►
From YouTube: Kubernetes Office Hours 20190919
Description
Office Hours is a live stream where we answer live questions about Kubernetes from users on the YouTube channel. Office hours are a regularly scheduled meeting where people can bring topics to discuss with the greater community. They are great for answering questions, getting feedback on how you’re using Kubernetes, or to just passively learn by following along
https://contributor.kubernetes.io/events/office-hours
A
Okay
and
we're
live,
welcome
everybody
to
today's
kubernetes
office
hours.
It
is
September
19,
2008
een.
This
is
the
first
of
two
sessions.
We
have
another
one
in
about
six
hours
to
catch
the
rest
of
the
United
States.
So
what
is
this?
This
is
a
live
stream
where
we
answer
your
user
questions
concerning
kubernetes.
How
do
I
use
turbine
at
ease?
How
do
I
set
up
kubernetes
that
are
configured
kubernetes?
All
that
is
fair
play.
You
can
find
us
in
cash
office
hours
on
slack
and
we'll
be
keeping
an
eye
on
kubernetes
users
as
well.
B
C
D
F
A
All
right,
I'm,
Jorge,
Castro,
we're
gonna
have
to
as
a
community
manager
and
I'm
kind
of
the
co-host
here
with
Baba
and
I'm
streaming
and
I
work
as
partisan
contributor
experience,
so
I
kind
of
have
my
hand
in
everything.
So
the
way
this
is
gonna
work,
let's
get
some
ground
rules
here.
So
first
of
all
keep
feel
free
to
keep
asking
your
questions
in
the
hash
office
hours
channel.
You
don't
have
to
wait
just
post
it
in
there
and
what
we
do.
We
have
a
link
there
to
the
hack,
MD
document.
A
That's
what
we'll
keep
notes
and
stuff
we'll
put
your
questions
in
there.
So
we
organize
them.
You
know,
tuck
answer
them
in
order
that
kind
of
thing
and
then
at
the
end
we're
gonna
have
a
raffle,
and
if
we
ask
your
question
online,
you
have
a
chance
to
win
the
highly
coveted
kubernetes
t-shirt,
not
this
one,
the
other
kubernetes
t-shirt.
We
continue
to
fail
to
actually
wear
the
t-shirt
that
we
give
away
during
the
show,
so
some
quick
ground
rules
here,
a
judgment-free
zone.
A
So,
if
you're
in
the
channel-
and
you
see
people
asking
what
seemed
like
basic
questions,
that's
literally,
why
we're
here
so
make
sure
you're
following
the
kubernetes
code
of
conduct
and
treating
everyone
professionally,
because
there
is
no
one
expert.
So
that's
that's
part
of
the
reason
that
we
have
the
shows
to
give
everyone
a
safe
space
where
they
can
not
get
judged
and
then
just
helping
them
at
work
getting
their
stuff
set
up.
We
will
do
your
best
to
answer
your
questions.
However.
A
Please
understand
we
don't
have
SSH
access
to
your
cluster
distributed
systems
are
complicated,
so
we
might
not
be
able
to
give
you
the
specific
answer
you
want,
but
at
a
bare
minimum
either
send
you
to
the
right
sig
to
get
more
information
or
perhaps
give
you
the
right
piece
of
Docs
or
the
right
information
that
you
need
to
troubleshoot.
So
we'll
do
the
best
we
can
but
yeah.
We
can't
like
SSH
into
your
control,
plane
and
figure
out,
while
your
cubelets
broken
panelists
you're
encouraged
to
expound
on
your
answers,
given
your
expertise.
A
So
it's
not
just
about
giving
some
of
the
docs.
It's
also
about
sharing
any
knowledge
that
you
might
have
on
on
any
implementation
that
you
might
might
done
in
production.
So
please
do
that.
So,
please,
post
your
questions
in
the
channel.
We
also
have
a
thread
on
discussed
at
kubernetes
I/o,
which
I
will
post.
If
you
have
any
written
up,
question
or
Stack
Overflow
link
that
you've
pre
written
or
something
for
you
to
plot
those
in
there
and
we
will
we
will
get
to
them.
A
You
can
also
help
us
out
by
tweeting
spreading
the
word
and
paying
it
forward.
So
if
we
answer
when
your
questions
or
if
your
say
any
value
at
us
doing,
is
we
always
love
getting
tons
of
feedback,
even
if,
even
if
we
helped
you
on
the
littlest
thing,
it
really
helps
us
out,
because
we
keep
a
running
tag
on
that
and
then
we
use
that
to
help
make
the
show
better.
A
At
some
point,
we
keep
saying
we're
gonna
do
this
I
would
like
to
have
topical
ones
where
it's
like
here's,
an
entire
office
hours
on
just
storage,
just
things
like
that,
but
we
need
feedback
from
you,
the
audience
on.
What's
working
for
you
and
what
isn't
each
session
is
recorded
and
available
on
YouTube
there's
a
playlist
there,
the
kubernetes
channel.
We
have
an
entire
backlog
of
information
going
back
almost
a
year
now,
I
think
so.
There's
a
ton
of
expertise
there,
so
you're
you're
welcome
to
browse
that.
A
And,
lastly,
if
you
want
to
sit
on
this
panel,
these
are
all
volunteers.
So,
if
you've
had
some
experience,
deploying
kubernetes
and
feel
like
you
want
to
share,
the
commitment
is
one
hour
a
month,
which
is
not
that
bad
and
even
then
you
have
to
show
up
every
month
or
so
we
haven't.
We
usually
get
enough
volunteers
to
show
the
panel.
So
if
you're
feeling
like
paying
it
forward,
you
can
always
do
that.
A
We're
also
always
looking
for
marketing
help
so
help
us
spread
the
word
and
make
sure
you
hang
out
towards
the
end
is
when
we
do
the
raffle.
So
let's
begin,
let's
kick
on
over
to
the
hack
and
deed
notes
and
see
what
the
first
question
is:
am
I
just
going
in
order
here
top
top
to
bottom.
Alright,
first
question
comes
from
Aaron
in
the
officers
channel.
A
Welcome
Aaron
says
we're
currently
operating
a
cube,
a
big
cluster
with
one
master
at
the
moment,
and
we
need
to
migrate
the
one
master
machine
to
a
new
machine
as
it's
running
on
a
machine
with
very
poor
resources.
Is
it
possible
any
advice
on
how
to
do
it,
eventually
we're
going
to
migrate
to
using
multi
multi
master,
but
we
don't
really
have
the
infrastructure
for
it
at
the
moment.
So.
C
Have
a
stop
so
I
haven't
used
Q
B
IDM,
just
as
a
so
I'm,
not
entirely
sure
on
that,
but
sort
of
generic
kubernetes.
If
you
want
to
do
this,
if
you've
got
one
master
I'm
guessing,
it's
got
HCG
on
that
master,
node
you're
going
to
need
to
bring
up
another
load
and
join
those
two
XE
DS
together.
Now,
there's
obviously
documentation
on
the
SD
website.
C
For
that
that's
quite
dangerous
though,
because
then
you've
got
to,
and
so,
if
you
get
bit
brain
like
you're,
just
gonna
be
in
a
bit
of
a
difficult
place,
so
try
and
make
sure
you
migrate
those
quickly.
But
if
you
do
get
CDs,
then
the
API
server
on
the
new
node
and
the
API
server
on
the
old
node
will
be
serving
the
same
data.
So
that's
perfectly
safe.
The
controller
manager
and
the
scheduler
should
in
theory,
do
the
leader
election.
So
that
should
also
be
safe.
It's
the
main
thing.
C
A
C
A
So
Aaron
hope
hope
that
answers
your
question
feel
free
to
stick
around
and
ask
some
follow-ups
and
details.
I'm
I'm,
almost
feeling
bummer
for
not
checking
the
latest
cube
admin
releases.
As
far
as
how
this
works,
cuz
I'm,
pretty
sure
you
could
just
add
two
more
right
provision
to
machines,
add
two
more
wait
and
then
mark
the
one
that
you
want
to
get
rid
of.
D
A
So
I
do
have
a
follow
up
because
I
feel
like
a
natural
question
here
and
Aaron
I'm
gonna
ask
a
follow-up
for
you
here
running
at
CD
and
the
control
plane
on
the
same
machines
like
I.
What
what
size
do
you
get
to
the
point
where
you
have
to
have
like
dedicated
at
CD
machines
or
do
most
people
just
the
control
planes,
the
control
plane.
C
In
our
environment,
we
have
separate
EDD,
also
scaling
groups,
so
we
tend
to
have
5x
CD
nodes
in
each
of
our
environments.
They
manage
our
staff.
If
we
want
to
replace
the
spindle
node
there,
we
can
just
kill
that
it's
just
got
a
CD
on.
We've
actually
got
a
script
that
came
from
Monsanto,
something
like
that.
C
That
means
that
you
can
just
kill
a
node
and
when
it
comes
back
up
it
fixes
all
the
missing
stuff,
like
the
movie
all
know,
joins
itself
to
the
cluster,
that's
quite
cool,
and
then
you
had
the
controller
playing
control
playing
the
committee's
like
scheduler,
API
server,
controller
manager.
You
don't
well,
we
haven't
seen
the
need
for
Moreland's
or
two
of
these
in
any
bar
clusters,
so
far
we're
running
them
on
C
for
large
and
AWS,
and
they
just
seem
to
cope
with
sort
of
40
nodes.
So.
E
C
A
A
A
D
C
A
E
A
A
All
right,
so
we
don't
know
this
one,
so
hopefully,
hash
gke
can
help
you
out
and
I
will
try
on
the
when
we
get
the
West
Coast
once
hopefully
we'll
get
some
Googlers
in
here.
That
might
be
able
to
help
you
out.
So
that
seems
to
be
the
best
answer.
Real
quick
before
we
move
on
when
Messam
was
talking
about
configuring,
kubernetes
masters
to
forward
audit
log
into
an
internal
service,
he
goes.
A
I
was
referring
to
the
web
hood
back-end
mention
here
in
the
docs,
slash
debug
application
cluster
slash
audit
web
hook
back-end,
but
the
documentation
does
not
mention
how
to
configure
the
forwarding
properly.
My
use
case
is
that
I
need
a
forwarded
to
an
internal
service,
a
checks
for
certain
events
to
send
them
to
a
slack
channel
in
my
organization.
B
A
So
do
y'all
see
the
link
there
to
the
web
hook.
Back-End
all
right
just
give
us
a
minute
there
to
digest
for
those
of
you
just
tuning
in
thanks
for
listening
on
YouTube.
This
is
the
kubernetes
office
hours.
Well,
we'll
address
your
questions
in
the
hash
office.
Hours
channel
on
the
kubernetes
slack
so
feel
free
to
just
whack
your
question
in
there
or
you
can
tweet
them
to
us,
just
tweet
them
to
act
kubernetes
not
actually
at
kubernetes
I/o
or
at
castro
Jo
and
I'm,
just
following
those
there.
E
Any
help
for
him
here
on
this
one
so
about
the
audit
webhook
I,
assume
it
works.
Like
the
admission
web
hooks,
you
would
have
to
create
a
controller
register.
You'll
control
in
the
API
generator
certificate
for
your
controller,
so
the
cube
API
will
call
on
TLS
endpoint
from
there.
You
will
probably
get
JSON
data
that
you
can
forward
to
really
if
real
big
you
are
using
should
be
Jason
so
from
there.
A
All
right
moving
on,
let's
see,
Laith
welcome
to
office
hours
s,
hi
I've,
had
reports
of
masternodes
becoming
unresponsive
on
our
cluster
and
the
Cates
API
becoming
unavailable
requiring
manual
intervention
should
masternodes
reliably
self-heal
are
there
use
cases
or
a
single
master
node
can
become
stuck
in
an
in
risk
unresponsive
status.
So
do
you
suggest
a
particular
way
to
monitor
these
nodes,
so
I
would
suppose
whether
I
could
heal
or
not
would
depend
on
what
the
problem
is
right.
C
You
there's
a
useful
tool
actually
from
aprender
called
Kiba
rang
that
you
might
be
able
to
set
up.
So
we
have
this
running
in
a
cluster
and
what
it
does
the
schedules
pods
period
you
can
set
up
scheduled
pods
periodically
and
it
tests
a
bunch
of
the
like
or
communities
or
operation
of
the
control
plane.
E
C
You
are
looking
to
monitor
the
kinetics
control
plane.
The
core
OS
Prometheus
operates,
has
a
bunch
of
really
good
alerts
that
they've
set
up
for
the
API
server
control
manager,
scheduler
that
kind
of
stuff,
and
so
I
would
recommend
going
through
the
in
the
definitions
that
they've
got
in
that
repository.
A
B
A
B
A
Yeah,
so
if
we
get
to
follow
up
on
what
exactly
you're
having
to
fix,
it
would
probably
help
us
out
to
help
you
but
we'll
circle
back
around
to
you.
Here
we
have
a
question
from
Patrick
says:
hey
there,
I
want
to
scrape
endpoints
of
pods,
which
are
part
of
a
deployment
all
I
can
think
of
is
using
cute
cuddle,
get
deployment
deployment,
Oh
JSON
to
get
the
IP
port.
Is
there
a
better
way
to
do
this,
or
maybe
I
should
rethink
how
my
data
flows.
A
A
A
E
A
So,
generally
speaking,
is
you
do
you
recommend
if
you
find
yourself
cute
cuddling,
a
certain
thing
over
and
over
again
at
that
point,
it's
like
my
tooling
needs
to
talk
to
the
API
directly.
Is
there
like
I
sort
like
a
rule
for
this?
What
do
people
do?
Do
they
generally
keep
a
bunch
of
scripts
that
are
cute
cuddle
stuff,
or
do
they
at
some
point,
make
like
a
judgment,
call
and
say?
Okay?
At
this
point,
we
need
to
formalize
pinging.
B
The
API
server
looks
like
it's
gonna,
be
programmatic,
I
can
import
the
Korea's
lives
into
my
app
mm-hm
and
the
other
thing
you
can
actually
do,
and
this
is
how
sort
of
service
discovery
happens
and
stateful
sets.
Is
you
can
create
a
headless
service
that
targeted
deployment
with
a
selector?
And
then,
when
you
get
that
service,
it
will
literally
just
spit
back
the
Potter
keys
for
all
the
end
points.
I
A
Do
this
sort
of
thing
yeah
that
looks
like
probably
what's
he
need?
It
looks
like
he's
typing
furiously
bunch
of
people
typing
furiously
all
right.
Patrick
we'll
give
you
a
second
to
collect
your
thoughts,
kind
of
think
about
what
they
said
and
then
ask
you
know,
keep
giving
us
information
and
we'll
get
we'll
get
back
to
you
after
this
next
question.
Those
of
you
that
are
joining
us,
we're
answering
questions,
live
on
hash
office
hours.
Oh,
it
sounds
like
this
question
is
answered.
Bingo
stick
around
Patrick
for
your
chance
to
win
a
kubernetes
t-shirt.
A
B
C
A
A
And
Dean,
if
you're
listening,
this
feels
like
one
of
those
things
that
if
we
knew
your
use
case,
we
might
be
able
to
give
you
a
best
practice
on
on
why
you
wanted
to
do
it
that
way,
and
maybe
maybe
there's
a
better
way
to
do
that.
Also
Patrick.
We
like
to
remind
us,
there's
a
10-second
delay
from
only
talking
so
give
him
a
chance
to
type
more,
but
it
sounds
like
we
answered
his
questions.
A
Lathe
has
some
follow-up
I
will
get
to
you
in
a
second
lathe.
Please
stand
by
Isaac
asked
how
could
I
list
my
installed?
Flex
volume
drivers
stand
allocation
should
be
a
user
live
execs
like
uber
Nettie's
such
cubelet.
That's
plug
in
slash
volume,
/
exec,
but
that
directory
does
not
exist.
Meanwhile,
I
know
we
use
NetApp
flex
volumes,
but
I'm
not
sure
about
the
driver.
Names
I
need
to
know
so.
I
can
put
it
into
a
pod
security
policy.
G
Sure
my
name
is
Matt
I'm
starting
a
new
role
at
the
beginning
of
October
I
was
at
Red
Hat
for
a
number
of
years
and
a
bunch
of
places
before
that
working
on
OpenStack
and
I'm
gonna
be
helping
a
startup.
It's
not
so
small
anymore
break
up
a
monolith
and
migrate
from
legacy
a
day
we
have
stuff
to
kubernetes
for
their
micro
services,
so
I'm
their
new
cloud,
architects,
so
I'm
just
kind
of
yeah.
J
A
Let's
see
these
someone's
are
coming
from
kubernetes
users
right,
not
the
channel
right.
Okay,
so
I'm
going
to
prioritize
the
questions
in
the
office
hours
channels,
those
they're
actually
hanging
out.
Let's
see
Laith
Astley,
he
goes
okay,
all
right.
Let's
coming
back
to
this
one
is
the
one
with
the
master
nodes
becoming
unresponsive
and
leaf,
has
to
go
in
and
do
a
manual
intervention,
and
he
says
shouldn't
they
Auto
heal
and
then
we
said
well,
it
would
help
if
we
figure
out
what
what
you
were
fixing.
So
he
comes
back.
He
goes.
A
The
incidents
happened
before
I
joined,
so
I
don't
have
much
information.
All
I
know
is
that
they
stopped
the
node
and
it
came
back
up
the
folks
didn't
the
I.
Think
that
should
be
IT
guys
didn't
investigate
the
issue.
I
appreciate
you
to
have
enough
information
to
work
with
I'll
have
a
look
at
the
tools
you
suggested
in
the
meantime.
I
feel
like
I.
A
Don't
know
like
I
feel
like.
If
you
can,
we
talk
about
some
best
practices
when
you
perhaps
show
up-
and
you
had
someone
else
set
up
a
cluster-
maybe
to
talk
about
some
of
the
due
diligence
that
you
would
do,
maybe
give
you
you
know,
maybe
even
give
you
some
more
information
as
far
as
what
you
would
do.
Cuz
I
gotta
feel
like
the
first
thing
I
would
do
is
like
run
like
Cooper
Aang.
Let's
see
what.
A
What
the
sonobuoy
results
are,
so
that's
a
oh
we've
got
the
notes
Matt.
We
have
the
notes
in
in
this
hack
and
e-file.
If
you
want
to
follow
along
kind
of
see
what
the
sonoran
blue
results
are,
and
that
will
kind
of
tell
you.
Oh
no
there's
like
a
Narvik
I
just
have
this
feeling
that
everyone's
our
back
is
always
messed
up.
It's
probably
like
the
first
given.
B
A
A
A
Some
people
do
do
that
is
there
anything
else?
Is
there
like
a
checklist
of
stuff
I.
C
Think
I'll
make
sure
I
know
where
the
XE
d
in
this
or
control
plane
components
are
sort
of
set
up,
like
some
people
have
them
static.
Pods
some
people
have
them
as
system
D
units,
then
obviously
there's
the
configuration
for
that
as
well
like.
If
there's
any
like
special
options,
you
might
need
to
be
aware
of
when
you're
debugging,
that's
kind
of
stuff
yeah.
A
C
We've
had
money
for
few
months
it
so
we've
got
a
using
prom
run,
and
so
we
can
explore
previous
metric
of
whether
it
passed
or
not.
We
can
over
on
that.
So
like
we
get
paged,
sometimes
when
it's
like
key
bearing
failed,
it's
like
okay,
the
control
planes
broken
somehow
doesn't
get
out
too
much
detail
from
that.
But
when
you
actually
look
at
the
web
page,
it
gives
you
it's
like
okay,
pod
student
schedule
on
this
node.
Is
there
a
problem
with
no
just
a
scheduler?
You
can
slow
it
down.
A
bit
has.
C
A
That's
probably
a
good
thing:
yeah
I,
don't
know
I'm
hoping
it
catches
bad
stuff.
Hopefully
it's
all
boring
and
never
returns
anything
all
right.
Anybody
else
have
any
follow-up
questions
before
I
move
to
kubernetes
users,
questions
I'm,
gonna,
go
ahead
and
I'm
just
gonna
repost
again
talks
amongst
yourselves
for
a
second
I'm
just
going
to
repose
in
the
kubernetes
users
channel
that
we
are
live.
C
I
think
is
a
historic
thing:
isn't
it
because
Cube
control
used
to
do
a
lot
of
validation
of
objects
before
it
send
them
to
the
API
server,
whereas
I
think
now,
with
the
newer
versions
of
communities
it's
now
sending
directly
to
API,
so
many
API,
so
I
was
doing
the
validation
of
the
schema.
I,
don't
know
if
you
can
actually
validate
false
anymore.
C
I,
remember
them:
sort
of
they've
basically
been
moving
a
lot
of
these
sort,
client
side
stuff
from
keep
control,
taking
it
into
the
API
server
itself.
Things
like
dry,
bronze,
moving
into
the
API
server
in
112.
A
lot
of
those
flags
are
going
to
become
just
sort
of
Robin,
actually
Q
control
logic,
a
query
parameter
on
the
API,
so
I,
don't
think
I,
don't
know
that
they
will
do
anything
anymore
because
I
would
assume
you'd
want
to
validate
it
broke
out
into
xcd
anyway,.
G
A
Awesome
if
c
comes
back
so
sorry,
I
got
dragged
away,
no
worries,
buddy,
Matt,
gke
and
I
basically
said
we're.
Stumped
probably
has
gke
might
be
a
place
for
that
one
and
specifically,
all
right.
I'm
gonna
move
on
to
some
kubernetes
users
questions.
Those
of
you
that
are
just
joining
us
feel
free
to
drop
your
question
and
hash
office
hours
in
the
kubernetes
slack,
and
we
will
get
to
it
very
chance
to
win
a
t-shirt.
A
All
right,
Derrick
ass,
not
sure
if
it's
a
kubernetes
issues
or
a
docker
issue,
I've
run
a
kubernetes
cluster
and
a
few
bare-metal
servers
for
several
weeks.
But
recently
the
master
components
will
kill
themselves
and
even
the
related
docker
images
get
removed,
tried
to
Rhian
it
the
cluster
with
Q
Batman,
but
the
problem
still
the
same.
Anyone
seen
this
kind
of
problem
before.
Thank
you
very
much.
I.
A
A
H
A
Yeah
Derek,
that
sounds
very
not
not
normal,
especially
of
like
the
local
images
get
removed.
It
sounds
like
there's
a
job
or
some
kind
of
automation,
somewhere,
that's
showing
up
and
killing
stuff,
but
that's
where
I
would
start
looking
long
says,
I
have
a
very
embarrassing
question.
Don't
worry,
there
are
no
embarrassing
questions.
Moving
on
the
Ahmed
asks
hello,
I
heard
about
vertical
scaling
in
kubernetes.
I
just
want
to
make
sure
I
got
it
right.
A
Assuming
I'm
running
Nettie's
on
a
cloud
provider
can
I
just
give
it
a
job
requiring
say,
12,
gigabytes
to
run
and
kubernetes
will
auto
provision
this
machine
and
run
the
job
on
it.
How
magic
is
this
I'm,
adding
that
last
part?
How
magic
is
this
and
everyone's
shaking
their
head?
No!
Ok!
So
how
does
this?
How
does
this
actually
work,
then,
with
the
vertical
scaling.
C
The
first
component,
so
scalar
sort
of
monitors
how
much
memory
and
CPU
that
our
pod
is
using
over
time
and
then,
when
the
pod
gets
restarted,
it
can
mutate
the
pod
to
actually
put
more
accurate
request
in.
So
if
you
keep
getting
boom
killed,
for
instance,
it
will
start
upping
the
memory
request
rather
than
you
just
having
the
same
static
requests.
It
can
also
over
time
if
it
notices
that
you're
under
bridging
kill,
pods
and
restart
them
and
sort
of
set
them
lower.
It's
not
for
nodes.
C
So
if
you
want
to
bring
up
new
nodes
when
you're
running
out
of
resource,
you
need
to
custom
alter
scalar,
but
that
doesn't
have
any
sort
of
sense
of
what
size
know.
Do
you
need
you
have
to
give
that
a
bunch
of
like
waterskiing
groups
would
be
like
choose
from
are
for
larges
or
awful
trigger
next
largest,
or
something
like
that,
and
it's
more
for
pods
than
for
the
cluster
itself.
E
There
is
a
proposition
to
to
make
the
vertical
autoscaler
be
able
to
mutate
the
limit
without
killing
the
pods.
That
would
be
more
useful
that
what
it
does
today,
because
when
he
needs
to
mutate,
something
it
has
killed
upon,
restart
the
pod
right.
So
let's
say
you
want
to
increase
the
memory,
have
to
kill
your
service
you'll
get
down
pain
because
of
that
once
it
can
mutate
the
limits
without
restarting
reported
per
your
zone,
but
increasing
the
vertical
scaling.
The
node.
That's,
not
not
kubernetes
issue
and
that's
a
cloud
provider
problem.
A
Yeah,
almost
too
magical
I
don't
know,
I
don't
want
things
that
affect
my
bill
to
be
too
magical,
I
guess
I,
don't
know.
That's
just
me.
Alright.
Moving
on
a
long
ass,
welcome
back
long
and
I
know,
you've
showed
up
to
previous
sessions
before
it
goes
for
people
whose
nodes
use
DHCP.
Is
there
an
easy
way
to
deal
with
auto
scaling
group
creating
/
deleting
nodes,
often
or
the
existing
node
changing
the
IP
address?
I
here
certificates
going
invalid
after
IP
address
has
changed
at
least
once
per
week.
B
A
A
B
E
Yeah,
there
are
several
reasons
one
is:
you
can
set
a
time
of
the
day
when
you
want
Jackie
to
automatically
update
your
nodes,
so
it
should
be
careful
setting
that.
Also,
if
you
are
using
the
pre
editable
the
ends,
those
hands
will
get
killed
every
time
for
hours
and
you
cannot
control
that
they
can
also
get
killed
faster,
but
I
would
say.
E
A
E
If
you
use
the
fateful
served
with
PVC
or
something
like
that
right
code,
your
app
when
it
starts
up
to
resume
from
disk
right
right,
even
if
the
note
goes
away
when
the
note
gets
replaced,
that
volume
gets
reattached,
creates
on
your
board.
So
you
can
reinstate
from
there
or
use
something
like
a
bucket
or
whatever
track
you're
saying
yeah.
There
is
no
way
to
avoid
the
the
maintenance
yeah
any.
H
A
What
I'm,
trying
to
get
you
to
say
is:
don't
don't
bet
on
hoping
that
a
node
won't
restart
I
guess
what
I
was?
Can
you
all
tell
I
just
read
the
kubernetes
fundamentals
book
I'm
like
okay
I
need
to
do
that,
but
all
right
good
to
know.
Let's
see,
I
see
people
typing
in
the
channel.
If
you
have
questions
you
feel
free
to
keep
them
going.
We
are
wrapping
up
here
in
about
10
minutes
or
so
so
no
worries.
A
B
A
So
you're
kind
of
using
off
of
these
and
familiar
with
all
of
these.
What's,
let's
say
let's
say
I
was
setting
this
up
today.
What
am
I
leaning
towards
because
I
know
one
came
from
coral
last
ones,
red
hat
and
then
I
had
back
core
OS
like
I.
They
are
they
merging
those
things.
What's
the
TLDR
which
one
you
know
we
still
waiting
to
see?
What
happens?
Are
they
both
going
to
use?
They
maintains.
B
I
think
we're
still
waiting
to
see
what
happens
and
I'm
like
I
can
at
least
speak
to
keep
oak
as
that's
what
we
use
internally
and
that's
been
rock
solid
for
us.
That's
mostly
because
our
LDAP
structure
is
kind
of
wack
being
a
very,
very
large
academic
institution.
With
many
years
of
history,
shoved
into
Elda
that
shouldn't
be
shoved
in
there.
I
believe.
A
Okay,
so
key
cloak,
that's
lots
of
options
still
out
there.
Hopefully
that
will
enter
your
question.
Ad
ad
fan
asks
very
novice
question,
but
I'm
trying
to
stand
up
a
mini
cube
cluster
inside
on
a
Bluetooth
VM
for
development,
but
the
cube
DNS
container
doesn't
seem
to
be
able
to
reach
the
Cates
API
server.
It's
reaching
out
to
the
correct
IP
import
of
the
kubernetes
service
and
the
service
is
correctly
pointing
to
the
API
server.
Any
ideas
on
where
I
should
be
looking
next
I.
E
B
A
A
D
C
D
G
A
Having
your
own
bare
metal
sounds
expensive,
but
then,
when
you
start
to
add
up
all
the
cloud
things
so
I
bought
a
nook
and
I've
been
doing
it.
A
blue
2
BM.
We
just
rock
you
bad
man,
similar
Stefan,
but
the
openSUSE
folks
have
been
lately
integrating
cube
admin
and
supporting
it
and
out
of
the
box,
so
I'm
actually
trying
their
mini
OS
thing,
so
Richard
Brown,
if
you're
listening,
I'm
actually
using
it
and
playing
it
other
than
for
streaming.
A
C
E
A
An
AWS
cluster
in
which
I
needed
to
schedule
pods
on
the
masternode
I,
have
had
a
taste
of
the
master
and
was
able
to
run
a
deployment
with
some
pods
of
master.
But
when
I
created
an
Associated
service,
AWS
doesn't
add
the
master
instance
to
the
ELB
I'm
missing
something
besides
the
team.
So
it
sounds
like
stops
getting
scheduled
so.
C
A
C
Guess
you
have
so
have
you
got
moths
if
they've
got
master
nodes
in
a
separate,
auto
scaling
group?
You
need
to
make
sure
that
auto
scaling
group
is
attached
to
the
EOB
and
by
default
I.
Don't
think
you'll
do
that
Norman.
So
we've
got
asked
the
Oh
beans
in
front
of
masters
and
workers,
and
you
have
to
manually
add
those
attachments
to
also
scaling
groups.
I've
not
used
the
service
type
where
it
goes
and
creates
load.
A
Alright
gongura,
please
feel
free
to
give
us
a
follow-up,
hopefully
that
will
that
will
help
you
and
he
passes
along
I,
wanted
to
thanks
for
these
sessions.
Guys
I
really
appreciate
your
time.
Thank
you
for
listening.
Patrick
wants
to
add,
I
use,
cube
AB
in
those
which
have
many
cubes
I
switch
back
to
cube
admin
with
vagrant,
because
I
can
do
multi.
You
know
things
so
that's
one
thing
I
like
about
about
cube
admin
is
I,
could
just
add
machines,
and
it's
like
a
real
thing.
It's
not
like.
A
C
Just
distracted
by
dad
they
so
I
haven't
seen
any
traction
on
that
in
a
while.
I've
got
couple
of
pubes,
he
always
open
to
it,
but
I
didn't
see
much
activity
and
then
there's
a
new
release
like
a
week
ago.
I
don't
know
who's,
maintaining
it
now
so
I
know
the
original
maintainer
has
left
chorus,
Erik
Chang,
so
I
don't
know,
what's
happening
with
our
product,
I'm
hoping
it
carries
on,
but
to
change.
A
A
And
the
last
chance,
alright,
do
you
have
the
do?
You
have
the
t-shirt
names
a
chance,
a
lot?
Yes,
I
do
it's!
Alright,
we
are
going
to
do
the
rat.
Let
me
do
the
outro
first
then
we'll
do
this
we'll
do
the
raffle
so
go
and
just
give
us
a
second
here
to
ran
this
totally
guaranteed
to
be
random
script.
I
do.
A
Me
ask
Jeff
to
get
as
a
result.
Unfortunately,
Jeff
Jeff's
getting
housework
done,
so
he
is
not
joining
us
in
the
meantime,
I'll
start
with
the
outro
sand.
Is
this
loads,
so
real
quick?
While
the
raffle
goes
thanks
to
the
following
companies
for
supporting
the
community
with
these
developer
volunteers?
These
sessions
are
run
by
volunteers,
your
giant
swarm,
hefty
o
stock
X
packet
net
pusher,
calm,
red
hat,
weave
works,
VMware,
zing,
hope,
I
got
that
right
and
the
University
English,
let's
see,
remember,
clack
UConn
is
coming
up
here
in
December.
A
We
should
be
hearing
about
the
sessions
here
pretty
soon
actually
and
for
those
of
you
that
are
contributing.
The
election
will
start
today,
probably
when
I
send
the
email
after
this,
and
with
that
thanks
Erin,
says
big.
Thanks
to
the
office
hours
panel,
very
helpful
gonna
give
all
these
options
go
thanks.
Erin
we
do
this
a
third
Wednesday
of
every
month,
just
keep
on
following
on
Twitter
and
stuff,
and
then
I
hop
into
the
users
Channel
every
once
in
awhile,
and
remind
everybody
that
we
exist
and
the
winner
of
the
t-shirt
is.
B
A
Hold
on,
let
me
click
and
see
if
you
have
a
real
name
Marcel,
you
have
won
a
t-shirt,
so
I
will
PM
you
after
this
and
give
you
the
code
for
the
CNC
app
store.
Oh
wait:
Bob's
wearing
the
shirt
stand
up
mom.
This
is
the
t-shirt
that
we
get.
We
give
away,
give
away
a
teacher,
every
recession
that
we
have
and
with
that.
Thank
you
very
much
and
we'll
see
everybody
out
and
enjoy
kubernetes.