►
Description
OpenShift 4.11 was recently released with a large number of changes and new features. For this week’s livestream we’ll focus on a number of those features and capabilities which are relevant for OpenShift administrators.
Want to learn more?
Try Red Hat OpenShift: https://red.ht/3uVxkya
Follow us:
Andrew Sullivan https://twitter.com/practicalandrew
Jonathan Rickard https://twitter.com/jrocktx1
OpenShift https://twitter.com/openshift
Subscribe to Red Hat's YouTube channel: https://www.youtube.com/redhat/?sub_confirmation=1
About Ask an OpenShift Admin:
Do you have questions? We want to answer them! Every Wednesday at 11am ET, hosts Andrew Sullivan and Jonathan Rickard sit down to answer your questions about Red Hat OpenShift.
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 on redhat.com https://red.ht/3rYz5ZS
OpenShift on GitHub https://github.com/openshift/
#RedHat #OpenShift #Kubernetes #ContainerPlatform
A
B
Hello
and
welcome
to
this
week's
ask
an
open
shift.
Administrator
office
hours
live
stream,
I
am
andrew
sullivan
and,
as
always,
I
am
joined
by
my
co-host,
mr
johnny
ricard.
How
are
you
today,
sir?
My
man
I'm
good
to
go?
How
are
you
you
know?
I
can't
complain
too
much
my
I
I
will
say
I'm
slightly
discombobulated.
B
I
I
swapped
out
deaths.
You
know
we
had
a
long
weekend
last
week
or
the
week
before.
I
don't
remember
which
so
I
think
I
said
last
week
that
I
swapped
out
desks.
Well,
my
my
you
see
my
microphone
arm.
Here
is
it's
in
my
way,
and
today
I
gotta
type
and
the
anyways.
So
I
was
sitting
here
adjusting
things
and
trying
to
get
it
out
of
my
way
right
before
we
were
starting-
and
I
just
I
feel
discombobulated,
but.
B
Yeah,
the
condensers
are
fine
in
a
quiet
room.
This
is
not
a
quiet
room
like
I
can
hear
right
now.
The
people
across
the
street
are
getting
their
lawn
mowed,
which
is
very
loud
anyways.
So
I
I
thought
about
maybe
doing
like
a
shotgun
microphone
or
something
you
know
putting
it
putting
it
right
up
here
up
next
to
the
camera,
anyways
enough
will
be
whining
right.
You
know.
B
So
today's
topic
is,
you
know
something
that
is
exciting
to
johnny,
and
I
we've
got
this
absolutely
massive
list
of
things
to
talk
about
to
chat
about.
You
know
for
folks
who
watched
the
it
was
a
90
minute.
What's
new
and
4.11
session,
it
covered
kind
of
9
000
things
in
a
very
short
period
of
time,
and
it
covers
things
that
span
a
broad
range
of
stuff.
B
So
we
tried
to
go
through
and
you
know
distill
that
list
and
and
get
it
down
to
just
things
that
we
think
that
you
all
our
audience
of
administrators
are
going
to
care
about.
So
hopefully
we
we
hit
the
mark
on
most
of
these
things.
If
there
are
things
that
you
are
interested
in,
but
you
don't
hear
today
fear
not
because
we
won,
we
were
always
willing
to
talk
about
new
things,
so
don't
hesitate
to.
Let
us
know
if
you
want
to
hear
about
something,
but
we
also
have.
B
Yeah
I'll
I
I
I
was
going
to
talk
about
it
on
top
of
mine,
but
I'll
talk
about
it
now,
which
is
so
few.
Let's
see
here,
future
4.11
streams
so
next
week
we're
doing
a
vmware
special.
So
I
will
be
at
vmware
explorer,
which
is
the
new
name
for
vmworld,
so
we're
gonna
focus
on
all
things:
vmware,
there's
a
number
of
changes
that
happened
with
vmware
in
4.11.
We'll
talk
about
those
we'll
talk
about
vmware
at
a
big
picture
level.
B
Hopefully
I
can
find
and
drag
a
vmware
person
in
with
me,
but
no
guarantees
so
next
week
will
be
vmware.
I
think
a
week
or
two
after
that,
we'll
be
talking
about
composable
openshift,
we'll
be
talking
about
the
alternative
recommender
for
the
vertical
pod,
auto
scaler,
along
with
custom
metric,
auto
scaling
the
tech
preview.
Kita
we've
got
a
stream
planned
around
logging
5.5,
so
talking
about
changes
to
loki
vector
right.
All
of
that
other
stuff.
That's
that's!
Just
what
we've
got
like
concrete
planned.
We've
got
a
bunch
of
other
stuff
coming
up
as
well.
B
Software
security,
software
supply
chain
security,
one
that
I'm
personally
excited
for
which
is
optimizing
resource
utilization
using
the
descheduler
pod
eviction
and
the
secondary
schedulers.
So
yeah.
All
of
that
to
say
you
know,
that's
like
what
eight
eight
episodes
you
know
the
next
two
months
worth
of
content.
So
please
subscribe,
you
know,
do
do
the
thing
on
the
platform
wherever
you're
at
so
that
way,
you
know
when
we
come
online,
you
can
also
go
to
red.ht,
livestream
and
you'll,
see
the
live
streaming
calendar
and
all
that
other
stuff
there.
C
B
B
For
4.11,
it's
funny
because,
with
every
release
I
talk
with
product
marketing
and
you
know
product
marketing
says
you
know.
Oh
this
is
you
know
this
isn't
a
big
release.
You
know,
there's
not
a
lot
going
on,
there's
just
these
little
things
and
they
think
about
it
completely
differently,
because
they
think
from
the
terms
of
like
marketing
campaigns
and
like
oh,
can
we
go
out
and
talk
about
this?
You
know
big
picture,
you
know
what
do
they
call
them
like
stories
or
market
problems.
B
You
know
how
are
we
addressing
the
edge
space
and
they
look
at
it
and
say:
oh,
like
ztp,
you
know
big
deal
like
well
yeah,
but
that
enables
you
know
easy
deployment
and
provisioning
of
clusters
at
the
edge
when
used
in
conjunction
with
acm,
like
it
adds
up
to
this
bigger
thing,
and
I
feel
like
anyways,
where
I'm
going
with.
All
of
this
is
we've
been
stuck
in
that
rut?
B
If
you
will
for
the
last
four
versions
where
product
marketing
is
just
like
meh,
it's
a
small
release
and
meanwhile
I'm
like
no
no
there's
so
much
going
on
and
4.11
is
like
the
start
of
a
lot
of
really
big
things
happening
with
openshift.
I
think
4.12
and
4.13
we're
going
to
see
a
lot
more
you're,
seeing
it
in
tech
previews
now
so
things
like
kita
kids
is
a
big
deal
right.
B
It's
really
powerful
functionality,
kcp
hosted
control
planes.
You
know
what
we're
currently
calling
micro
shift,
but
I
I
think
they're
trying
to
come
up
with
a
new
name
for
it,
there's
just
so
much
going
on
so
yeah
yeah
logging,
5.5
isn't
just
for
ocp-411
you're
right,
our
hope,
although
it
is,
they
align
their
releases
with
open
shift
releases
for
whatever
reasons,
along
with
several
other
things,
what
else
was
released
recently
vlogging.
A
B
Yeah
openshift
virtualization
4.11
is
also
in
the
near
future.
They're
we
were
talking
with
them.
Did
I
include
you
on
that
conversation?
I
think
I
did
anyways.
I
was
talking
with
them
about
having
them
come
on
and
have
a
session
with
us
about
what's
happening
in
openshift
virtualization,
with
their
new
release
as
well.
So
you
know
never
ending
st
for
pun,
intended
stream
of
topics
right.
B
So
I
I
guess,
no
reason
not
to
keep
rolling
with
some
top
of
mind
topics
while
we're
at
it
so
just
to
formalize
our
introduction
to
top
of
mind.
So
these
are
things
that
johnny
and
I
have
identified
that
are,
we
think,
important
or
maybe
interesting
to
you
all
that
aren't
necessarily
related
to
today's
topic.
So
by
all
means
ask
us
questions
ready
to
add
in
things
as
you
see
fit
we're
happy
to
have
a
chat
about
any
of
these
they're,
just
not
directly
related
to
today's.
B
B
So
remember
when
a
release
goes
ga,
it
is
ga
for
new
installs,
but
it
takes
some
time
for
it
to
be
updated
or
added
to
the
update
channels.
I
I
say
that
so
it
upd
added
to
the
stable
upgrade
channel.
So
when
it
goes
ga
you
can
use
fast
in
order
to
upgrade,
in
this
case
your
4.10
clusters
to
4.11.
Remember.
That
is
a
fully
supported
update
as
well.
B
So
if
you
want
to
use
fast,
you
want
to
get
to
4.11
right
now.
Something
goes
wrong.
Pick
up
the
phone
call
us
and
they'll
help
you
out
with
that,
however,
and
why
I
started
saying
this
is
because
there
is
currently
a
it
is
a
the
new
phrase
that
they
told
me
to
use.
It
is
not
recommended
but
supported
due
to
or
they
they
pulled
the
the
default
upgrade
edge
so
that,
if
you
just
go
in
where's
my
window
here.
B
B
C
C
B
B
B
B
There
is
a
here
we
go.
There
is
an
issue
that
was
found
with
quote
unquote
highly
loaded
clusters,
so
you
can
see
here.
Nodes
with
substantial
numbers
of
containers,
for
example,
200
plus
and
contention
for
cpu
time
are
likely
to
have
the
rpmosd
tree
service
failed
to
start
so
from
what
I
understand,
there's
no
danger
to
workload
here
effectively.
B
I
hope
that
was
clear.
I
feel
like.
I
was
really
long-winded
there.
It's
a
lot
of
information,
yeah
random
question.
If
nm
state
operators
installed
but
no
instances
will
this
still
affect
ovn
kubernetes,
I
don't
believe
so
as
far
as
I
know
so
in
them
states
if
you're,
not
even
if
you
are
configuring
things
so
long
as
you
are
not
configuring
or
adjusting
the
interface,
that
is
the
quote-unquote
primary
interface
for
ovn
kubernetes.
It
should
be
fine.
B
B
B
If
you
look
at
the
host,
that's
also
the
one
that
has
the
ovs
bridge
and
ovian
bridge
and
all
that
other
stuff
on
it.
So
as
long
as
you're,
not
adjusting
that
network
interface
so
day,
two
you
go
in
and
say:
oh,
I
really
wanted
to
bond
e0
and
eth1
right
and
then
like
that,
would
cause
issues.
If
you
say,
oh,
I
really
wanted
to
bond.
You
know
e2
and
e3.
That's
not
going
to
affect
anything
you'll,
be
perfectly
fine
to
do
that.
B
B
In
markets
around
the
world,
every
time
you
upgrade
a
product
need
to
re-certify
it
yeah
yeah.
Another
interesting
thing:
there
are
hope
nine.
So
we
for
certified
operators
we
don't
require,
but
we
strongly
encourage
our
partners
to
re-certify
with
each
one
of
the
eos
releases,
so
just
keep
an
eye
on
and
the
partners
they
have
to
go
in
and
update
their
operator
to
add
support
for
each
new
version.
Anyways.
B
B
B
Let's
see
oh,
so
we
talked
a
little
bit
before
about
I'll
open
up
a
new
tab
for
this.
So
we
talked
before
about
the
pod
security
admission
or
pod
security
policies,
changing
with
openshift
4.11,
where
the
old,
restricted,
psp
or
scc
is
no
longer
being
used
by
default
for
new
openshift
4.11
deploys.
It
now
uses
restricted
v2,
which
basically
drops
everything
as
opposed
to
only
dropping
a
subset.
B
So
if
you
look
through
this
blog
post,
it's
a
really
thorough,
really
in-depth
blog
post
that
walks
through
all
of
these
changes,
and
this
one
is
written
in
the
context
of
you
know
you
all
customers,
users,
right
administrators,
as
opposed
to
the
other
one
that
we
first
linked,
because
that
one
was
more
targeted
at
partners.
Just
so
happened.
The
partner
team
was
faster
to
get
their
blog
post
out.
So
the
the
content
here
or
the
the
end
result
here
should
be
basically
identical.
B
A
B
C
B
That's
funny
see,
I
think
I
only
got
one
more,
which
was.
This
is
an
interesting
one.
So
I
I
don't
know
why.
I
guess
there's
two
things
here.
So
one
back
in
the
four
eight
or
four
nine
days
we
changed
the
cdn,
that's
used
for
distributing
openshift
images,
so
hopefully,
by
now
we've
all
you
know
caught
up
on
this.
So
if
you
had,
for
example,
specific
firewall
exceptions,
you
know
hey,
I
need
to
allow
any
external
access.
B
There
is
a
kcs
that
was
that
had
all
of
the
urls
all
of
the
domains
that
you
needed
to
whitelist
the
openshift
clustered
access
so
that
got
updated.
You
know
three
or
four
versions
ago,
but
for
whatever
reason,
with
openshift
4.11
when
we
went
to
you,
know
mirror.openshift.com
and
looked
at
in
particular
the
core
os
dependencies
inside
of
here.
This
wasn't
populated
for
a
while.
C
B
B
B
If
I
go
all
the
way
up
to
the
top
you'd
be
able
to
see
the
format
of
this,
but
here's
x86
and
I
can
kind
of
scroll
through
and
say:
okay,
here's
eleon,
here's
aws,
azure,
azure
stack,
gcp,
ibm
cloud
metal
and
all
the
different
components
of
metal.
Like
here's,
the
iso.
If
you
wanted
to
download
the
iso-
and
you
can
just
use
this
url
to
directly
do
like
a
curl
or
a
wget,
or
something
to
pull
that
in
so
nutanix
openstack
qmu
vmware.
B
B
I
mean
I
guess
you
could
always
you
know
where
where'd
my
thing
go
over
here.
I
guess
you
could
always
you
know
dependencies
coreos,
4.11
latest,
and
then
these
are
named.
You
see
both
with
the
version
number
4.11
at
zero,
as
well
as
down
here
without
a
version
number,
so
you
could
always
script
against
that.
B
A
That's
awesome.
Yes,
you
just
reminded
me
of
something
so
like
you're
talking
about
like
you've
always
seen
it,
but
you
haven't
really
paid
attention
to
it
so
like
when
I
was
when
I
would
be
learning
something
new
right.
What
I'd
do
is
I'd
run
the
cli
so
like
oc
and
then
do
dash
health,
and
I
would
kind
of
like
walk
through
each
one
of
those.
You
know
things
to
kind
of
see
the
different
options.
So
like?
A
Oh
okay,
if
I
want
to
do
this,
like
you
know,
here's
kind
of
that
workflow,
but
as
I've
gotten
older
I've
gotten
lazier.
So
I
don't
do
that
nearly
as
much
but
like
that's
a
good
idea
too,
if
you're
trying
to
learn
something
I
take
that
cli
and
just
run
it
and
then
do
dash
dash
help
or
just
kind
of
like
tap
it
out
and
see
what
pops
up
and
then
kind
of
walk
it.
So
that
way
you
can
kind
of
get
an
idea
of
like
what
you
can
and
can't
configure
from
the
cli.
B
Yeah,
I
I
I
always
encourage
folks
to
to
the
explain
you
know
you've
heard
how
many
times
have
you
heard
me
say
you
know,
use
use,
oc,
explain
and
or
openshift
install
explain
to
dig
through
all
the
options
and
all
the
features
that
are
inside
of
there.
So
that's
just
another
version
of
that
of
use,
dash
dash
help
and
explore
each
one
of
those
commands
yeah.
That's.
A
B
Although
it
does
kind
of
remind
me,
years
and
years
and
years
ago,
when
I
was
a
linux
admin
and
they
would
ask
me
to
do
technical
interviews
for
folks,
I
would
usually
find
like
the
most
esoteric
crazy
like
what
is
this
option
on
this?
You
know
command
that
nobody,
since
the
1980s
has
used-
and
it
was
more-
you
know
it
wasn't
a
stump.
The
chump
type
of
thing
right
or
you
know
I
expect
people
to
know
every
aspect
of
every
command
of
every.
A
Yeah
yeah
mine,
mine,
mine's,
more
of
a
bully,
but
just
because
people
did
it
to
me
is
name
name
10
commands
in
linux.
That
are
three
letters,
and
so
it's
like
you
know,
there's
the
easy
ones
where
you
go:
oh
okay,
ssh
and
then,
as
you
start
getting
like
further
out
you're
like
oh
wait,
wait
hold
on.
I
love
that
one
that
one
or
the
two
letter
commands.
B
A
B
B
I
don't
forget
to
all
right,
so
let's
talk
about
openshift,
4.11
and
what's
new
for
administrators,
so
just
real
quickly.
This
is
a
reminder
or
just
a
reminder
that
this
is
office
hours,
so
office
hours
means
that
we
are
here
for
you
for
our
audience
to
ask
or
to
answer
rather
any
questions
that
you
may
have,
so
it
doesn't
matter
if
it's
related
to
4.11
or
anything
like
that.
If
you've
got
questions
about
openshift,
ask
us
we'll
do
our
best
to
answer
here
on
the
stream.
B
If
we
don't
know
the
answer,
which
does
happen
rather
frequently,
we'll
do
our
best
to
find
those
answers
while
we're
on
the
stream
by
reaching
back
to
folks
or
searching
you
know
doing
that
type
of
stuff.
If
we
can't
get
that
answer,
while
we're
here
on
the
stream
we'll
follow
up
in
next
week's
stream
or
whenever
the
next
stream
happens
to
be,
you
can
also
reach
out
to
us
anytime,
on
social
media.
B
So,
if
you're
not
watching
us
live,
if
we
don't
satisfactorily
answer
your
question
here,
I
I
we
try
our
best
not
to
miss
any
questions.
If
we
happen
to
miss
a
question
here
in
chat
or
something
like
that,
don't
hesitate
to
reach
out
to
us.
You
can
reach
me
at
andrew.sullivan
redhat.com.
I
love
email.
Email
is
great,
it's
the
the
best
least
distracting
method
of
anything.
B
Ever
he
says
sarcastically,
but
no,
don't
don't
don't
hesitate
to
send
us
email
or
twitter
at
practicalandrew,
as
well
as
on
reddit
and
all
the
other
places.
Thank
you
stephanie
our
producer
here.
In
the
background
for
popping
up
my
information
there
and
johnny
is
at
jrocktx
on
social
media
or
at
johnny
j
or
excuse
me
not
at
johnny,
but
johnny
j-o-n-n-y
redhead.com.
B
B
So
the
problem
with
that
is
kvm
when
we,
when
we
think
of
or
specifically
think
of
like
libvert
kvm,
it's
limited
to
a
single
physical
node
and
for
better
or
worse,
I
andrew's
interpretation,
so
andrew's
opinion,
which
means
that
I
cannot
be
held
responsible
for
what
comes
out
of
my
mouth
is
that
we
discourage
that
because
it's
not
a
not
a
good
deployment
right.
So
if
I
deploy
a
five
nodes,
you
know
virtual
kvm
virtualized
openshift
cluster
onto
a
single
physical
server.
It's
still
a
single
point
of
failure,
so
rather
we
would
want.
B
You
know,
want
encourage
you
to
use
something
like
red,
hat,
virtualization
or
openshift,
virtualization
or
openstack,
or
something
like
that
to
do
the
kvm
based
piste
kumar.
Is
it
stable
enough
to
upgrade
4.11.1
from
4.10.26,
so
I'll
say
I
have
not
had
any
issues
upgrading,
so
I've
upgraded
four
clusters.
Now
I
did
have
one
compact
cluster,
where
one
of
my
control
plane
nodes
was
throwing
a
random
error
and
I
literally
just
rebooted
it
and
it
fixed
itself
and
went
back
to
normal.
B
So
stable
is
relative.
You
have
to
decide
whether
or
not
you
want
to
you
know,
accept
any
real
or
perceived
risk
associated
with
doing
those
updates.
Red
hats
are
our
official
determination
of
quote
unquote.
Stable
is
when
it
reaches
the
stable
upgrade
channel,
but
just
remember
that
fast
is
fully
supported.
If
you
choose
to
do
that
before
it
reaches
stable,
yeah,
our
hope,
nine
oven,
kubernetes,
ov
and
kubernetes
is
slowly
becoming
the
default
sdn.
B
You
know
we
talked
about
it
gosh,
it's
been
like
40
or
50
episodes,
since
we
had
mark
curry
on,
and
it
was
back
in
like
the
four
six
or
four
seven
days,
and
at
that
time
you
know,
the
thought
was
well
in
4.9
or
or
something
like
that.
4.8
or
4.9
ovn
would
become
for
every
platform
that
defaults,
ovn
kubernetes
would
become
the
default
sdn
for
various
reasons
that
got
pushed
out
hasn't
happened,
yet
we
should
probably
follow
up
with
mark.
B
Have
him
back
on
the
stream
because
it's
been
a
while,
and
I
recently
learned
that
one
mark
lives
here
in
the
raleigh
area.
I
actually
saw
him
in
the
tower
and
we
chatted
and
like
I
thought
you
lived
in
boston
or
something
but
no
and
two
while
we
were
chatting.
I
also
learned
a
bunch
of
stuff
about
gateway
api.
So
there's
a
lot
of
really
cool
stuff.
That's
coming
down
the
pipe
so
hopefully
we'll
be
able
to
have
him
back
on
in
the
not
too
distant
future.
A
B
Yeah
yep
yeah
mark
is
a
fountain
of
information.
If
you
will
yeah
our
hope,
9
is
supposed
to
be
defaulting
4.8,
so
yeah
that's
gotten
pushed
out.
I
think
it
was
made
the
default
for
it's
some
some
deployment
types
and
places.
I
think
single
node
open
shift,
it's
the
default.
Maybe
compact,
I'm
not
sure
I
maybe
bare
metal
ipi,
it's
it's
in
places
here
and
there
I
don't
have
the
list
off
the
top
of
my
head
dave.
I
think
412
is
the
new
target
yeah.
I
I
honestly
don't
know
like
that's.
B
So
I
I
know
there's
lots
of
cool
stuff
in
ovn
kubernetes.
My
stance
has
always
been
stick
with
the
default.
Unless
you
know
you
don't
need
the
default
and
the
good
news
is
with
openshift
sdn.
You
can
do
a
update
or
a
two
or
a
change
modification
redeployment.
I
don't
know
you
can
change
to
openshift
or
ovn
kubernetes
after
deployment.
You
can't
go
the
other
direction,
but
you
can
go
openshift
sdn
to
ovn
kubernetes
day.
Two,
let's
see
what
else
did
I
have?
What
have
I
got
open
here?
B
B
B
B
So
if
I
switch
over
to
my
4.11
cluster
here,
oh
yeah,
a
lot
of
people
seem
to
think
it's
a
big
deal.
If
there's
a
dark
mode,
I
I
agree.
I
like
dark
mode.
If
you
are
curious,
so
my
I'm
on
a
mac,
my
system-wide
preference
is
dark
mode,
so
it
automatically
picked
up
on
that
and
used
it.
But
if
you
go
to
your
user
up
here
and
go
to
user
preferences,
you
can
change
the
theme.
B
Yeah,
I
I
agree.
I
I
find
it
generally.
I
am
in
a
slightly
darkened
room,
so
this
this
helps
with
the
whole
screen
glare
issue
and
all
that
other
stuff.
Anyways.
Sorry,
I'm
losing
focus
if
we
come
down
here
to
observe
you'll
notice
that
if
we
go
to
dashboards
right,
it's
all
the
normal
dashboards
that
are
here.
B
If
we
go
to
metrics,
there's
no
longer
the
little
link
to
open
in
prometheus
or
anything
like
that,
and
if
we
come
up
here
to
networking
and
go
to
routes
and
select
all
projects,
there
is
no
routes
for
either
grafana
or
for
prometheus.
So
you
can
see
the
prometheus
federate
endpoints.
So
this
is
these.
Are
the
metrics
and
points
that
you
can
query
to
get
data
out,
but
it
is
not
the
like.
B
You
know
if
you
deploy
grafana
or
something
like
that,
that's
what
you
would
point
to
grafana
add
as
a
data
source,
but
there's
not
the
query.
Interface,
that's
available.
You
know
that
traditional
ui
and
grafana
is
just
not
present
at
all.
So
we've
been
talking
about
this
for
a
long
time.
It's
finally
finally
happened.
If
you
will,
the
logic
here
is
that
from
what
I
understand
I'll
caveat
it
with
that
there
was
some
sort
of
problem
issue
trouble
hardship.
I
don't
know
with
translating
openshift
rbac
into
grafana
rbac.
B
So
andrew,
you
know
in
my
an
solo
username
up
here.
You
know
if
I
only
have
access
to
two
namespaces
when
I
would
open
that
grafana
dashboard.
There
was
no
way
for
it
to
limit
me
to
just
seeing
the
two
namespaces
that
I'm
allowed
to
see,
whereas
with
the
built-in
dashboards-
and
I
happen
to
be
a
cluster
administrator
here
with
the
built-in
dashboards,
I
can
only
see
the
namespaces
that
I
have
permissions
to.
B
So
a
lot
of
folks
ask
well
what
if
I
wanted
to
customize
those
well
that
grafana
instance
that
shipped
with
the
openshift,
the
core
openshift
cluster,
has
always
been
read.
Only
we
never
supported
making
any
modifications
to
it.
We've
always
said
since
day,
one
right,
you
have
to
deploy
your
own
grafana
to
create
custom
dashboards.
B
So
the
exact
same
is
true
now,
and
we
talked
about
that
back
in
episode,
63
where's,
my
notes,
yeah
63.,
so
in
episode
63
we
did
a
demo
of
exactly
how
to
deploy
grafana
using
the
operator,
how
to
point
it
to
the
endpoint,
both
thanos
as
well
as
prometheus,
and
you
know,
set
up
those,
those
dashboards
and
stuff.
So
our
hope,
nine
griffon
are
back.
Isn't
that
a
license
module,
maybe
yeah?
I
honestly
have
no
idea
I
have
I.
I
will
admit
that
I
have
not
dug
deeply
into
grafana.
C
B
Or
hope:
nine!
Yes
correct!
If
you
want
to
use
windows
nodes,
you
do
have
to
set
the
hybrid
overlay
modes,
which
can
only
be
done
at
install
time.
But
if
you
have
no
intention
of
using
windows
nodes,
you
can
make
that
day.
2
shift,
I
want
to
say
I
thought
I
heard
that
they
were
going
to
do
away
with
that
requirement
as
well.
A
B
B
So
it's
basically
saying
I'm
going
to
wait
until
network
manager
says
everything
is
up
and
running
right.
The
node
has
an
ip
all
of
that
other
stuff,
and
then
the
next
set
of
services
can
do
their
thing.
Apologies
for
any
network
manager
or
systemd
people
who
are
far
smarter
than
me
on
both
of
those
technologies
that
I
just
butchered
a
very
basic
explanation,
so
why
that
would
be
causing
ovs
or
why
ovs
would
would
cause
that
issue.
I
don't
know
off
the
top
of
my
head.
B
That
sounds
like
something
has
gone
wrong
in
the
deployment
of
the
node
or
something
like
that
and
you
may
need
to
redeploy
it.
I
will
say
that
I
do
occasionally
see
nodes
that
that
I'll
log
in
I'll
sshn
it'll
say
that
that
service
has
failed,
but
it's
functioning
normally.
So
I
don't
know
what
that
means
necessarily.
A
B
Hey
called
yeah
things
are
going
very
wrong
for
this
cluster
yeah.
If
ovs
isn't
able
to
come
up,
then
the
sdn
won't
come
up,
which
means
that
none
of
the
rest
of
the
stuff
will
be
able
to
to
configure
so
telemeter
anatoly.
I
don't
know,
but
just
based
off
of
that
little
bit
of
information.
It
doesn't
ring
a
bell
if
you
can
send
an
email
to
johnny
or
I
j-o-n-y
at
red,
hat
or
andrew.sullivan
at
red
hat.
B
Where
would
you
go
to
look
for
logs?
I
would
check
the
system
logs.
So
if
you
can
ssh
into
the
node,
so
if
it
has
a
node
ip
but
not
a
sdn
ip,
I
would
ssh
in
there.
I
would
check
for
system
logs
around,
like
network
manager
configuration.
I
would
also
check
for
any
pods
or
containers
related
to
the
sdn.
B
B
This
is
a
result
of
csi
and
we'll
talk
more
about
csi
in
a
minute
as
well
so
effectively.
The
vmware
csi
driver
requires
virtual
hardware
version
15.
virtual
hardware
version
15
is
not
available
until
6.7
update
3..
So
that's
why
it
has
been
removed
right,
we're
on
the
road
to
removing
the
entry
or
stopping
the
entry,
vmware
provisioner
from
being
present.
C
B
Being
removed
from
upstream
kubernetes,
not
not
just
from
openshift,
so
update
one
vsphere
7.0
update
one
or
earlier,
I
think
again,
so
don't
hold
me
personally
responsible.
I
think
that
this
is
a
result
of
vmware's
support
policy.
You'll
notice,
that's
just
deprecated
and
not
removed.
So
if
we
do
vsphere
supports,
I
think.
C
B
I
don't
remember
specifically
why
now
and
why
we're
deprecating
update
one
or
earlier
I
don't
know
if
that's
lumped
in
with
the
vmware
support
policy
or
something
like
that,
but
just
be
aware
that
if
you're
not
on
at
least
vsphere
7.0
update
2,
it
is
considered
deprecated
right
now
and
then
related
to
that
6.7
update
to
earlier
here's
virtual
hardware
version.
13.
support
is
removed.
So
keep
that
in
mind.
B
A
C
B
Yeah
this
one-
I
don't
know
why
this
isn't
up
above,
but
here,
red
hat
virtualization
is
a
host
platform
for
openshift
container
platform
will
be
deprecated,
so
so
red
hat
virtualization
is
moving
into
so
currently
it
is
fully
supported
for
the
next
week.
So
the
end
of
this
month
it
goes
into
the
maintenance,
support
phase
of
its
life
cycle.
C
B
Mean
to
support
in
so
end
of
life
phase.
Sorry
maintenance.
So
it's
currently
in
full
support
goes
to
maintenance
yeah,
so
full
support
ends
august
31st.
So
next
week,
then
it
goes
into
the
maintenance
phase.
Maintenance
support
ends
in
two
years
and
then
it
is
in
basically
end
end
of
life
or
extended
life
phase.
B
So
you
can
see
extended
life
phase
out.
Here
is,
basically,
you
have
access
to
the
bits.
If
we,
if
it's
a
known
issue
or
something
like
that,
support,
will
help,
but
there's
no
new
patches,
no
bug
fixes
no
security,
fixes
right.
None
of
that
other
stuff
inside
of
there.
So
this
is
that
that
deprecation
of
rev
as
a
platform
for
openshift
is
a
result
of
rev
slowly
going
away.
A
I
just
want
to
back
up
on
one
thing:
really:
quick,
the
so
the
automatic
generation
of
service
account
secrets
just
keep
in
mind
if
you're
doing
anything
where
you're,
maybe
like
parsing
out
the
secret
or
that
token
from
a
service
account
that
you're
expecting.
So
if
you're
spending
up
a
new
cluster
and
you're
starting
a
service
and
you're
expecting
like
the
that
service
account
that
you
create
to
get
that
token
automatically.
You're
gonna
have
to
generate
that
now.
So
just
keep
that
in
mind.
A
So
there's
tools
like
vault
and
stuff,
like
that,
where
you
can
have
that
externally,
create
that
for
you
and
then
loop
it
into
like
an
external
secret
or
something
like
that.
But
just
if
you
do
any
type
of
automation
where
you're
doing
that
like
I.
This
is
specific
to
me,
because
I
we
have
a
tool
where
we're
actually
running
something
like
this,
and
it
will
break
some
of
the
things
like
once
we
get
there.
So
just
I
wanted
to
kind
of
just
emphasize
that.
B
A
B
B
You
know
definitely
worth
looking
into
and
here's
some
details
about
removing
support
for
esxi
vsphere
stuff,
like
that,
it's
funny
that
they
have
both
vsphere
6.7
update,
2
and
esxi
6.7
update
2..
I
wonder
why
they
did
that
so
yeah
all
right
moving
on
a
little
bit,
so
we
already
talked
about
it.
Last
week
we
had
a
whole
stream
about
it,
nutanix
ipi.
So
if
anybody
happened
to
miss
that
go
back
and
check
out
episode,
78,
but
yeah
nutanix
ipi.
B
B
B
One
thing
I
didn't
do
so
heterogeneous
heterogeneous
heterogeneous,
I'm
terrible
at
english.
My
apologies
are
in
tech
preview,
so
normally,
when
we
talk
about
or
when
we
do
these
streams,
I
only
talk
about
you
know.
We
only
want
to
talk
about
ga
features,
things
that
most
folks,
you
know,
are
comfortable
using
today.
This
one,
I
think,
is
really
cool
and
I
think
it's
something
that
a
lot
of
folks
are
going
to
be
interested
in,
so
I'd
definitely
encourage
you
to
to
test
it
out
to
try
it
out.
B
Keeping
in
mind
it
is
tech
preview.
So
the
goal
here
is:
you
can
mix
cpu
architectures
inside
of
the
same
cluster.
To
me
where
this
is
really
cool
is
with
things
like.
You
know,
aws
azure,
google
clusters,
where
I
can
use
less
expensive
instances
for
things
like
the
control
plane,
so
I
can
deploy
an
arm
based
instance,
for
the
control
plane,
maybe
for
infrastructure
nodes
and
then
use
x86
specifically
for
my
applications
that
require
x86,
and
then
I
can
reduce
my
cost
inside
of
each
one
of
those
clouds.
B
You
can
also
do
things
like.
I
was
chatting
with
some
red
hat
folks
this
morning
that
were
excited
about.
You
know
I
can
deploy.
You
know
arm
based
or
something
like
that
control
planes
inside
of
a
cloud
and
then
have
x86
based
worker
nodes,
using
remote
worker
nodes
at
edge
sites
and
stuff.
Like
that
again,
you
know
reducing
that
cost
and
in
that
instance,
being
able
to
keep
the
control
plane.
Maybe
closer,
you
know
my
edge
edge
nodes
are
in
apac,
so
I'll
use.
B
A
A
However,
you
want
to
pronounce
it
but
had
it's
daniel
messer
and
I
can't
remember
the
other
person
right
now:
oh
benchmarks,
yeah
and
so
yeah
they
put
together
some
stuff
with
oc
mirror
which
went
ga
and
4.11,
so
oc
mirror
if
you're
not
familiar
with.
It
is
a
tool
that
you
can
use
to
simplify
your
your
content
collection,
so
you
think
about
like
the
4.11
release
or
4.10
release
some
operators.
A
You
know
if
you've
ever
gone
through
that
process
to
try
and
pull
the
operators
outside
of
you
know
the
oc
adm
mirroring
process.
It
can
be
a
little
cumbersome
and
I'm
saying
that
very
tongue-in-cheek,
because
it's
it's
pretty
cumbersome,
it's
a
pain,
but
this
tool.
You
know
it
helps
a
ton
with
that
and
then
you
know
you
combine
that
with
the
mirror
registering
capability
that
you
know,
dan's
team
put
out
earlier.
This
year,
I
think
it
now
you've
got
like
a
whole
suite
of
tools
available
to
you
that
are
they're.
A
A
B
Very
cool,
yeah
and-
and
don't
forget,
mirror
registry
right
so
being
able
to
have
it,
deploy
an
instance
of
quay
on
a
rail
node
so
that
you
can
deploy.
You
know,
pull
all
of
those
images
over
into
a
fully
supported
registry
instance
yeah.
So
I
just
realized
that
I
should
probably
go
faster
here,
because
we've
got
like
15
things
still
to
cover
so
I'll
I'll,
stop
running
my
mouth
so
much
here,
external
dns
operator
has
gone
ga
so.
B
Already
have
it
up,
probably
not
so
external
dns,
so
this
was
another
one
that
we
talked
about.
I
think
in
episode,
63
as
well
right
johnny,
so
external
dns
operator
allows
you
to
yeah
episode.
63
was
grafana
and
external
dns,
so
this
allows
you
to
reach
out
and
configure
external
dns
for
these
services
and
routes
inside
of
your
cluster.
B
B
It's
a
downstream
of
the
external
dns
operator,
that's
publicly
available
that
supports
all
kinds
of
stuff.
We
talked
about
that
in
the
other
stream.
So
I
I
wonder
if
it's
possible
to
do
like
an
rfc
compliant
dns
server
on-prem
like
bind-
and
you
know
sure
it's
not
supported,
but
I
don't
know
anyways
project-based
default
subdomain
for
routes,
so
this
was
a
slide
that
piqued
my
interest.
B
So
here
I'm
still
in
the
docs.
Where
am
I
in
the
docs?
Oh
I'm
in
the
release.
Notes
so
inside
of
here
we
have
this:
where
did
it
go?
Really,
I've
completely
lost
here
we
go.
Basically
what
this
does
is
if
you
have
route
shards,
so
I
deploy
with
you
know
my
standard.
You
know:
apps
start.apps.clustername.domainname
domain
name,
yeah
we'll
go
with
that.
But
then
I
want
to
create
multiple
shards.
B
B
When
you
create
a
route
that
used
a
subdomain
you'd
have
to
specify
the
spec.host
and
it's
you
know:
spec.host
equals
you
know
routename.subdomain.domain
or
cluster
name
domain
name,
blah
blah
blah.
Now
I
can
just
specify
spec.subdomain
and
it
uses
rate
it
assembles
the
full
domain
name
based
off
of
that.
So
it'll
pull
in
the
route
name;
it'll,
pull
in
that
subdomain
value;
it'll
pull
in
the
cluster
name,
the
domain
name,
all
that
other
stuff
to
make
it
easier
and.
B
I
could
do
you
know
whatever
my
sub
domain
is
for
this
particular
project.
I
don't
have
one
set
up
here,
but
that's
all
I
would
need
to
do
as
opposed
to
having
to
do
the
you
know.
Sub
domain
is
say
and
solo
for
this
one,
whereas
before
I'd
have
to
do
host
and
I'd
want
to
do
like
okay,
this
is
my
hello
dot
and
solid
dots.
Whatever
my
cluster
name
is
aoa
dots,
you
know,
domain.tld
blah
blah
blah
blah,
which
you
can
see
in
mine,
is
like
this
super
long
thing.
B
B
Here
we
go
using
dns
forwarding.
So
if
we
look
in
the
docs
here,
I'm
under
networking
and
understanding
the
dns
operator
and
then
we
have
a
using
dns
forwarding
section
here.
If
I
scroll
down,
we
have
this
upstream
resolvers
segment
and
I
can
set
the
transport
to
tls
and
then
provide
the
ca
bundle.
So
if
you
want
to
do
you
know
encrypted
dns
queries
over
tls,
you
can
use
that
with
the
cluster
level.
Dns
resolver
now.
B
B
B
B
You
know
if
there's
three
or
five
or
ten,
however
many
pods
backing
that
route,
it
will
choose
the
one
that
has
the
least
number
of
connections
now
you
can
see
it
has
switched
over
to
random,
so
just
be
aware
of
that,
it
may
cause
some
different
behavior
or
you
may
need
to
scale
those
pods
differently
depending
on
how
that
works,
because
it's
no
longer
aware
of
any
of
those
connections
inside
of
there.
I
don't
know
if
that
will
make
scalability
better
or
worse,
but
it's
different
so
just
be
aware
of
that.
B
Looking
at
chat,
randomness
is
always
is
my
wife's
favorite
word
by
the
way
or
hope
none,
I
don't
know
I
I
don't
know
why
that
why?
Why
that's
her
favorite
word,
but
that
is
her
favorite
word.
Let's
see
yes
for
any
red
hatters,
there
was
a
a
lengthy
internal
email
chain
about
the
which
english
to
use
to
pronounce
I'm
gonna,
I'm
gonna,
say
kiwi
or
koi,
or
I
don't
know
how
do
I
most
brutally
mispronounce
it
to.
A
B
Let's
see
some
core
os
related
things,
so
core
os
has
been
rebased
to
rel
8.6.
So
if
you
had
any,
you
know
driver
dependencies.
Anything
like
that
that
you
needed
8.64.
Please
be
aware
of
that.
Somebody
asked
about.
I
think
this
was
on
reddit.
Somebody
asked
about
fips
for
4.11
and
or
4.10
4.11
on
power.
B
B
So
when
rel
has
that
achieves
that
certification,
then
basically
the
openshift
folks,
openshift
engineering
will
go
through
and
validate
it
and
then
it'll
be
available,
but
that
that's
the
dependency
there
for
anybody
who
might
be
waiting
a
fairly
interesting
one
for
me
or
to
me
anyways-
is
that
the
way
that
machine
config
operator
does
updates
has
changed
so
historically,
it
kind
of
just
randomly
selected
nodes
right.
So
in
the
case
of
my
cluster
here,
let's
see
I've
got.
B
A
C
B
The
other
normal
stuff
worker
latency
profiles,
so
this
one
I
had
to
think
about
this
one
for
a
minute
right,
because
it's
at
surface
value,
when
you
think
about
the
name
and
even
just
looking
at
like
this
headline,
improving
cluster
stability
using
worker
latency
profiles,
you
think,
oh,
this
is
going
to
be
something
like
you
know
in
via
in
vsphere
deployments
we
usually
recommend
folks
at
latency
sensitivity
at
the
hypervisor
to
high.
So
that
way
it
has
more
access
to
resources
and
stuff.
Like
that.
B
So
my
first,
when
I
first
learned
about
this,
I
was
thinking.
Oh
okay,
it's
going
to
be
a
way.
You
know
something
that
we
can
use
through,
like
tunde
or
the
performance
add-on
operator,
or
something
like
that
to
basically
tune
for
latency
for
the
workload.
That's
not
it
at
all.
So
it
turns
out
what
this
actually
does
and
there
is
a
blog
post
johnny.
I
don't
know
if
you
or
or.
B
Yeah
it's
on
cloud.reddit.com,
blog
but
effectively
it
determines
how
long
or
let
me
rephrase
that
if
we
start
at
the
top
here,
the
documentation
explains
that
the
control
plane
will
heartbeat
each
one
of
the
worker
nodes
every
10
seconds
by
default
right.
So
we
have
this
node
status,
update
frequency
if
it
doesn't
respond
within
the
node
monitor
grace
period
of
40
seconds.
It
stops
scheduling
workload
to
that
node
after
300
seconds.
B
It
declares
the
node
as
unreachable
and
will
begin
rescheduling
workload
that
was
originally
on
that
node.
So
that's
the
default
behavior
and
what
this
does
is.
It
allows
us
to
tune
those
values
based
on
the
reliability
of
the
network,
connection
that
we
have
so,
for
example,
if
I
go
with
medium
worker
latency
profile,
you
can
see
that
it
now
checks
only
every
20
seconds
and
it
will
wait
until
there
are
two
minutes.
B
That's
an
interesting
one.
So
I
wonder
why
that
is
pot
eviction
period
is
oh
reduced
to
60
seconds.
Okay,
so,
basically
we're
adjusting
for
how
much
disconnect
how
flaky
the
network
in
between
the
control
plane
and
the
worker
node
happens
to
be
inside
of
there.
Oh,
I
see
what
it
is
so
after
two
minutes
it
will
then
wait
an
additional
60
seconds
before
rescheduling
workload,
so
it
takes
longer
to
declare
that
the
node
is
unreachable,
but
it
will
more
quickly
begin
to
reschedule
workload
down
here.
B
B
So
this
is
particularly
useful
again
remote
worker
node
type
scenarios
other
other
places
where
you
might
have
a
less
than
reliable
network
inside
of
here
I
will
say
that
be
cognizant
of
how
this
might
affect
you
know
any
pod.
You
know
workload
aj
type
of
activities,
because
it
could
mean
that
it
takes
longer
for
them
to
get
rescheduled.
It
could
mean
that
they,
you
know
all
of
that
other
type
of
stuff.
So
just
be
aware
of
that,
and
I
think
they
talk
about
that
in
the
blog
post.
If
I'm
remembering
correctly.
B
B
B
B
So
just
a
reminder-
and
we've
been
talking
about
this
for
a
while
as
well,
I
think,
since
the
4.8
days
when
this,
the
red
hat
shipped
vsphere
csi
driver,
was
first
tech.
Preview
of
eventually
all
of
the
entry
storage
provisioners
are
being
removed,
so
everything
will
be
csi,
all
all
storage
providers-
all
storage
drivers
will
be
csi,
so
we're
starting
to
see
that
take
effect
rates
csi
for
vsphere
and
with
a
bunch
of
these
others.
B
B
So
I
think
that
do
I
have
it
here:
gcp
openstack
vmware.
So
I
think
that
let
me
rephrase
this
at
some
point,
for
example
with
vsphere,
if
you've
been
using
the
upstream,
the
vmware
shipped
vmware
csi
driver.
Essentially
today
you
would
continue.
You
have
the
option
of
continuing
to
use
that
or
you
can
use
the
red
hat
one
and
I
think
it
is
yeah.
B
So
here's
the
the
documentation
for
how
to
remove
that
driver
and
then,
basically,
when
the
red
hat
one
gets
deployed,
it
takes
over
management
of
all
of
those
volumes,
so
there's
no
real
risk
to
the
workload
or
to
the
storage,
because
they're
effectively
the
same
right.
It's
just
taking
over
the
management
piece
inside
of
there,
and
I
don't
think
any
of
these
are
a
part
of
the
data
path
like
csi.
I
csi
drivers
themselves
are
around
the
management
of
volumes.
B
That
doesn't
mean
that
some
storage
vendors
don't
have
or
don't
install
a
package
or
a
driver
or
something
like
that
to
connect
to
their
storage.
I'll
pick
on
odf
right,
odf
uses
ceph
rbd
for
block
based
volumes,
so
the
csi
provisioner
just
does
volume
creation,
volume
deletion
and
then
there
is
a
csi
agent
on
each
one
of
the
nodes
that
will
actually
mount
the
volume
and
do
other
actions
so
that
agent
on
each
one
of
the
nodes
is
what
will
actually
install
and
manage
like
the
rbd
driver
package
on
the
host.
B
B
B
So
yeah,
if
you're,
if
you're,
deploying
single
node
openshift,
which
single
node
openshift
itself,
is
fully
ga
today,
it
is
exactly
that
it
is
a
single
node
of
openshift
in
the
future,
or
I
guess
today
using
tech
preview,
but
in
the
future,
when
it's
fully
supported,
you'll,
be
able
to
add
worker
nodes
to
that.
So
it's
still
a
single
control,
plane
node.
That
is
also
a
shared
worker.
B
Node,
probably,
but
you
can
add
additional
worker
nodes
onto
that-
that
one
control
plane
node
note
that
there
is
no
way
to
convert
from
a
single
control,
plane
node
to
three
control
plane
nodes.
So
you
can't,
you
know,
start
off!
You
know.
Oh,
I
just
need
a
super
small
deployment,
single
node,
openshift
and
then
oh,
no,
my
application
has
grown
to
be
a
monster
and
it's
production.
B
Oh
actually,
I
think
I
broke
my
my
stage
here.
I
was
gonna
show,
there's
a
command
shift,
install
create
single
node
ignition
config,
we'll
just
do
a
dash
dash
help
here
and
basically
it
works.
Just
like
a
regular
openshift
install
create
ignition
config
command.
It'll
output,
two
things,
though
so
one
will
be
the
bootstrap
in
place
for
that.
First
single
node
instance,
so
it'll
create
that
bootstrap
in
place
ignition
file,
which
you
could
then
use
the
coreos
installer
command.
B
To
add
that
ignition
file
to
the
iso-
and
you
can
take
that
iso
and
then
go
and
boot
your
node
to
it
and
get
everything
up
and
running
or
you
can
use
assisted
installer,
but
it
will
also
generate
a
worker
ignition
file.
So
you
can
then
use
that
worker
ignition
file
to
boot
any
additional
nodes
and
have
them
look
to
our
single
node
openshift
instance
for
machine
config
and
go
and
do
its
thing.
B
B
So
I
I
don't
think
that
you
can
add
those
before
then
or
even
generate
retrieve
the
worker
dot
ignition
tour
guidelines
to
find
out
which
latency
profile
you
should
set.
Is
there
a
tool?
No,
I
think
the
guidelines
are
really
around
those.
Wherever
my
thing
went
here,
it's
really
around
those
latency
numbers.
B
So
if
you're,
if
you're
expecting
there
to
regularly
be
periods
of
time
where
the
node
cannot
be
reached
for
two
minutes,
then
you
know
you
would
want
to
consider
setting
you
know,
maybe
that
worker
latency
profile
or
five
minutes.
However,
the
case
may
be,
you
can
see
setting
it
is
done
on
a
per
node
basis.
You
set
it
in
the
node
object
worker
latency
profile.
B
B
B
We
look
at
remote
worker
nodes
like
if
you
look
at
the
documentation
here,
you
don't
do
anything
to
configure
those
nodes
right.
You
see,
there's
no
like
nothing
that
you
do
here.
It's
all
about
awareness
of
what
happens
when
those
remote
worker
nodes
lose
connectivity,
because,
basically,
what
will
happen?
Is
it
can't
heartbeat?
So,
therefore,
new
workload
stops
being
scheduled
after
five
minutes,
the
node
that
thinks
it's
disconnected
will
start
shutting
down
those
pods,
because
it's
assuming
at
that
point
that
the
control
plane
is
surviving
and
the
control
plane
is
now
rescheduling
that
workload.
B
So
it
starts
shutting
it
down.
So
the
documentation
here
is
really
about.
Like
you
know
down
here
where
you
did
see
a
little
bit
of
yaml
is
like
how
do
I
allow
my
pods
to
continue
to
run,
even
if
my
remote
worker
node
becomes
disconnected,
you
see,
you
set
a
whole
bunch
of
tolerations,
so
that
way
it
doesn't
get
booted
by
the
by
kublin,
so
that
that's
where
in,
in
my
opinion,
my
perspective,
this
is
really
helpful.
Remote
worker
nodes
is
the
main
use
case
for
those
latency
profiles.
B
B
They
were
created
by
you
by
a
human
or
some
automation
that
that
one
that
somebody
other
than
red
hat
created,
whereas
with
ipi
the
installer,
goes
out
and
creates
all
of
those,
but
the
end
result
should
be
a
cluster
that
is
99
percent
the
same
between
up
upi
and
ipi.
The
biggest
difference
will
be
with
ipi.
The
worker
nodes
are
created
by
a
machine
set,
so
you
will
start
off
as
soon
as
the
installer
is
done.
B
You'll
see
that
there's
at
least
one
machine
set
there
that
you
can
use
to
scale
with
upi
and
in
the
case
of
vsphere,
assuming
the
credentials
and
permissions
are
correct
with
upi.
All
of
that
should
be
there
so
that
you
can
go
in
and
create
your
own
machine
set
day
two
and
it'll
work
exactly
the
same.
B
A
B
The
inverse
is
true
as
well,
so
you
can
manually
create
a
node
and
then
join
it
to
an
ipi
cluster
as
well,
and
have
it
not
be
managed
by
by
machine
api
and
there's
a
couple
of
different
ways.
You
can
do
that.
So
one
is-
and
I
showed
this
a
long
time
ago
I
want
to
say,
like
before
episode,
20
of
like
vsphere
ipi,
I
have
my
cluster.
I
go
and
create
a
new
virtual
machine.
B
You
know
I
and
I
go
through
the
normal
upi
at
a
worker
process
and
it
just
joins
the
cluster,
but
there's
no,
you
know
node
object
associated
with
it.
So
if
we
look
over
here
in
my
or
excuse
me,
no
machine
object
associated
with
it.
If
we
look
inside
of
my
cluster
here,
so
you
can
also
hit
the
create
machine
button
here
and
if
you
fill
this
thing
out
correctly,
it'll
trigger
machine
api
to
go
and
provision
a
node
that
way,
so
you
can
kind
of
do
it
one
one
of
two
ways.
A
Yeah
I
knew
the
inverse
we've
had
some
customers
where,
like
we've
deployed
ipi
and
we're
trying
to
you,
know,
add
some
workers
and
you
know,
do
like
the
upi
method.
Like
that's
always
worked
us.
I
I
just
feel
like
early
on.
Maybe
there's
just
misinformation
going
on
around
about
like
oh
you
know,
if
you
didn't
deploy
with
the
machine
api,
you
can
never
add
them
in.
So
that's
that's
nice
to
know.
C
B
B
B
Where'd
that
other
cluster
go
no,
it's
not
in
this
one
either,
so
you'll
have
a
little
button
up
at
the
top
that
will
deploy
a
terminal
and
you'll
be
a
terminal.
Pod
and
it'll
have
a
little
window
that
you
can
pop
up
here,
and
you
can,
you
know,
do
your
oc
and
and
other
commands
inside
of
there.
So
it's
a
convenient
way
to
interact
with
your
cluster.
If
you're
like
me
and
sometimes
forget
where
things
are
in
the
gui,
let's
see
project
resource
quota
reporting.
B
B
A
B
One
one
answer
or
one
question:
I
don't
know
the
answer
to
is
whether
or
not
you
can
assign
any
persistence
with
that.
So,
like
azure
cloud,
shell
right,
it
provisions
a
volume
that
gets
attached
to
your
cloud.
Shell
instance.
So
that
way
you
can,
you
know,
drop
your
powershell
scripts
or
whatever
in
there
and
they'll
persist
across
sessions.
I
I
don't
know,
maybe
it
does.
Maybe
it
doesn't.
I
have
no
idea.
I
haven't
tried
whether
or
not
that
exists
with
the
web
terminal.
B
B
I
I
learned
just
before
the
stream
that
it
is,
I
think,
a
specific
use
case
or
a
specific
scenario
where
it's
most
appropriate.
So
I
want
to
get
some
more
details
before
I
share
information
on
that
one,
but
if
you're
curious
about
it,
feel
free
to
reach
out
again
andrew.sullivan
redhat
or
practicalandrew
on
twitter,
reddit
et
cetera.
B
B
B
B
Yeah
so
like
this,
they
are
setting
somewhere
down
here.
I
thought,
like
I,
o
threads
for
each
one
of
the
virtual
machines
and
just
various
other
things,
so
it
was
an
interesting
read,
something
that
I
learned
a
couple
of
things
from
so
I
also
like
that
they
did
a
bunch
of
different
testing
to
show
you
like
different
performance,
metrics
things
you
can
expect
with
various
scenarios
if
you
are
interested
in
using
openshift
virtualization
and
how
that
affects
things.
B
So
if
you've
got
any
hard
questions,
any
interesting
stuff
to
talk
about,
be
sure
to
you
can
submit
questions
beforehand,
I
think,
as
soon
as
we
get
the
stream
staged
on
youtube,
you
can
start
chatting,
which
is
weird,
so
we
try
to
check
the
youtube
channels
to
see
if
there's
any
chat
that
has
happened
before
it
starts
see.
If
there's
any
questions,
but
you
know
social,
media
and
email
are
gonna,
be
the
best
way.
A
Nice
yeah
that
that
reference
architecture,
that's
awesome,
I'm
glad
that
they
did
that
because
that's
you
know:
they're
both
huge
products
right,
they're,
both
huge
products
on
openshift
that
people
are
really
interested
in.
So
I'm
glad
that
they
put
something
together
like
that
because
it
kind
of
it
simplifies
it
or
really
it
gives.
It
gives
context
to
things
that
you
know
might
be
abstract
when
you're
looking
at
the
documentation.
A
B
Yeah
speaking
of
johnny's
shirts,
I
won't,
I
won't
state
anything
obvious,
but
I
will
encourage
folks
to
go
to
the
cool
stuff
store
and
check
out
the
openshift
collection.
They
they
added
some
new
products
there
that
are
pretty
cool,
so
I
say
cool
knowing
that
I'm
I'm
not
cool,
not!
I
haven't
been
cool
in
a
long
time,
but
I
think
they're
they're
neat,
I
think
you're
cool,
I
think
you're
cool
thanks.
That's
all
that
counts.
B
B
As
we
said
at
the
beginning,
here,
we've
got
like
the
next
two
plus
months
of
content,
all
planned
out,
and
it's
a
lot
of
really
cool
stuff.
I
you
know
the
summer
has
been
a
little
bit
of
a
doldrums.
I
feel
like
we've
been,
you
know,
sometimes
a
little
last
minute,
stephanie's
laughing
at
me
right
now.
I'm
sure
I
can
laughing,
because
sometimes
we
we've
had
some
things
a
little
last
minute,
but
not
not
for
the
next,
basically
through
the
end
of
the
year,
so
yeah.
B
A
Yeah,
always
man
thank
you
and
yeah
thanks
for
walking
through
and
you
know
doing
the
lion's
share
of
you
know
all
the
demos
and
click-throughs
today
you
know,
thanks
for
all
the
info,
make
sure
you
like
and
subscribe
on,
twitch
or
youtube,
or
whatever
you're
using
and
stephanie.
Thank
you
for
everything,
because
without
you
we
would
we'd
be
exposed
as
unprofessional,
which
is
most
terrible.
So
no,
but
thank
you
and
we'll
see
you
guys
next
week.