
state
of
the
given
okd
releases.
The
very
top
of
the
list
is
the
stable
channel
right
and
those
are
in
quay.
Dot
io
under
open
shift,
slash
okd.
So
if
I'm
going
to
build
one
of
these,
which
is
what
we're
going
to
do
today,
I
I
would
build
from
this
channel,
but
this
will
also
build
nightlys.
A
But
if
I
want
to
build
from
a
4.7
nightly,
which
I
have
done
actually
recently
as
I
was
looking
to
see
tracking
particular
bugs
as
we
get
them
wiped
out,
then
you
can
come
down
here
to
the
nightlys
or
you
can
even
under
the
stable,
build
older
versions
of
4.6
or
4.5.
Okay.
So
today
we're
going
to
build
a
stable
release
from
from
4.7
and
it
will
come
from
quay
dot
io.
If
I
did
want
to
build
a
nightly,
then
I
would
change.
A
Excuse
me,
this
okd
registry
environment
variable
to
registry.ci.openshift.org
origin,
slash,
release,
okay,
so
that's
really
the
only
difference
there
and-
and
I'm
realizing-
that's
probably
really
small
font
for
you
guys,
so
there
I'll
crank
that
up
a
little
bit
all
right.
So
that's
where
you
go
to
find
the
the
current
releases
and
what
their
disposition
is.
A
A
The
first
thing
I'm
going
to
do
is
I'm
going
to
enable
name
d
so
that
it
can
start
and
then
I'm
going
to
run
actually
before
I
do
that
now,
I'm
going
to
run
it
first
and
then
I'll
show
you
what
the
the
magic
behind
the
covers.
So
I'm
going
to
run
my
setup
dns
and
what
it
did
leveraging.
Those
environment
variables
is
it
set
up
a
name
server.
A
A
I
did
create
an
acl
that
includes
my
anything.
That's
on
this
24
network
right
under
my
10.11.11.0
network
and
my
localhost.
Now,
if
you're
also
running
a
podman
or
docker
on
this,
you
will
need
to
add
the
ip
address
for
that
network
as
well,
so
that
any
of
the
containers
that
you
run
can
do
name
resolution
that
has
bit
me
before
in
the
past,
when
I
forgot
to
do
that.
A
A
And
it
created
the
a
records
for
all
of
the
hosts
as
well.
Now
I
am
going
to
spend
just
a
couple
of
minutes
talking
about
a
couple
of
things
here:
one
I'm
not
setting
up
a
load
balancer,
because
this
is
a
single
node
cluster
right.
So
I
don't
need
to
balance
across
different
ingress
nodes.
I
don't
have
infrastructure
nodes
in
my
other
tutorial,
you
would
set
up
a
load
balancer
to
manage
your
traffic,
but
here
it
was
complexity.
We
didn't
need.
A
However,
during
the
bootstrap
you
have
to
be
able
to
talk
to
either
the
bootstrap
node
or
then
the
master
node,
and
they
have
to
both
be
able
to
resolve.
So
what
I'm
doing
here
is
a
little
dns
trick
by
having
the
same
address
with
the
two
different
ip
addresses
in
there.
The
dns
round
robin
will
give
me
a
poor
man's
load
balancer.
A
These
two
records
that
have
the
remove
after
bootstrap
annotation
on
them
will
be
pulled
out
of
this
when
we
destroy
the
bootstrap
node,
leaving
just
the
master
node
in
place
and
the
the
master
node
has
multiple
records.
We
have
the
actual
host
for
the
master
node.
We
have
the
etcd
node
and
then
we
have
a
wild
card
for
our
cluster
and
the
api
endpoint
for
our
cluster
and
the
api
internal
for
the
cluster.
So
all
of
those
you
see
they
all
resolve
to
the
same
address
they're
all
the
master
node.
A
A
All
right
so
pretty
simple,
it's
it's
said
magic
in
the
repository
I've
got
stub
files
for
the
named.com,
the
zone
files
and
the
named.conf.local
and
those
those
stub
files.
There
are
records
in
them
that
look
like
this.
That
gets
substituted
with
some
said
magic
and
when
these
files
get
put
in
place
and
then
the
last
thing
it
does
is
refresh
name
d
so
that
you
get
a
clean
start.
A
One
word
of
caution:
don't
do
this
on
a
machine
that
you've
already
configured
name
d,
because
this
will
destroy
your
name
deconfiguration,
okay,
so
this
is.
This
is
set
up
to
to
be
a
clean,
install
you're,
not
using
this
particular
machine
for
anything
else,
except
a
single
node
cluster.
Then
this
is
safe.
A
A
A
I've
got
this
conveniently
set
aside,
so
what
I'm
doing
is
I'm
doing
an
nmcli
connect
against
my
bridge
connection
and
setting
its
name
server
to
the
environment
variable
that
is
set.
This
is
why
I
use
that
that
set
enb
in
my
bash
rc
is
because
then
I
don't
have
to
remember
numbers
and
things.
I
know
that
the
variables
are
set
from
a
consistent
thing,
so
it's
going
to
put
the
name
server
and
the
domain
in
there.
A
A
A
A
That
also
works,
if
you're
running
code
ready
containers
on
your
local
machine
as
well
so
helpful
tip
there
all
right.
The
next
thing
is
to
prepare
to
install
our
cluster
and
the
utility
scripts
that
I've
put
in
place
are
hopefully
going
to
make
this
pretty
straightforward
for
you
all
right.
So
first
thing
we're
going
to
do
is
get
this
to
go
to
the
top.
There
we
go,
so
you
guys
can
see
it.
I'm
gonna
set
that
okd
release
variable
that
I
was
showing
you
guys
previously.
A
A
And
then
we
have
to
prepare
our
installation
file
the
manifest
and
I've
got
a
stub
of
that
for
you
here
as
well,
so
you
can
see
what's
going
to
happen,
your
base
domain
is
going
to
again
with
some
said
magic.
Your
base
domain
is
going
to
get
populated
the
cluster
network.
These
defaults
should
work
for
you
out
of
the
box.
A
If,
on
your
home
network,
you
have
a
10.100
or
a
172.30,
you
might
need
to
change
either
the
cider
for
the
cluster
network
or
the
service
network,
but
since
most
people
are
sitting
on
a
10.10
or
a
192.168,
these
should
work
for.
For
most
of
you,
the
compute
section,
you
want
zero
replicas
under
the
worker
section,
and
this
is
what
makes
it
a
single
node
cluster.
A
I'm
telling
it
to
do
one
master
because
we're
doing
a
bare
metal
install
you,
don't
you
put
none
under
the
platform
since
we're
not
doing
aws
or
vsphere
or
azure
right.
If
you're
doing
this
on
one
of
those
platforms,
then
you
can
take
advantage
of
their
capabilities
and
you
would
put
your
settings
here.
This
pull
secret
is
just
a
fake
pull
secret.
That
is
base64
encoded,
fubar
right
there,
and
this
is
going
to
get
replaced
with
the
ssh
key
that
we
created
earlier.
A
A
So
now
you
can
see
I've
got
that
install
config
in
my
working
directory
and
I'm
going
to
run
this
set
command
against
it.
To
put
the
domain
in
place
like
that,
then
I'm
going
to
populate
a
environment
variable
with
my
ssh
key
that
we
created
earlier
with
the
public
key.
Don't
put
your
private
key
in
here.
A
A
Which
I
am
and
I'm
going
to
install
the
single
node
cluster
now
before
I
kick
this
off,
I'm
going
to
go
ahead
and
kick
it
off
and
then,
while
the
install
is
going
I'll,
pull
up
the
code
and
and
show
you
guys
what
is
actually
going
on
under
the
covers
in
this,
because
I've
abstracted
away
building
the
virtual
machines
and
configuration
and
a
whole
bunch
of
stuff.
So
so
I'm
going
to
show
you
guys
the
the
code
that
I
wrote.
A
A
A
A
Now
it
is
at
this
stage-
and
I
say
this
in
the
in
the
tutorial
documentation-
it
is
not
critical
that
you
have
the
latest
or
the
the
version
of
oc
that
you're
going
to
be
installing
from
it
just
needs
to
be
a
recent
4.x
version
of
oc,
because
the
the
script
that
we
run
to
build
this
is
actually
going
to
retrieve
the
the
real
version
of
oc
alright.
So
I
just
pulled
that
from
the
page.
A
A
A
A
A
Once
the
boot
strapping
begins,
it's
actually
going
to
replace
the
operating
system
with
the
version
of
fedora
core
os
that
is
bundled
with
this
version
of
openshift,
so
that
everything
is,
is
controlled
and
managed
by
the
openshift
cluster
itself,
and
that's
actually
one
of
the
things
I
didn't
mention
this
morning.
That
is
somewhat
unique
about
openshift
versus
other
kubernetes
distributions.
A
A
I'll
show
I'll
show
you
here
in
a
minute
it's
actually
embedded
in
the
in
the
binary
or
not
the
binary,
but
the
script
that
I
ran.
So
when
I
pop
open
the
script
I'll
show
you
there,
because
you
can
override
that
it
again,
it
doesn't
have
to
be
the
specific
version
of
fedora
core
s
that
is
needed
for
for
that
install
of
openshift.
It
just
needs
to
be
a
fairly
recent
one.
B
A
All
right
now,
because
my
script
disconnected
from
those
when
it
kicked
off
the
initial
install
they
did,
they
did
not
actually
reboot
themselves.
They
just
shut
themselves
down.
So
so
the
next
thing
I
need
to
do
is
to
boot
these
up.
But
when
I
do,
they
are
going
to
begin
the
single
node
cluster
install.
A
A
A
A
And
I'll
start
the
master
node
and
I'm
actually
going
to
switch,
and
let
you
see
the
master,
because
I
want
to
point
out
something
that
it's
going
to
be
doing
all
right.
You
see
it
actually
didn't
really
do
anything.
You
see
it's
pending
on
this
start
job
here
it
is
waiting
to
get
its
ignition
config
from
the
bootstrap
node.
A
E
A
See
the
rpm
os
tree
stuff,
that's
going
on
what
it's
doing
here
is
it's
downloading
the
correct
version
of
fedora
core
os
that
this
cluster
needs
to
run
and
it's
going
to
apply
it
and
there
it
just
finished.
So
you
saw
this
guy
kicked
me
out,
because
the
bootstrap
is
now
rebooting,
so
I'll
go
ahead
and
pin
that
there.
Let
me
separate
this
one
out
shrink
it
down
a
bit.
A
Everything
in
here
is
containers
we're
rendering
the
manifests
and
in
just
a
little
bit
this
api
is
going
to
come
up,
and
here
you
will
see
the
masternode
node
finally
get
a
hold
of
its
ignition
file
and
start
its
install,
see
there
class
starting
cluster
bootstrap
and
there
you
go
so.
The
master
node
now
has
its
ignition,
and
now
it
is
booting
itself
up.
It's
going
to
do
a
very
similar
thing:
it's
going
to
pull
its
core
os
image
and
apply
it
and
then
start
its.
A
A
All
right
now,
if
you
watch
these
logs
like
this,
don't
don't
panic
when
you
see
errors
and
things
flying
by
like
I
saw
all
this
failed
to
fetch
failed
to
fetch
failed
to
fetch.
This
is
doing
a
whole
bunch
of
stuff
in
parallel,
and
so
there
will
it,
it
waits
until
states
come
into
being
so
it
can
do
things.
So
during
this
time
you
will
see
stuff
like
this.
If
you
don't
see
progress
down
here
like
you
can
see,
the
api
is
up
and
running
now.
A
A
Upset
so
I'm
going
to
export
my
cube
config
variable
to
where
my
install
is
so
that
I
can
issue
oc
commands
and
I'm
going
to
see
if
the
sed
object
exists.
Yet,
since
I
was
talking
to
you
guys
yeah
it
does.
If
we
had
done
this
40
seconds
ago,
you
would
have
seen
an
error.
I
call
that
out
in
my
tutorial,
because
you
do
need
to
wait
a
little
bit
for
the
scd
object
to
show
up
now
that
it's
there
we're
going
to
patch
its
specification.
A
A
You
can
see
now
under
the
spec
there's
this
unsupported
config
overrides
where
non-ha
is
set
to
true.
So
that's
that
is
one
thing
that
in
4.8,
because
single
node
will
be
an
officially
supported
configuration
in
4.8.
We
won't
have
to
do
that
anymore
right
now.
We
have
to
do
that
because
it's
still
not
a
fully
supported
configuration,
but
it
does
work
right
so
now
we're
waiting
for
the
bootstrapping
to
complete.
A
A
A
A
A
A
To
the
actual
start
of
the
script,
so
one
thing
I'm
doing
I'm:
this
is
a
this
is
a
cheesy
little
random
generator
here.
So
I
apologize
to
everybody
up
front,
but
I'm
generating
mac
addresses
for
my
bootstrap
and
my
master
node,
so
I'm
generating
those
mac
addresses
and
and
the
reason
I'm
doing
that
it
really
has
nothing
to
do
with
openshift.
A
So
so
then
I
use
a.
I
use
a
dig
to
get
the
ip
addresses
that
are
already
in
dns
for
the
bootstrap
and
the
master
node,
and
I
populate
variables
with
those
here's
where
I'm
pulling
the
okd
tooling
for
the
openshift
installer
and
for
the
oc
command
and
I'm
putting
those
into
my
home
bin
directory
and
then
getting
rid
of
that
that
temporary
folder
that
I
created
here,
I'm
grabbing
a
tool
called
fcct.
A
This
is
a
fedora
core
os
tool
that
allows
you
to
manipulate
ignition
files
for
machines
and
I'll
show
you
the
configuration
further
up
in
this
script,
for
what
I'm,
what
I'm
using
fcct
for
that
enables
you,
especially
if
you're
doing
bare
metal
cluster
work
that
enables
you
to
configure
what
the
operating
system
is
going
to
look
like.
Let's
say,
you're
deploying
some
nodes
and
you
want
you
know
ssd
devices
that
are
in
them
to
be
preserved
for
rook
ceph.
A
A
The
next
things
here
from
this
syslinux
on
down,
I'm
actually
preparing
an
iso
for
this
machine
to
boot.
From
again
because
I
didn't
want
to
create
a
pixie
environment,
I
I'm
not
using
vsphere.
This
is
all
bare
metal,
so
I'm
creating
an
iso
for
the
the
bootstrap
node
and
for
the
master
node
that
have
the
specific
configuration
in
them
that
they
need
just
to
come
up
and
then
start
being
part
of
the
cluster.
A
And
the
last
thing
here
after
creating
those
those
isos
is
then
to
create
the
virtual
machines
themselves
and
the
creation
of
the
virtual
machines
themselves
is
using
the
vert
install
and
passing
the
appropriate
parameters,
including
the
particular
iso
file
that
I
just
created
all
right.
So
the
last
thing
I'm
going
to
show
you
in
here.
Let's
check
on
our
our
bootstrap
bootstrap,
is
still
running,
make
sure
nothing's
blowing
up
over.
D
A
What
what
the
fcct
tool
works
with
is
yaml
files,
an
ignition
config
and
I'm
setting
this
up
as
a
merge
type.
So
it's
going
to
merge
the
the
following
configuration
into
the
ignition
config.
That
then
becomes
the
initial
ignition
config
for
this
machine
to
boot.
Off
of
this
is
why
I
had
to
know
the
mac
address,
because
I
am
explicitly
hard
coding,
the
name
of
that
nick
to
be
nick
0,
so
that
it's
always
predictable,
and
then
I'm
configuring
network
manager
with
the
network
information
about
that
machine.
A
So
it's
fairly
short,
but
it
gives
you
a
clue
into
the
power
of
ignition
configs
and
what
you
can
do
with
them.
So
I
am
prescriptively
setting
up
the
network
configuration
for
this
guy
before
he
boots
up
and
that's
the
deploy
okd
snc
script.
It
does
all
of
that
for
you
and
looks
like
the
bootstrap
is
now
done.
A
A
A
A
A
B
A
Back
off
is
not
good,
but
I
don't
see
any
of
those,
so
everything
appears
to
be
fat
and
happy
right
now.
So
a
couple
of
things
that
we
need
that
we
need
to
do.
The
ingress
operator
is
going
to
be
unhappy
because
it
expects
to
have
two
replicas
and
the
authentication
operator
is
also
going
to
be
unhappy
because
it
is
expecting
three
etcd
nodes.
So
we
have
to
tell
those
two
that
we're
running
in
a
single
node
configuration.
A
A
A
A
A
A
A
A
A
A
A
B
A
A
A
A
A
A
A
A
A
B
A
A
A
A
A
D
A
I
have
seen
that
a
couple
of
times
in
my
lab
and
I'm
wondering
if
it's
a
race
condition
on
when
I
yank
the
bootstrap
node
out
that
etsyd
goes
sideways
somehow,
because
I
I
have
it's
very
rare
that
I've
seen
this
and
of
course
it
would
happen
on
the
day
that
we're
we're
doing
a
video.
But
I
have
actually
seen
that
happen
with
a
full
cluster
build
a
couple
of
times,
but
not
enough
time
to
warrant
actually
getting
in
to
see.
If
I
can
figure
out
why
it's
doing
that,
but.
E
A
A
A
All
right
so
in
my
environment,
I
should
probably
be
using
ansible
for
some
of
this,
but
I
know
shell,
so
I
write
shell.
So
in
my
environment,
when
I,
when
I
get
a
new
nook
and
I'm
adding
it
to
my
lab,
I
have
a
utility
script
here,
deploy
kvm
host
that
all
I
have
to
give.
It
is
the
host
name
that
that
new
nook
is
going
to
have
the
mac
address
off
the
butt
of
that
nook.
A
A
A
A
It
does
a
dig,
just
like
you
saw
in
the
in
the
single
node
cluster
one
to
get
the
ip
address
that
I've
already
pre-populated
in
dns
creates
a
file
that
is
named
the
mac
address,
with
the
colons
replaced
with
dashes
and
a
dot.
I
pixie
and
creates
an
ipixie
file
for
that
machine
so
that
it
will
boot
from
this
kernel
and
this
initial
ram
disk
off
of
my
nginx
server
and
it
pre-populates
the
ip
information
that
I
want
this
machine
to
have.
A
All
right,
okay,
so
now
bootstrapping
is
still
running.
Okay
and
then
I
create
a
kickstart
file
for
that
machine
and
the
the
kickstart
file
then
has
all
of
the
configuration
that
I
want
it
to
have.
And
here's
where
I
feed
in
the
disk
information
it's
going
to
create
some
of
the
file
systems
with
prescriptive
sizes
and
then
just
blow
the
rest
of
them
out
to
whatever
the
the
size
of
the
disk
is.
A
A
A
You
incur
down
time
when
you
do
an
upgrade
yeah,
so
I'm
using
kickstart
in
my
lab
and
it's
it
sets
up
everything
that
I
want
this
machine
to
be
doing.
So,
since
this
is
a
kvm
host.
It's
you
know
it's
installing
the
libvert
tooling,
that
I
need,
and
then
it
reboots
the
machine
and
when
the
machine
comes
up,
it's
got
its
full
personality
and
it's
ready
to
be
part
of
my
lab.
A
A
I
have
not
tried.
I
have
not
tried
this
on
on
my
mac.
Actually
code
ready
containers
will
will
run
on
your
mac
and
once
we
once
we're
waiting
on
a
a
bug
fix
for
the
bare
metal
installer
that
actually
the
dean
will
probably
drop
the
new
release
today
and
if
that's
working,
then
I'm
going
to
build
the
4.7
version
of
code
ready
containers
for
okd
the
version
that's
currently
out.
A
A
All
right
things
are
creating
it's.
Okay,
there
will
be
errors.
If
you
look
at
this
before
bootstrapping
is
complete.
You
will
see
containers
in
error
states
again
just
like
when
we
were
watching
the
log
scroll
most
of
the
time
that
is
okay.
It
just
means
that
there's
some
collaborator
resource
that
this
particular
operator
is
waiting
on
that
isn't
there
yet,
and
so
it
continues
to
error
until
its
collaborator
is
in
place
and
then
it
will,
it
will
finish
its
installation.
A
A
And
one
thing
once
once
you
get
this
up:
there
are
some
of
the
operators
from
the
marketplace
that
might
be
assuming
a
a
fuller
cluster.
I
know
like
rook
seth
out
of
the
box,
you
if
you
want
to
run
rook
ceph
on
a
on
a
single
node
you'll
have
to
adjust
the
replica
count
for
some
of
the
components
because
it
will
expect.
You
know
at
least
three
nodes
for
particular
things:
code,
ready
containers,
yes
and
no
patrick
code.
D
A
A
A
A
Okay,
single
node
cluster
from
code
ready
containers
it
it's
actually
doing.
It
is
useful
to
give
you
some
insights
into
something.
It's
it's
actually
doing.
An
ipi
libert
install
of
your
openshift
cluster,
which
is,
is
pretty
cool
in
and
of
itself
in
a
similar
fashion.
To
what
I'm
doing
here,
in
fact,
let
me
see
I
was
hacking
on
code
ready
containers.
A
B
A
A
All
right
crank
the
font
up,
for
you
guys,
apologies
that
that
makes
the
screen
really
crowded.
So
what
the
punch
line
in
this
it
it
is
very
prescriptive
toward
building
code
ready
containers.
So
it
strips
a
bunch
of
things
out
of
the
openshift
cluster
like
monitoring
and
things
like
that
to
make
it
fit
in
a
really
compact
size
or
compact
for
open
shift.
A
A
Using
the
bare
metal
to
then,
let's
write
in
here's
some,
oh
here
we
go
creating
the
cluster,
so
this
is
using
the
I
it's
using
an
ipi
install
where
the
installer
actually
does
all
of
the
provisioning
and
stuff
for
you.
So
not
you
know
not
giving
you
control
over
the
network
configuration
or
things
like
that.
This
is
more
like
the
I'm
deploying
in
the
cloud
on
aws
or
on
gcp
or
or
on
on
vmware,
or
you
know
something
along
those
lines.
So
that's
what
that's
what
the
code
ready
containers
is
built
from.
A
So
this
big
long,
snc.sh
script
here,
what
it's
actually
doing
is
standing
up
a
single
node
cluster
on
libert,
and
then
you
run
another
script.
That's
part
of
this
bundle
called
where's
it
at
create
disk
words
create
disk
right.
There
called
create
disk
that
then
creates
a
qcal's
image
of
the
the
cluster
it
strips
a
whole
bunch
of
operators.
Out
to
you
know,
like
I
said,
to
try
to
get
the
size
down
as
much
as
possible
and
once
it's
once
it's
done.
A
So
so,
for
each
of
these
there's
functions
further
up
here
you
see
it
it's
creating
the
q
q
mu
image.
It
creates
then,
from
this
images,
for
windows,
for
linux
and
for
mac,
os
right
and
that's
basic
and
and
when
you're
done
with
that,
then
you
have
a
a
file,
a
virtual
machine
file,
that
is
a
single
node
cluster
that
would
run
on
one
of
those
platforms
code.
A
Ready
containers
is
then
a
wrapper
around
that
that
bundles
the
the
image
so
that
you
can
start
and
stop
it
with
the
crc
and
that's
why
the
crc
binary
is
so
fat.
If
you
go
download
it
you'll
see
that,
like
the
the
mac
os
crc
binary
is
like
two
and
a
quarter
gigabytes.
I
think
most
of
that
is
the
virtual
machine
image,
because
it's
when
it
uncompresses
it
uncompresses
to
about
a
910
gigabyte
image.
A
A
A
A
A
A
A
Oc
edit
and
I
could
go
down
here
and
change
that
replicas
to
one
and
save
it,
but
if
you
really
want
to
impress
your
friends,
the
oc
patch
command
looks
like
magic
because
from
the
command
line
without
doing
anything
else-
and
this
is
also
how
you
can
do
this
kind
of
thing-
with
infrastructure
as
code.
So
what
I'm
going
to
do?
This
is
a
json
patch
and
what
I'm
saying
is
under
the
under
spec.
A
I
want
replicas
to
be
one
and
I'm
doing
a
patch
type
of
merge,
rather
than
replace
so
I'm
gonna,
I'm
gonna,
merge
this
into
the
existing
config
and
and
I'm
going
to
patch
the
ingress
controller
named
default
in
the
openshift
ingress
operator,
namespace
with
that
patch
and
there
it
is
now
it's
patched,
and
if
I
do
this
dasho
yaml
again
and
scrolling
scrolling
up,
you
can
see
that
now.
Replicas
is
one
okay!
A
A
Let's
go
ahead
and
go
get
the
authentication
okay,
you
see
this
ok
d4
installed
there
that
was
created
by
my
script
and
that's
where
it
put
the
manifests
and
things
and
then
told
the
installer
where
to
go
to
get
its
manifest.
So
that's
all
of
this
stuff
under
here
was
created
by
the
installer.
So
so
there's
the
there's,
the
ignitions
that
were
created.
You
see
this
guy
right
here,
the
worker
ignition.
A
A
Before
I
step
away
from
this,
though
you
see
this
worker
ignition,
if
you
were
in
the
main
stage,
when
we
were
first
starting-
and
we
were
talking
about-
you
know
the
fact
that
you
can't
go
from
a
single
node
cluster
to
a
full
cluster.
That's
really
just
talking
about
the
control
plane.
We
could
use
this
ignition
file,
this
worker
ignition
file.
I
could
build
a
virtual
machine
boot
that
virtual
machine
off
of
this
worker
ignition
file
and
it
would
join
the
cluster
as
a
worker.
A
I
would
not
have
an
h
a
sed
control
plane
because
I
would
be
using
an
unsupported,
unsafe,
non-ha,
non-production,
unstable
betsy
d,
but
I
can
create
lots
of
worker
nodes
now
and
and
have
a
single
master
cluster
with
with
worker
nodes.
So
since
our
cluster
is
up,
let's
go
ahead
and
log
into
it.
You'll
notice
down
here
it.
It
gave
me
a
url
and
it
gave
me
a
username
and
a
password.
A
If
you
lose
those
you
can
get
them
back.
What
I
was
going
to
show
you
is
under
this
auth
directory
here.
This
cube
config
is
what
I've
been
using
when
you've
seen
me.
Do
this
that
cube,
config
is
kind
of
like
the
you
know,
passwordless
ssh,
so
it
gives
me
immediate
access
to
it.
That's
like
my
door!
Key!
A
Don't
share
that
unless
you
want
to
give
people
cluster
admin
access
to
your
cluster,
and
if
you
delete
it,
you
can't
get
it
back,
but
it's
probably
a
good
idea
to
either
keep
it
in
a
super
secret,
safe
place
or
delete
it
and
use
the
the
identity
management.
That's
there.
The
other
thing
that
it
gives
you
is
this
cube
admin
password,
which
is
a
file
that
contains
this
randomly
generated,
string
right
there,
that
is
the
password.
A
A
B
A
A
Let
me
show
you
in
one
of
my
larger
lab
clusters.
You
see
that's
not
there,
because
all
of
the
pods
are
in
a
good
state.
You
will
always
see
this
in
the
4.7
open
shift
cluster,
because
there
are
two
etsy
decorum
guard
pods
that
don't
have
anywhere
to
run
so
nothing
to
worry
about
here.
It's
not
broken.
It's
just.
You
will
see
that
because
these
two
etcd
quorum
guard
pods,
don't
have
a
node
to
run
on
when
we
get
the
full
bootstrap
list.
4.8.
A
You
shouldn't
see
that
anymore.
But
at
this
point
now
the
cluster
is
ready
to
run.
There's
actually
a
couple
of
other
things
that
I
included
in
the
tutorial.
That
will
be
helpful
for
you.
One
of
them
is,
let's
go
ahead
and
set
up
a
real
user
account
and
a
developer
user
account
all
right
and-
and
I
did
include
in
the
in
the
tutorial-
and
let
me
get
back
to
get
back
to
that.
A
A
Within
the
cluster
itself,
and
what
we're
gonna
do
right
now
is:
let's
create
I'm
just
going
to
pull
these
straight
out
of
the
out
of
the
tutorial,
so
I'm
going
to
create
a
working
directory
for
the
credentials
and
locally
on
my
box.
I'm
going
to
this
was
actually
when
you
saw
initially,
we
were
installing
the
http
tools
that
was
to
get
to
the
hd
password
command
so
that
I
can
create
hd
password
files.
A
A
A
A
And
it
has
that
the
content
of
that
hd
password
that
we
just
created
all
right.
So
that's
step
one.
Now
it's
in
the
openshift
config
step
two
is
to
apply
that
custom
resource
right
there,
the
hd
password
custom
resource-
and
it
will
do
this
because
I'm
using
an
oc
apply
and
not
an
oc,
create
I
just
I'm
in
a
habit
of
always
using
oc
ply.
D
A
A
A
A
A
A
A
D
A
Two
users,
a
dev
user
and
a
cluster
admin,
a
couple
of
other
things
just
to
keep
it
from
complaining
at
you.
If
you're
going
to
use
the
internal
image
registry,
you
either
need
to
give
it
a
host
volume
or
an
ephemeral
volume
in
the
tutorial,
I've
got
the
command
for
you
to
give
it
a
ephemeral
volume.
A
So
I'm
going
to
switch
the
image
registry
to
a
managed
state
and
I'm
going
to
give
it
an
empty
dir,
which
is
an
ephemeral
on
volume.
But
since
you
only
have
one
node
that
ephemeral
volume
is
really
sitting
on
the
file
system
of
that
one
node.
So
the
images
that
you
put
there
that
they're
living
on
that
one
node.
So
you
can
use
the
internal
image
registry
at
this
point
now
and
you
saw
in
the
background
there,
it's
actually
doing
some
things
because
of
that
you
see
the
image
registry
kicked
up.
A
And
that
one
is
shutting
down
and
replaced
itself.
So
now
we've
got
a
configured
image
registry
and
the
last
thing
is
image
pruners.
It
will
complain
if
you
don't
have
an
image.
Printer
scheduled
you'll,
get
warnings
in
your
console,
so
let's
just
go
ahead
and
set
up
a
image
pruner
that
will
prune
out
images
based
on
this
configuration
at.
A
B
A
Yeah,
I
know
you
know
I
need
to
run
enough
in
stalls.
I
might
kick
off
a
loop
that
just
spins
them
and
shuts
them
down
to
see
if
I
can
catch
what
it
is.
That
causes
etsy
d
to
fall
over
because,
like
I
said,
I've
seen
that
just
a
few
times-
and
I
literally
ran
this
at
least
three
times
this
morning
and
four
or
five
times
yesterday.
While
I
was
cleaning
yeah.
A
B
D
A
That
that's
right!
So
what
I've
generally
done!
I
I've
I
have
found
that
ht
password
is
pretty
stable,
even
across
upgrades
at
a
previous
place,
where,
where
I
was
doing
this
for
real
in
the
data
center,
what
I
set
up
for
them,
because
they
they
had
azure
ad
as
their
authentication
mechanism,
so
but
but
if
something
went
either
wrong
with
that
or
with
the
cluster's
integration
to
azure
ad
that
they
would
lose
access
to
their
cluster.
A
So
we
set
up
ht
password
for
just
a
few
of
the
cluster
admins
and
that
was
kind
of
the
always
there
back
door,
and
then
you
just
set
up
a
you
know
a
password
rotation
and
complexity
policy.
That
periodically
replaces
that
secret
and
that
way
you've
at
least
got
a
way
to
get
into
your
cluster.
If
you,
you
lose
your
primary
oauth
provider,
I've
also
set
it
up.
A
Yes,
yes,
and
and
mike
you're,
absolutely
right.
In
fact,
you
can
do
that
with
code
ready
containers
too.
That's
the
way
to
make
code
ready
containers
available
off
the
workstation
that
you're
running
it
on
is
to
slap
h.a
proxy
or
your
favorite
proxy,
in
front
of
it
and
and
use
that
to
route
the
traffic
in
this.
In
this
case,
I'm
using
the
etsy
the
etsy
resolver
on
my
macbook
to
get
to
it.
In
my.
A
B
A
Yeah,
which
I
I
did,
I
bought
a
domain
name
because
I
plan
on
actually
one
of
these
days.
My
wife
is
encouraging
me
to
do
this,
and
I
just
never
get
around
to
it
is
to
start
right
is
to
start
keeping
an
official
blog
of
this
stuff.
So
I
did
buy
a
domain
upstream
without
a
paddle.
There's,
nothing
there
right
now.
A
If
you
go
to
it,
but
that's
me
upstream
without
a
pass,
so
I
actually
do
to
to
create
some
real
certs
and
do
this
with
with
real
certs,
so
that
I
can.
I
can
show
folks
how
that
works
as
well
I'll
get
around
to
it.
One
of
these
days,
one
of
the
other
things
that
I
I'm
not
sure.
If
I
I
would
have
to
make
any
endpoints
available
on
the
internet
to
make
that
work.
But
but
that's
another
thing
I
don't
do
is
make
any
of
the
endpoints
available
on
the
internet.