►
From YouTube: Ask an OpenShift Admin (Ep 38): High availability
Description
Most people think of Kubernetes and OpenShift as hosting “cloud native” applications, where cloud native refers to a system that expects failure and can compensate automatically, for example, using horizontal scaling. But does that apply to the infrastructure hosting OpenShift? What about OpenShift itself?
In this episode we’ll look at high availability for OpenShift, including what it expects from the infrastructure, what it’s capable of providing for applications, and some example scenarios.
Learn more:
https://cloud.redhat.com/blog/deploying-highly-available-applications-openshift-kubernetes
About Ask an OpenShift Admin:
Do you have questions? We want to answer them! Every Wednesday at 11am ET, hosts Andrew Sullivan and Chris Short sit down to answer your questions about Red Hat OpenShift.
Follow us:
Andrew Sullivan https://twitter.com/practicalandrew
Chris Short https://twitter.com/chrisshort
Christian Hernandez https://twitter.com/christianh814
Subscribe to Red Hat's YouTube channel: https://www.youtube.com/redhat/?sub_confirmation=1
About OpenShift:
Red Hat OpenShift is an open source container application platform based on the Kubernetes container orchestrator for enterprise application development and deployment.
Openshift.com https://red.ht/33Fh6uV
OpenShift on github https://github.com/openshift/
OpenShift on redhat.com https://red.ht/2o4ccDk
#RedHat #OpenShift #ContainerPlatform
Twitch: https://red.ht/twitch
YouTube: https://youtu.be/_YT3mB-jyzQ
A
A
A
Good
morning
good
afternoon
good
evening
and
welcome
to
a
very
special
ask
an
openshift
admin.
I
am
chris
short
host
of
most
things
on
red
hat
live
streaming.
I'm
joined
by
two
of
my
other
favorite
hosts
andrew
sullivan
and
christian
hernandez
you're
all
my
favorite
host,
if
you're
watching
out
there,
but
it's
like.
A
C
C
Yeah,
that
would
be
great,
so
yes,
hello,
everyone.
I
am
andrew
sullivan,
technical
marketing
manager
with
the
hybrid
platforms
business
unit.
Now
at
red
hat
and
yes,
it
is,
it
is
a
great
red
hat
day
as
every
day
is
a
great
red
hat
day,
so
I'm
very
privileged
to
be
here
not
only
with
chris
shorts,
the
the
producer
with
the
most
the
hostess
with
the
mostest,
but
also
my
other
other
favorite
co-host
christian
hernandez.
So
christian,
if
you
don't
mind
introducing
yourself
for
anybody
who
may
not
know
you.
B
Yeah,
what
can
I
say
about
myself?
You
guys
haven't
already
said
so:
yeah
also
principal
technical
marketing
here
senior
principal-
I
guess
now
and
yeah
here
happy
to
be
here
as
always,
andrew.
Thank
you
for
the
for
the
invitation.
I've
always
been.
I've
been
admitted
at
heart
for
most
of
my
most
of
my
life,
and
so
it's
it's
always
good
to
get
back
and
talk
about
admin
stuff.
B
C
So
and
there's
a
reason,
a
specific
reason
why
I
asked
you
to
come
and
join
me
on
this
show
and
that's
the
three
of
us
all
three
have
a
strong
ops
background.
You
know
we
we
all
three.
Were
you
know
customers
right.
We
were
all
three
administrators
of
rel
of
other
types
of
systems
at
some
point
in
our
careers,
and
I
I
think
that
that
gives
us
a
lot
of
background
and
can
give
us
a
lot
of
perspective
on
today's
topic,
and
this
is
one
that
we
we
love
to
talk
about.
C
Internally,
we
get
a
lot
of
questions
on
you
know
coming
from
our
fields
through
email
through
chat
through
all
of
the
different
forums.
We
also
hear
about
it
from
you
all
from
our
audience,
from
our
customers
so
on
and
so
forth,
and
that
is
in
case.
You
can't
see
the
title
of
the
stream
high
availability
and
we
wanted
to
take
this
opportunity
where
we
wanted
to
use
the
stream
to
talk
about
high
availability
from
a
couple
of
different
aspects.
C
So
we
want
to
talk
about
application
level,
high
availability
in
the
context
of
openshift
and
all
of
the
different
things
that
can
affect
that
things
that
can
go
into
it
so
on
and
so
forth,
and
we
really
want
to
have
this
be
an
open
conversation
right.
I
will
fully
admit
as
christian
and
chris
or
you
we're
on
a
lot
of
meetings
together.
C
You
all
hear
me
say
all
the
time.
I
don't
know
all
the
answers,
so
I
also
can
only
dream
up
so
many
use
cases
in
so
many
scenarios
right
in
addition
to
the
ones
that
I
personally
have
experienced.
So
I'm
very
curious
for
anybody
watching.
You
know
whether
you're
at
home
or
the
office,
where
you're
at
what
are
the
aha
things,
scenarios
questions
that
you
have
you
know
please
don't
hesitate
to
put
those
into
the
chat
here.
C
B
What
what
what
I
like
about
getting
different
admins
perspectives,
especially
having
you
guys
here
with
military
background,
is
the
different
use
cases
right
because,
like
you
said
andrew,
not
all
of
us
have
all
the
answers
right.
I
come
from
e-commerce
background
right,
startup,
sort
of
thing,
and
you
know
you
guys
have
to
worry
about
other
things
in
it
like
what,
if
a
tank
rolls
over
the
fiber
cable
like
like
that's.
B
A
C
I
I
think,
there's
also
a
lot
that
goes
on,
and
this
is
something
that
I
learned
early
in
my
career.
There
is
two-
and
I
know
it's
easy
to
generalize
right,
but
but
two
kind
of
general
personalities
that
I
see
when
it
comes
to
a
lot
of
these
types
of
admin
actions
even
aha,
sometimes
right,
and
that
is
there
are
people
who,
unless
they
are
told
they
can
do
something
they
won't
do
it
right,
like
there's
risk
adverse
people,
yeah.
C
Generalizations,
you
know
they
are
what
they
are
they're
a
generalization,
but
that
mentality
and
recognizing
that
is
super
important.
When
you're
talking
about
some
of
these
things,
I
used
to
work
with
a
guy
still
a
really
great
friend
of
mine,
he's
over
at
vmware
now
and
he
he
was
a
windows
administrator
who
became
a
virtualization
admin.
I
was
a
linux
admin
who
became
a
virtualization
admin
and
like
windows,
the
common
you
know
at
the
time
we're
talking
windows,
2000
windows,
xp.
You
know.
B
C
2003
right,
the
the
common
fix
was
just
rebooted.
It'll
come
back,
it's
fine,
you
know,
and
I'm
I'm
like
an
old
solaris
admin.
You
know
and
and
all
this
other
stuff
right
deck
alpha
systems
like
you,
never
reboot
systems
or
or
rarely
reboot
them.
So
it
was
a
very
different
mindset
shift
and
you
know
virtualization,
especially
in
my
virtualization
career.
We
used
a
lot
of
netapp,
nfs
and
nfs.
I
scuzzy
right.
They
all
ride
the
network,
so
you
got
to
train
the
network
admins
like
hey,
don't
just
reboot!
C
A
I
remember
early
in
my
air
force
career.
I
was
helping
troubleshoot
something
at
like
the
base
level.
When
I
say
the
base,
that's,
like
you
know,
think
of
it
as
like
a
metropolitan
area
network
right.
It's
massive
there's
a
lot
to
it,
but
I
walked
into
the
the
main
data
center
for
the
base
and
I'm
looking
at
one
of
the
like
not
a
help
desk
person
but
like
tier
two
guys
and
he's
standing
in
front
of
the
router,
which
is
something.
B
A
Me
in,
for
my
you
know,
bgp
experience
and
everything
else
and
they're
like
we're
having
route
issues
I'm
like,
but
we're
up
right.
Yes,
it's
like
well,
I
was
just
gonna,
restart
the
router
and
see
what
happens
now.
If
anybody
knows
anything
about
cisco
routers
in
the
early
2000s,
you
don't
ever
restart
them.
Never,
unless
you're,
applying
a
firmware,
update
right,
like
restarting
them,
will
just
put
you
back
in
the
same
bad
state
you
had
before
it
doesn't
fix
anything
so
yeah.
A
A
C
B
A
B
C
A
C
Technically,
you
could
deploy
using
the
non-integrated
platform,
agnostic,
bare
metal
method,
and
we
consider
that
a
non-tested
platform,
which
means
that
there
were
some
limits
around
the
support
that
was
possible.
So
now
that
caveat
has
gone
away.
So,
even
though
this
first
set
of
integration
released
is
still
a
non-integrated
deployments,
it
is
now
a
tested
platform.
It
is
fully
supported
in
all
of
the
aspects.
C
C
Can
see
that
they
have
a
certified
csi
operator
in
here,
so
lots
of
great
stuff
happening
inside
of
there.
I
believe
that
we'll
have
a
session
at
nutanix
next.
Whenever
that
happens,
I
think
that's
in
late
september.
That's
talking
about
this
as
well.
So
if
you
are
a
nutanix
customer,
keep
an
eye
out.
There's
lots
of
stuff-
that's
happening
here.
C
So
the
other
thing
or
the
next
thing
I
wanted
to
talk
about,
which
is
conveniently
in
line
with
the
question
that
han
solo
asked,
which
is,
if
you
missed
it
again,
a
little
bit
of
noise
happening
in
the
community.
These
days,
the
nsa
right,
the
the
u.s
intelligence
community's
national
security
agency,
yes
released
a
kubernetes
hardening
guide.
A
C
A
C
Yeah
and
I'll
note
that
this
blog
post,
which
was
written
by
and
it's
a
little
bit
terse
in
my
opinion,
but
so
michael
eppley,
I
know
kirsten
newcomer
so
two
of
the
security
experts
inside
of
openshift
and
in
the
bu
here
they
contributed
to
this
blog
post
and
there's
a
lot
of
stuff
internally
about
how
openshift
already
meets
many
of
these
guidelines.
Right.
There's
a
lot
of
stuff
inside
of
here
that
you
just
get
out
of
the
box
with
openshift.
I
don't
have
details
there.
C
C
Yeah,
nor
is
it
a
standard
right.
It's
it's
not
a
stig.
It's
not
a
you
know
a
hipaa
guideline
or
anything
like
that
right,
the
compliance
operator,
all
of
the
other
tools
available
inside
of
openshift,
are
still
applicable.
This
is
just
the
nsa's
perspective
on
that
so
and
I
I
see
freeman,
I
don't
know
what
version
it's
actually
for.
I
haven't.
I
didn't
pay.
A
A
C
Yeah
we
we,
my
family,
are
cable
cutters,
so
we
stream
everything
and
we
do
likewise
yeah.
So
something
like
two
terabytes
of
internet
a
month
and
it's
it's
all
streaming
services.
We
have.
You
know
the
every
time
the
tv
is
on
it's
streaming.
You
know
some
some
channel
yeah
anyways
exactly
not
to
get
distracted.
C
So
the
third
and
final
thing
I
wanted
to
talk
about,
and
chris
and
christian
appreciate
your
opinions
here,
along
with
the
audience
is
we
had
a
question
that
sparked
some
debate
internally
and
that
was
or
on
one
of
the
internal
chats
and
that's
when
should
I
use
or
when
should
I
create
and
use
more
than
one
namespace,
because,
like
we,
you
know
you
look
in
the
docs
and
you
see
mention
you
know
you
hey,
deploy
this
into
a
new
namespace.
Hey,
create
this
here
from
an
application
perspective.
C
When
is
it
appropriate
to
use
more
than
one
andrew's
response?
Was
more
or
less
it's
a
mechanism
to
apply
our
back
and
other
kind
of
control
policies,
but
there's
no
beyond
that
right.
I
need
to
enforce
a
quota
at
the
name
space
level.
I
need
to
prevent
this
user
from
accessing
other
resources
right,
name,
space
resources
beyond
that
there
isn't
a
technological
reason
right,
you're
you're,
not
you
know,
changing.
B
Yeah
also,
the
network
policies.
I
don't
know
if
you
mentioned
it
right
because,
like
if
you
need
network
isolation
right
between
certain
components,
that'll
also
dictate
what
goes
where
and
what
you
do.
What
I
always
say,
it's
a
reflection
of
your
your
environments.
Like
current
policies,
your
environments
like
how
you
do
work
right,
it's
a
reflection
of
of
all
of
that
right,
so
it
just
it
just
depends
on
how
you
have
your
your
environment
laid
out
right.
How?
B
A
I
will
alert
the
alex
handy
if
need
be,
but
as
far
as
name
spaces
go
like
I
treat
name
spaces
like
I
treat
folders
direct.
Basically
right
like
I
know
that
I
can
lock
them
down
to
a
point
but
past
that
it's
not
a
security
feature
anymore.
It's
just
a
resource
utilization
kind
of
you
know,
coalescion
collection
of
things
or
even
tracking
purposes.
A
Right
like
it
really
depends
like
what
I
what
I
have
you
know
to
to
broaden
the
question
even
further,
would
I
have
a
name
space
for
dev
stage
and
prod
on
the
same
cluster?
No,
don't
run
your
dev
environment
on
the
same
cluster.
That
runs
your
production
environment.
That's
kind
of
my
opinion
there,
but
that's
you
know,
aside
from
just
organizational.
A
That's
kind
of
just
how
I
treat
namespaces
just
to
organize
things
or
you
know,
isolate
them.
If
you
know
some
name,
space
needs
to
be
isolated,
a
certain
way
versus
another
namespace
in
the
organization
because
of
regulatory
reasons,
for
example,
it's
name
spaces
aren't
a
security
feature,
so
I
did
not
treat
them
like
that.
It
is
just
isolation,
basically
so
yeah
these.
A
C
A
C
Which
is
a
good
segue
right
to
that
that
threat
model
or
risk
assessments
also
applies
to
today's
topic,
right,
high
availability
of,
and
it's
funny,
because
in
our
our
little
cheat
sheet,
the
notes
thing
that
I
create
in
here
you
know
one
of
the
first
things
that
I
have
inside
of
there
is
essentially
to
identify
what
it
is.
You
actually
want
to
protect
against
right
and
be
kind
of
a.
C
Yeah,
so
I
I
see
here
freeman,
can
you
run
devin
proud
on
the
same
cloud?
Absolutely
you
can
there's
no
reason
not
to.
I
think
it's
entirely
based
off
of
what
are
your
requirements
for
protecting
those
workloads
and
also
for
you
know,
dev
to
me
is
pre-prod
right.
It's
non-it's
non-production.
If
it
goes
down.
Yes,
it
impacts,
you
know
the
dev
team
and
all
of
that
it's
not
impacting
your
customers.
C
C
C
Because,
just
like
with
disaster
recovery,
we
have
to
identify
our
recovery
time
objective
right?
How
long
will
it
take
for
us
to
get
back
up
and
running
after
a
failure
and
recovery?
Point
objective:
rpo
right
how
how
much
data?
How
much
information
am
I
comfortable
losing?
Hopefully
the
answer
is
zero,
but
it
may
not,
may
not
be
I'm
an
old
storage
guy.
So
zero
is
always
the
answer.
C
So
you
use
those
two
metrics
in
order
to
gauge
well
how
much
infrastructure,
how
much
investment,
how
much
time
you
know
how
how
much
money
do
I
have
to
use
in
order
to
protect.
You
know,
does
that
or
protect
against
disaster
recovery,
and
then
you
take
into
account
the
scenarios.
Am
I
protecting
against
a
rack
falling
through
the
floor?
Am
I
protecting
against
a
data
center
falling
into
an
earthquake
crevice,
maybe
only
applicable
for
christian
out
in
california,
yeah.
B
C
C
So
is
similar,
so
disaster
recovery
is
essentially
is
in
andrew's
definition.
Disaster
recovery
is,
the
site
is
down,
the
application
is
down
whatever
that
happens,
to
mean.
How
do
I
get
it
back
up
is
more
traditionally,
it
hasn't
gone
down,
but
there
is
risk
of
it
going
down
right
of
something
has
happened.
How
do
I,
how
do
I
bring
back
those
resources
to
prevent
it
from
actually
causing
an
outage?
C
C
C
I
think
my
the
funnest
one
I
had
to
plan
for
most
fun,
one
that
I
had
to
plan
for
was
back
to
data
center
flooding,
so
we
had
raised
floors
and
the
main
water
pipe
was
a
12
inch,
water,
pipe
chilled
water
pipe
and
the
volume
of
water
that
was
flowing
through
there
going
out
to
the
cracks
you
know
other
stuff.
It
was.
We
have
this
much
time
before
the
room
literally
fills
with
water.
C
Fun
so
yeah
it's
scenarios
like
that
that
you
got
to
take
into
account
so
h.a,
so
christian.
I
want
to
ask
you
and
get
your
opinion.
You
know
coming
from
again
chris
and
I
have
a
lot
of
military
background,
which
is
a
little
sometimes
different,
sometimes
the
same,
sometimes
different.
Coming
from
the
commercial
side
right,
what's
a
like
a
scenario
that
you
would
have
to
contend
with
from
an
aj
perspective
versus
a
dr
perspective,.
B
B
Well,
and
just
it's
just
localized
right,
localized
failures,
right
you're,
talking
about
like
a
server,
you
know
just
going
out
right,
just
just
frying,
you
know
disks
failing
right.
Server
going
offline
and
high
availability
is
trying
to
protect
against
those
things
right.
Those
localized
things
worse,
like
my
not
my
entire
data
center
hasn't
gone
down,
maybe
particular
components
have
gone
down,
but
my
my
site
is
still
up
and
running
because
you
designed
it
in
such
a
way
where
you
have
high
ability.
B
You
know,
as
the
name
indicates,
high
availability,
whereas
right
disaster
recovery,
you're
talking
about
like
off-site
off-site
backups,
we
had
in
my
previous
job.
We
had
what
we
call.
Oh
no,
not
now.
Now
that
I'm
about
to
say,
I
lost
my
project.
Greenbrier
greenville,
youth
military
guys
should
know
that
off-site
where.
B
Yes,
so
where
essentially
is
like,
you
can
basically
bring
another
site
up
quickly
to
you
know
to
recover
from
from
a
disaster,
something
like
the
whole
date
like
you
guys.
The
situation
where,
like
the
data
center
filled
with
water
or
caught
on
fire,
or
you
know
whatever,
like
we
we've
always
here
in
california,
we
had
a
plan
for
like
the
big
one
right
like
if
an
earthquake
hit
in
our
sight
was
off.
C
B
C
Would
say
yes,
yeah
right
and
and
if
you
think
about
it
in
the
same
sense
of
let's
pick
on
vsphere
right,
if
a
vm
fails
for
whatever
reason
doesn't
matter,
if
it's
a
node
failing
right,
whatever
happens,
if
the
vm
fails,
there's
mechanisms
to
restart
that
virtual
machine,
so
you
could
argue
that
from
a
from
one
perspective
and
and
we're
all,
you
know,
data
center
system
admin.
You
know
ops.
C
Our
careers,
certainly,
I
built
my
career
off,
of
designing
and
implementing
infrastructures
that
provided
as
many
lines
as
possible,
more
or
less
as
cheaply
as
possible,
right
and
and
that's
how
we
we
kind
of
went
about
doing
things.
Certainly
until
probably
you
know,
cloud
native
became
a
thing
and
then
suddenly
there
was
this
dramatic
shift
right
of.
We
have
the
12-factor
application.
You
look
at
the
12-factor
application
right.
What
is
it
12factor.net.
C
C
C
Back
in
the
early
2000s
of
how
can
I
accommodate
this
massive
influx
of
traffic
coming
in
and
it
turned
out
that
that
was
actually
really
well
suited
for
things
like
okay,
now,
I'm
going
to
aws,
remember
aws,
launched
in
yep
and
aws
for
a
long
time
never
had
an
uptime
of
you
know
guarantee
an
sla
for
an
individual
ami
right
or
if
they
did.
I
think
it
was
two
or
maybe
three
nines
yeah,
so
the
application
teams
started
writing
and
creating
applications
to
accommodate
that
failure.
C
If
I've
got
a
hundred
instances
of
my
application
at
any
given
point
in
time
at
least
one
of
them
is
going
to
be
out
right
as
as
infrastructure
guys.
We
know
this
right.
I
used
to
have
a
gigantic
object,
store
right,
literally
hundreds
of
servers
each
one
with
I
think,
16
or
24
drives
in
it
and
every
week,
tuesday
mornings.
C
B
The
netflix,
the
whole
netflix
story
is
essentially
you
you
build
with,
and
that's
kind
of
like
a
good
thing.
I
think
it
came
out
aws
because
I
was
always
trying
to
get
my
my
developers
to
not
trust
the
infrastructure
right
is,
since
you
know
things
can
go
down
like
you
should
write
your
code
to
accommodate
for
that
right.
No.
C
Exactly-
and
this
is
one
of
those
I
used
to
have
conversations
with
folks
all
the
time
of
you
know-
the
infrastructure
is
really
good
at
certain
things
and
yeah.
I
can
keep
adding
nines,
but
every
nine
is
going
to
get.
You
know
exponentially
more
expensive
and
exponentially
more
complex,
whereas
the
application
can
be
written
to
be
super
simple
and
rely
wholly
on
that
infrastructure
or
the
infrastructure
can
be
super
simple
and
highly
unreliable,
and
the
application
can
do
that.
C
C
You
know
new
worker
nodes,
new
control,
plane,
nodes,
you're,
going
to
wipe
out
that
new
lcd
and
replace
it
with
the
old
fcd,
and
that
old
scd
is
going
to
say
who
are
all
of
these
nodes.
And
where
did
you
come
from
and
it's
immediately
going
to
try
and
make
things
look
like
it
used
to
look
like,
even
though
none
of
those
things
are
actually
what
they
used
to
be
and.
C
So
this
is
why
you
know
andrew's
personal
opinion
is
scd.
Backups
are
not
a
disaster
recovery
option.
Lcd
backups
are
for
I
lost
one
or
let
me
rephrase
that
they
are
only
a
disaster
recovery
option.
If
you
still
have
the
majority
of
the
cluster
in
place
and
it's
just
scd
that
failed,
so
something
went
high
haywire
inside
of
there
or
whatever
you
lost
two
of
the
control
plane
nodes.
Whatever
that
happens
to
be.
B
B
Yeah,
it's
only
a
keyboard,
you
don't
you
don't
treat
it
like
a
regular
database.
It's
it's
not
it's,
and
the
like,
like
andrew,
was
alluding
to
the
the
real
backup
mechanism.
Right
is
in.
I
think
two
places
right.
Your
application
manifests
back
up
in
a
get
ops
fashion
and
in
the
storage
right.
So
having
your
for
your
stateful
applications
having
that
storage
resilient,
I
believe
we
already
solved
that
problem
with
things
like
netapp
hitachi.
You
know
amc.
What
have
you
right?
A
Right,
the
the
biggest
thing
is
make
sure
you
have
a
copy
of
every
piece
of
yaml
you've
ever
applied
to
that
cluster
right
seriously.
Right
like
it,
I
know
we
make
it
very
easy
to
just
go
in
and
click
click
click
put
things
together
for
you,
but
you've
got
to
make
sure
that
you
have
a
process,
for
you
know,
saving
that
configuration
data
somewhere
yeah.
C
So
I'm
gonna,
I'm
gonna,
put
us
back
on
track
for
a
moment
yeah.
So
I
just
I
just
shared.
I
know
right
in
a
very
obvious,
very
yeah,
so
I
just
shared
in
the
chat
here
two
links
to
the
two
that
are
up
on
my
screen
here.
So
the
second
one
is
actually
the
first
one
I'm
gonna
talk
about,
which
is
this
kcs
around.
You
know
openshift
container
platform,
high
availability
and
recommended
practices,
so
this
this
kcs
is
rather
interesting.
It
walks
through
a
couple
of
different
aspects.
C
C
If
I
lose
that
one
instance,
my
cluster
is
now
toast,
so
I
could
recover
it
at
that
point
again,
assuming
that
I
have
my
worker
nodes
and
the
other
things
still
present,
but
during
that
time
everything
is
effectively
down,
so
we
always
recommended
having
three
or
more
nodes
for
that
resiliency.
Why?
Three
or
more?
Because
we
need
to
have
a
tiebreaker
right,
we
need
to
have
quorum,
and
that
requires
a
a
prime
number
of
nodes.
In
order
to
do
that,
so
openshift4
we
made
kind
of
the
the
the
command
decision
right
it
was.
C
So
a
lot
of
times
and
by
far
the
most
frequent
question
that
we
get
asked
when
it
comes
to
openshift
high
availability
is
I
have
two
sites.
How
do
I
do
openshift
high
availability
and
the
answer
is
with
one
cluster.
You
really
can't,
because
you're
still
going
to
have
one
site
that
has
the
majority
of
the
nodes
and
if
that
one
site
goes
down
well
you're
right
back
in
the
same
scenario
and
we've
kind
of
alluded
to
this
and
christian
I'll
be
curious
about
your
thoughts.
C
So
we've
alluded
to
this
on
other
shows.
I
think
I
talked
about
it
now.
It
escapes
me.
I
think
it
was
when
brian
bodwin
was
on
of
using
fault,
tolerance
or
similar
technologies
right
so
vmware
fault.
Tolerance
is
effectively
one
vm
that
is
executing
in
two
places
and
if
the
primary
one
it
happens
to
go
down
the
secondary
one,
basically
picks
up
without
any
interruption
to
what's
going
on.
C
So
can
I
use
fault
tolerance
to
have
that
third
control
plane
node
run
in
both
data
centers,
essentially
and
technically.
Yes,
in
practice,
that
is
the
in
practice.
The
problem
is,
there's
a
lot
of
overhead
associated
with
that
which
has
a
lot
of
performance
implications,
which
means
that
there
are
scale
implement
implications
to
your
openshift
cluster,
so
technically
possible,
but
not
recommended.
B
What
I
always
say
when
someone
asked
me,
or
maybe
I'm
a
little
bit
more
abrasive
when
someone
asks
hey,
I
have.
I
have
two
data
centers.
How
do
I
set
up
openshift
to
bha
and
my
answer
is
always
install
two
clusters
right
because
effectively
as
you
as
you
noted,
andrew
you
can't
right
like
you,
it's
it's
like
you're
you're,
asking
how
and
where
you
know
you
just
can't
it's
just
like
the
limitation
of
of
the
of
the
technology
of
fcd.
B
You
know
it's
wanting
it
to
work
out
is
not
gonna
make
it
work
out
right.
So
then
you
have
to
kind
of
change
your
thinking
a
little
bit
when
you
have
some
like
a
situation
right,
like
I
mean
not,
everyone
has
three
data
centers
right,
so
not
everyone
has
right.
Has
the
ability
to
have
three.
You
know
like
you
know,
I
I
totally
get
it
right
like
it's.
You
know.
You
know,
budget
budget.
B
A
B
C
Exactly
yeah
so
on
the
screen
here,
so
what
I'm
sharing
and
I've
posted
the
link
into
the
chat
is
a
blog
post
written
by
rafael
back
a
year
ago,
now
august
21st
time
flies
when
you're
having
fun.
So
talking
about
exactly
what
you
just
said,
christian
right
with
two
data
centers,
essentially
the
best
way
to
achieve
high
availability
is
effectively
an
active,
active
type
of
deployment
for
your
application
right
and
you
can
see
this
is
using
a
global
load
balancer,
you
know
and
it
it
sends
traffic
to
both
places.
C
C
A
B
C
And
so
my
point:
is
this:
isn't
red
hat
being
difficult
right?
This
isn't
andrew!
You
know
you
know
playing
ostrich
sticking
my
head
in
the
sand
and
saying
no,
no
right.
This
is
too
hard
or
you're
doing
it
wrong.
This
is
really
from
multiple
sources
right.
The
recommended
way
of
doing
this
and
kind
of
what
I
was
saying
before
of
the
infrastructure
team,
the
apps
teams
right.
We
need
to
work
together
to
find
the
best
solution
to
the
problem
and
oftentimes
that
requires
you
know
yeah.
C
Well,
you
haven't
done
anything.
Similarly,
at
the
application
level,
do
you
have?
You
know
scheduling
hints
and
what
I
mean
is
you
know:
hey
I've
got
this.
You
know
super
critical
core
service
to
my
application,
microservice
for
my
application
and
I
scaled
it
out
to
be
10
pods
in
size.
So
that
way
you
know
we're
good.
We
can
tolerate
a
node
in
the
cluster
going
down,
but
without
something
like
an
anti-affinity
rule.
You
know
request
in
place
on
that
pod
definition
or
deployment
definition.
C
C
So
there's
a
blog
post.
Let
me
I
actually
don't
have
this
one
up
over
in
this
browser,
so
I'll
pull
it
over
here
there
you
go
so
this
person
talks
about
similar
things
of
distributing
your
application
across
all
of
the
available
resources
right.
So
I
will
copy
and
paste
this
link
into
chat
or
into
the
so
effectively.
C
And
we'll
take
a
peek
here
and
go
to
scheduler
profiles,
so
the
scheduling
profiles
give
the
course
scheduler
a
hint
as
to
hey.
I
want
to
spread
out,
or
I
want
to
consolidate
my
pods
low
node
utilization
means
that
it
is
going
to
prefer
nodes
that
have
the
least
amount
of
workload
on
them.
For
pod
scheduling,
high
node
utilization
means
the
opposite.
It's
going
to
prefer
nodes
with
higher
utilization.
C
C
High
node
utilization
in
andrew's
opinion
is
most
applicable
for
those
times
when
you're,
like
I've
got
my
openshift
cluster
deployed
to
azure,
and
when
I
deploy
new
nodes
in
azure,
I'm
being
charged
for
those
nodes
right
for
those
virtual
machines
inside
of
azure.
So
I
want
to
have
as
few
nodes
as
possible
to
keep
my
costs
down.
You
know,
maybe
that's
one
of
the
reasons
why
you
want
to
do
something
like
this.
A
C
C
C
In
the
virtualization
worlds-
and
you
know
for
those
of
us
who
are
vmware
admins
right-
we
remember
things
like
admission
control
policies
in
our
drs
clusters
right.
We,
you
keep
a
set
amount
of
resources
available
for
those
aha
events.
So
does
the
same
thing
apply
with
kubernetes?
Does
the
same
thing
apply
with
our
openshift
clusters?
C
C
Going
back
to
that
risk
assessment,
yeah
yeah
yeah
of
what
what
am
I
you
know,
one.
What
is
my
application
doing
and-
and
this
is
a
conversation,
another
conversation
that
I
think
you
know
it's
not
always
an
easy
one,
because
we
have
to
have
kind
of
a
robust
understanding
and
a
good
relationship
with
the
application
team.
But,
like
hey
we're
we're,
you
know
all
in
on
openshift
everything's,
going
in
openshift
kubernetes
right
all
of
the
application
components
we're
going
to
bring
this.
You
know
mysql
in
or
you
know,
so.
C
Some
single
point
of
failure
is
that
the
right
thing
to
do.
I
pick
on
my
skill
because
it's
a
database
and
a
lot
of
times,
databases.
You
know
you.
You
have
one
place
where
you
write
to
it
becomes
that
single
point
of
failure,
but
you
have
to
really
carefully
consider
whether
or
not
containerization
is
even
the
right
answer
for
all
of
those
aspects.
C
Does
it
make
sense
to
keep
some
components
in
a
virtual
machine
and
rely
on
something
like
vsphere
to
give
it
what
it
expects
and
what
it
needs
instead
of
you
know,
okay,
maybe
I
do
need
to
refactor
re-architect
the
application
to
take
advantage
of
something
like
cockroachdb
right.
Where
now
I
have
that
same,
you
know,
sql
interface
same,
you
know
rdbms,
but
it's
designed
to
be
that
distributed
cloud-native.
B
Yes,
so
I
think
eric
eric
jacobs
put
it
quite
crassly
right
is
that
kubernetes
or
openshift
doesn't
fix
your
crappy
application,
design,
yeah
right
or
or
I
think,
as
as
kelsey
hightower
eloquently
put
it.
You
can't
rub
kubernetes
on
something
and
make
it
better
right.
There's
going
to
be
some
tax,
you
need
to
pay
in
order
to
take
advantage
of
some
of
these.
You
know
some
of
these
things
that
are
in
place
for
kubernetes
yeah.
B
C
C
I
think
sometimes
people
forget
that
containers
are
an
application
deployment
mechanism
rate
of
I'm
gonna.
Stop
sharing,
since
I
don't
have
anything
to
share
at
the
moment
of,
I
don't
need
kubernetes
to
have
a
containerized
application
and
I
can
use
you
know
podman
on
my
rail
node
to
quickly
deploy
an
application
inside
of
there
without
having
to
go
through
the
rpm.
C
You
know
install
or
a
dnf
local
install.
You
know
without
that
whole
process,
so
I
can
use
a
container
to
quickly
deploy
an
application.
I
used
to
pick
on
sql
server
for
this,
but
somebody
told
me
that
microsoft
stopped
shipping
sql
server
in
windows,
containers
like
if
you've
ever
deployed
sql
server
on
a
windows,
server
right-
and
I
used
to
do
this
right
with
a
virtual
machine.
You
go
and
you
pop
the
dvd
into
the
drive
right
physically
or
virtually,
and
then
you
open
the
installer,
even
if
you
just
blindly
click
next
next.
B
C
Installing
the
app,
but
not
even
getting
it
up
and
running,
not
even
attaching
the
new
database
all
that
other
stuff,
you
know,
can
be
a
significant
amount
of
time
or
I
can
do
well
in
windows.
It
would
be
a
docker
pull
docker
run
and
now
I
have
it
up
and
running.
You
know
on
rel,
sql
server,
you
know
podman
podman
run
and
now
I
have
it
up
and
running
it's
as
quick
as
my
internet
connection.
Basically
so
yeah
containers
can
be
a
deployment
mechanism.
B
Well,
yeah,
the
inputs
like
like
the
infrastructure,
should
be
a
reflection
of
the
of
the
need
right
you
shouldn't
have
to
you
know
you
you
shouldn't,
have
to
try
to
contort
your
your
your
application
or
or
your
needs
into
the
technology
right.
The
technology
be
in
service
of
you
know
whatever
it
is,
that
you're,
providing
whatever
service
you're,
providing.
C
C
It's
five
minutes:
five
minutes:
okay,
yeah,
okay,
yeah.
So
if,
if
a
node
fails
for
whatever
reason
it's
five
minutes
before
the
scheduler
will
take
action,
basically
it
gets
declared
unreachable
and
available
and
then
it'll
reschedule
that
workload.
You
can
shortcut
that
by
doing
things
like
deleting
the
node
from
the
cluster
yeah,
and
so
that's.
C
Well
so
so
this
is
an
issue
with
openshift
virtualization
right
we've
we
have,
you
know
hey.
I
need
my
virtual
machine
to
restart.
You
know
when
it
fails
and
if
kubernetes
takes
five
minutes
to
recognize
that
the
vm
you
know
the
node
is
gone,
but
we
want
the
pod
to
restart
before
that.
How
do
we
do
that?
C
So
reese
actually
figured
out
that
if
you
do
like
ipi
bare
metal
ipi,
you
can
use
a
node
health
check
that
will
quickly
identify
like
hey
this
node
stopped
responding,
it
must
be
gone
and
then
the
action
is
to
delete
the
node
from
the
cluster
and
then
all
that
workload
immediately
gets
rescheduled
and
we
can
rely
on
the
ipi
mechanism
to
add
a
new
node
back
in
so
where
I'm
going
with.
All
this
is
there's
another
aspect
as
well.
It's
not
just
you
know
this
deploy
a
pod
leave
it
alone.
Kubernetes.
C
B
B
C
Well,
the
pdb
pod
disruption
budget
yeah.
It
took
me
a
minute
to
remember
that
one
so
is
also
really
important
for
upgrades
and
updates
yeah,
because,
yes,
if,
if
basically
we're
taking
down
nodes
and-
and
so
you
know
in
order
to
do
reboots
and
all
that
other
stuff,
so
if
you,
you
know
cluster
level
things,
we
want
to
be
as
disruptive
as
least
disruptive
as
possible.
C
So
when
you
do
those
the
number
of
nodes
that
is
affected
by,
for
example,
a
machine
config
update
or
a
cluster
update,
is
controlled
in
the
machine,
config
pool
so
there's
a
setting
inside
of
there.
The
default
is
one
node
at
a
time,
but
you
can
turn
that
all
the
way
up
to
like
100
of
nodes,
because
it
can
be
number
based
or
percentage
based,
so
hey,
reboot,
100
of
the
nodes
at
the
same
time,
inside
of
this
machine,
config
pool
that
seems
bad
right.
C
C
B
Yeah
yeah,
I
know,
and
it's
it's
a
like
the
the
whole.
How
long
has
chris
you've
been
involved
in
devops
for
a
while?
How
long
has
that
that
whole
idea
been
around
it's
like?
We
need
to
work
closely
together
more
now
than
anything
else,
especially
in
this
cloud
native
world
in
this
kubernetes
world.
There's
there's
no,
no
more!
This
throwing
over
the
fence
sort
of
thing
or
where
infrastructure
guys
buy
something
and
put
it
in
place
and
they're
like
oh
by
the
way.
B
To
and
it
needs
to
be,
it's
a
holistic
approach
right
and
especially
when
you're
getting
into
something
like
in.
I
don't
wanna
derail
it
too
much,
something
like
acs
right
where
you're
talking
about
application
delivery
supply
chain
where
there's
no
such
thing
as
dev,
there's
no
such
thing
as
tests
right
like
because,
if
you
think
of
a
factory
right,
there's
no
like,
if
you
think,
if
you
think
of.
C
C
Team
should
have
that
you
know
ability
to
have
a
conversation
around
hey.
This
looks
wrong.
We're
going
to
impact
this
and
yeah
modern
times.
You
know
I'm
getting
way
over.
My
skis
here
of
you
know
we
call
it
devops
or
sre,
or
you
know,
whatever
whatever
in
vogue
term,
is
but
there's
a
a
ton
of
and
there's
actually
a
I
used.
I
had
this
conversation
with
a
number
of
customers
previously
around
brene
brown
brene
brown.
C
Does
a
phenomenal
ted
talk
around
vulnerability
in
relationships
and
you
kind
of
have
to
have
that
willingness
that
openness
between
dev
and
and
the
ops
teams
right
infrastructure
in
the
application
teams
to
be
able
to
be
successful
and
and
all
of
that
in
this
cloud
native
world.
So
I
know
I'm
getting
all
mushy
and
stuff
like
that.
A
A
If
you
don't
trust
the
other
person
sitting
aside
from
the
table
from
you,
then
you
need
to
do
things
to
build
that
trust,
or
vice
versa.
Maybe
they
need
to
do
something
to
help
build
that
trust
and
whether
they're
self-aware
enough
to
do
that
is
the
question
so
yeah,
so
to
kind
of
drive
home.
The
devops
point
right
like
this,
isn't
anything
super
new.
A
B
B
Nowadays,
right:
well,
I
I
always
see
it
as
like.
Now
we
actually
have
the
technology
to
do
all
the
things
we've
been
wanting
to
do
right
right
and
so
now
we
have
the
ability
to
do
like.
I
remember
talking
about
in
my
previous
shop
talking
about
autonomous
data.
Centers
right,
like
a
data
center
should
be
autonomous
and
if
one
goes
down,
it
doesn't
matter
right.
We
lose
you
know,
you
know
those
records
are
over
here
and
we
can
reprocess
those
orders,
and
you
know
that
was
hard
before
a
lot
of
this
stuff.
C
C
I
came
over
to
the
vendor
side
and
johnny,
who
now
works
for
red
hat
as
well.
He
and
I
were
co-conspirators
on
like
regional
disaster
recovery
for
virtual
desktops,
wow
just
yeah
create
and
again
it
goes
back
to
well
as
an
infrastructure
admin.
I
can
make
the
desktops
the
vms
available,
but
what
about
the
profiles?
What
about
this?
What
about
anyways?
So
I'm
reading
through
chat
here,
I
think
we
addressed
all
of
the
questions
I
do
see
so
cleed.
I
see
you
asked
about
descheduler.
C
C
Alejandro,
how
is
data
consistent
across
two
clusters,
so
that
was
going
back
to
our
global
load?
Balancer,
you
know
having
two
clusters
and
two
data
centers.
So
this
is
one
of
those
there's
kind
of
two
ways
to
do
it.
One
of
them
is
from
an
ops
perspective,
infrastructure
perspective
right,
punting
it
to
the
apps
team.
Like
yeah,
you
know
cockroachdb
right.
I
mentioned
that
before
of
having
a
database
instance
that
is
naturally
cloud
native.
C
You
know
something
like
that,
and
there
are
certainly
others
out
there.
Cockroach
just
happens
to
be
the
one
that
comes
to
mind,
because
I
was
recently
recently
fiddling
with
it
in
operator
hub
and
they
do
have
a
certified
operator
by
the
way.
Yeah
you
know,
another
way
would
be
to
do
something
like
storage
replication,
you
know,
use
odf
openshift
data
foundation,
which
is
the
new
name
for
openshift
container
storage,
and
you
know
replicating
that
data
from
one
cluster
to
another.
C
That
gets
a
little
weird
in
andrew's
opinion,
because
it
would
on
the
secondary
site
it's
either,
because
it's
two
separate
openshift
clusters,
but
one
storage
cluster.
So
you
might
have
to
do
some
things
inside
of
there.
It
might
even
be
an
external
instance
where
you
treat
it
as
an
external
instance.
That's
replicating
data
that
goes
back
to
the
get
ops
thing
of
you
know.
Do
I
just
use
git
ops
to
reintroduce
that
pvc
that
pv
in
a
controllable
fashion.
B
Yeah
it's
it's
always
like
you
said
you
can
punt
it
to
to
the
devs
and
say:
okay
well,
design
your
application.
This
is
what
you
got
right.
You
could
also
do
it
on
the
infrastructure-wise,
whereas
essentially
it's
an
active
passive.
Where
you
have
your
application,
you
have
your
things
running
you're
replicating
your
data
over
either.
You
know
using
netapp,
snapmirror
or
using
odf
replication.
Whatever
right,
your
storage
system
needs
to
replicate
offside
and
then
using
get
ops.
B
You
can
you
know
when
a
failure
happens,
you
just
you
know,
attach
the
storage
reapply
the
manifest,
and
you
should
like
come
up
right.
So
there's
there's
a
it's,
then
that's
like
kind
of
like
the
active
passover
or
hot
cold,
whatever
you
want
to
call
it
design
between
two
data
centers
or
you
can
get
really
really
cool
like
net
like
netflix
and
basically
have
two
data,
centers
and
they're
autonomous.
Essentially,
where
you
know
you
have
data
replicating
forward,
you
know
cross
replication
and
the
application
is
smart
enough
to
know
that
hey.
C
C
C
C
C
So
I
I
do
appreciate
you
joining
us
for
your
perspective,
both
from
the
ops
as
well
as
the
dev
side,
as
well
as
your
experience.
So
again,
audience
don't
hesitate
to
reach
out
at
any
point
in
time,
we'll
I'll
hang
out
on
the
chat
here
for
a
few
more
minutes
even
after
the
stream
ends.
So
you're
welcome
to
send
us
any
other
questions.
B
Night
tomorrow,
wow
yeah,
I'm
actually
so
I'm
following
suit
with
andrew.
So
tomorrow
get
ups
guide
to
the
galaxy.
I'm
having
josh
packer
from
the
acm
team
right
andrew
had
had
someone
from
the
say
jimmy
from
the
ac.
Now
I'll
I'll
go
to
the
actual
engineering
route.
I
decided
to
one-up
you
and
get
the
engineer.
B
C
Sorry
so-
and
I
will
be
tuning
in
to
that
one
because
it's
I
know
acm-
has
some
ability
to
apply
yaml
files
right,
kubernetes
objects
that
are
inside
of
a
git
repository,
but
that
is
not
git
ops.
That
is
not
argo
cd.
So
I'm
I'm
curious
to
hear
about
the
differences
between
the
two
different
capabilities.
C
Yeah.
So
I
see
a
promo
apologies
for
mispronouncing
anybody's
name
is
openshift
supports
every
kind
of
software
or
applications.
So
the
answer
there
is
yes
and
no
open
shift
is
kubernetes.
Kubernetes
is
an
orchestrator
for
containers
and
containers
are
basically
a
it's
a
linux
kernel,
it's
a
process
that
has
been
isolated
from
the
rest
of
the
system
using
some
kernel
features,
c
groups
etc.
C
So,
in
that
respect,
yeah,
if
you
can
execute
it
on
a
linux
system,
it
could
technically
go
into
openshift,
but
it's
more
complex
than
that,
because
kubernetes
has
its
own
set
of
things
that
it
wraps
around
that
around
scheduling
and
and
all
of
that
other
stuff.
So
the
answer
is
almost
certainly
yes,
but
the
bigger
answer
is
there
might
be
a
lot
of
additional
work
that
has
to
go
into
it.
For
it
to
be
kubernetes
ready,
even
if
it's
you
know
a
traditional
monolithic,
app.
C
B
B
A
A
C
C
C
C
So
I'm
going
to
pick
on
odf
because
it's
in
our
portfolio
and
I'm
not
offending
anybody
except
ourselves,
right
of
so
with
the
defaults.
You
know
with
odf,
and
let's
say
I
have
you
know
three
data
nodes
and
I
think
the
default
is
or
the
minimum
is
three
so
for
every
byte
that
I
store
it's
being
replicated
to
three
nodes,
so
my
application
store
is
one
gigabyte.
My
disks
have
three
gigabytes:
one
gigabyte
on
three
different
nodes
consumed
in
order
to
protect
that
data.
C
Writing.
That
three
times
has
performance
implications
right.
It
has
to
land
on
on
all
three
nodes:
there's
a
network
traversal
that
has
to
happen
there
as
well
as
they
all
have
to
validate
that
they
got
the
data
and
then
respond
back
that
they
got
the
data
right,
so
that
has
a
performance
implication
from
availability
perspective
on
the
other
hands
right.
I
can
now
tolerate
node
failure.
C
So
there's
a
bunch
of
different
solutions
out
there
right.
If
you
go
to
the
marketplace
and
look
there's
a
number
of
similar
solutions
that
are
that
are
available
to
be
deployed
into
the
cluster
really
comes
down
to
those
things.
On
the
other
hand,
there's
a
lot
of
external
storage
right.
Basically,
all
of
the
major
storage
vendors
now
have
the
ability,
through
csi,
to
connect
your
you
know,
pvcs
to
their
storage
solution.
So
you
know
chris,
you
mentioned
netapp
before
right.
Netapp
has
both
asynchronous
and
synchronous
replication,
just
like
every
other
storage
vendor.
C
B
Yeah
that
it
makes
me
think
well
with
podman.
It's
probably
doesn't
answer
the
question,
but
by
the
way,
with
podman,
you
can
take
your
containers
that
you're
building
with
podman
and
export
them
as
a
yeah
as
a
kubernetes
object
right.
So
you
can
like
export
this
as
a
deployment
and
you
can
even
pipe
it
into
cube.
Ctl
apply.
So
not
sure.
If
that
answers
your
question,
but
I
decided
to
throw
that
out
because
it's
actually
pretty
cool
and.
C
C
C
So
we
have
all
of
these
different
scenarios
on
here
so
like
oh,
I
want
openshift
basics.
Okay,
I
want
to
learn
about
here,
deploying
applications
from
scenarios
and
I
hit
start
scenario
and
I
am
dropped
into
you
notice.
I
didn't
log
into
anything.
I
didn't
have
to
do
anything.
It
was
like
five
seconds
even
with
me
running
my
mouth
and
I
now
have
a
whole
tutorial
to
run
through
here.
C
If
you
have
a
developer
entitlement,
you
are
also
entitled
to
up
to
16
cores
of
openshift,
so
anybody
can
go
completely
free
of
cost
and
have
quote
unquote
real
openshift
available
to
you.
It
is
not
supported.
It
is
only.
You
can
only
use
that
entitlement
for
single
user
non-production
work,
but
you
can
absolutely
do
that.
I
will
note
that
you
cannot
actually
title
it
yet
there's
an
error
on
the
back
end
that
we're
still
working
on
getting
sorted
out.
A
A
A
A
Managed
services
on
aws
and
azure-
and
I
forget
the
other
ones
but
yeah,
there's
ways
that
we
can
manage
it
on
your
cloud
or
you
can
manage
this
yourself
completely.
You
know
and
if
you
need
a
sandbox
to
plan,
we
give
you
that
for
free
for
30
days
it
does
have
some
limitations,
but
it's
pretty
open
right
now,
right,
like
yeah,
you.
C
Yeah
yeah,
so
this
the
sandbox
as
the
title
indicates,
is
really
geared
at
developers.
Folks,
who
I
just
want
to
get
in,
and
I
want
to
start
deploying
pods
writing
code
right.
It
includes
the
code
ready
workspaces
inside
of
there
just
to
kick
the
tires
and
see
what's
going
on,
and
I
don't
remember
if
it's
15
days
or
30
day
or
30
days,
30
days
yeah,
so
you
can
get
in
and
do
exactly
that
out
here
with
like
the
developer,
even
though
it's
a
developer's
entitlements
as
an
administrator
as
an
ops
person.
C
A
C
Oh
yeah,
so
it
took
me
over
to
console.redhead.com
which
I'm
already
logged
in,
and
it
tells
me
exactly
that
right.
You
know
hey.
I
want
to
get
started
with
your
assisted
installer
and
goes
through
and
does
everything
for
you.
If
you're
curious
about
assisted
installer,
we
have
a
couple
of
shows
on
assisted
installers.
A
C
B
C
Yeah
and
so
again
that
developer
account,
which
this
is
my
developer,
account
I'm
not
logged
in
as
an
employee
right
now,
you
can
absolutely
go
in.
If
you
have
everything
you
need
from
a
credential
perspective,
you
notice
I
can
go
and
I
can
deploy
to
aws
you
get
that
same
60-day,
eval
entitlement
or
16-core
developer
entitlement.
So
if
you
want
to
see
what
openshift
looks
like
in
aws
by
all
means,
you
know
test
out
the
openshift
ipi
experience
in
aws.
A
B
A
C
So
the
the
contemporary
to
minicube
is
code
ready,
containers
right.
So
if
we
do
a
code,
ready
containers
can't
type
and
talk
so
code
ready
containers
is
more
or
less
that
same
experience
for
again
local
development
with
openshift.
So
it
does
require
you
to
have
a
relatively
powerful
machine
to
run
on.
B
C
Just
just
like
you
know,
docker
on
mac
os
or
any
of
that
stuff
or
mini
cube
for
that
matter.
It
will
create
a
virtual
machine
and
deploy
openshift
into
that
virtual
machine.
So
I
don't
remember
what
the
minimums
are.
I
think
it's
like
4,
cpus
and
10
or
12
gigs
of
ram
for
that
virtual
machine,
which
means
that
your
physical
hosts
should
have
you
know
probably
six
or
eight
cpus,
and
you
know
16
gigs
of
ram,
but
that
that
is
certainly
an
option.
B
A
C
All
right,
oh
so,
a
deal.
I
think
this
will
be
the
last
question
because
we're
way
over
our
normal
time,
so
don't
don't
hesitate
to
reach
out
again
andrew.sullivan
redhat.com.
So
in
a
disconnected
environment,
how
does
red
hat
know
the
number
of
clusters
used
by
a
customer?
And
the
answer
to
that?
Is
we
don't
no
so
yeah?
So
essentially
we
we
rely
on
you.
We
trust
you,
the
customer,
to
entitle
your
clusters
through
subscription
manager
right
so
or
console.redhat.com.
C
B
C
C
A
Coming
up
later
today
on
the
channel
is
red
hat
enterprise.
Linux
presents,
I
believe,
we'll
be
talking
about
some
network
manager
type
stuff
so
tune
in,
for
that
we
have
a
new
host
coming
along
there
and
then
yeah
we're.
Actually,
we
we're
having
a
red
hat
recharge
day
is
what
we
call
it
on
friday,
so
we'll
be
off
air
completely.
On
friday,
we
will
I'll
be
around
on
the
internets,
though.
A
A
C
Of
friday
keep
an
eye
on
the
blog,
so
opengl.com
blog,
which
will
redirect
wherever
it's
at
now
so
we
every
week
following
these
streams,
we
have
a
blog
post
that
summarizes
everything
that
we
talked
about
with
links
to
the
recorded
video
and
where
we
talked
about
it,
so
keep
an
eye
out
for
that.
Usually
they
come
out
on
fridays,
assuming
I
can
get
everything
ready
and
over
to
alex
on
the
on
the
blog
team.
It'll
be
staged
and
ready
to
go
out
friday
morning
so
but
yeah.
C
So
thank
you
so
much
to
our
audience
appreciate
everybody.
Staying
with
us
today.
Thank
you
again
to
christian.
Very
much
appreciate
you
coming
on
be
sure
to
tune
into
his
stream
tomorrow.
3
p.m,
eastern
12
p.m,
pacific,
and
I'm
not
even
going
to
attempt
the
translation
into
utc
keep
an
eye
on
the
streaming
calendar.
I
know
chris
is
dropped.
Thank
you.
Chris
has
dropped
the
link
in
there
a
couple
of
times
and
with
that.