►
From YouTube: Kubernetes Office Hours 20190828
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.
Discussion thread for this episode: https://discuss.kubernetes.io/t/office-hours-for-28-august/7559/3
For more info: https://github.com/kubernetes/community/blob/master/events/office-hours.md
Liking what you see here? Continue the discussion at https://discuss.k8s.io
A
And
welcome
everybody.
It
is
Wednesday.
That
means
it's
time
for
the
kubernetes
office
hours,
the
show
where
we
live
stream
on
the
Internet
and
try
to
answer
as
many
of
your
kubernetes
questions
as
quickly
and
as
thoroughly
as
possible.
My
name
is
Jorge
Castro
special
edition
here
today.
Normally
we
go
on
the
third
Wednesday
of
every
month,
but
a
bunch
of
us
were
traveling,
so
we
are
going
today
on
the
28th.
Thank
you
to
those
of
you
joining
us.
We're
gonna,
do
quick,
intros
and
then
kind
of
let
you
know
how
this
get
started.
B
D
Hi
Joel
I
work
for
push
around
London
I
have
recently
been
working
on
CI,
so
we
can't
even
see
prowl
here,
ditchthecube
nessa
CI,
if
you
haven't
heard
of
it,
I've
done
a
lot
of
stuff
for
controllers
in
the
past.
So
if
you
got
any
control
our
questions
controller
on
time
queue
builder
source
stuff,
send
them
my
way.
Also
done
security
in
that
kind
of
stuff
impossible.
E
A
George
Kessler
I'm
a
community
manager
for
kubernetes
at
VMware
and
I'm
kind
of
the
emcee
organizer
of
this
little
shindig
here
all
right.
So
here's
how
it
works.
Welcome
everybody
we're
glad
you
you've
joined
us
here
today
we
are
live
streaming
on
YouTube
how's,
the
audio
sound
feel
free
to
type
into
the
slack,
where
you're
from
say
hello.
That
kind
of
thing
we
like
to
see
things
scrolling
in
the
channel.
It
motivates
us
to
put
on
a
great
show
for
you
before
we
start
here.
The
ground
rules
judgement
free
zone.
A
Everyone
had
to
surf
from
somewhere.
There
are
no
dumb
questions
so
feel
free
to
ask
whatever
question
is
that
you're
stuck
on
right
now
and
we
will
do
our
best
to
help
you
out
and
well?
We
will
do
our
best
to
answer
your
questions.
We
don't
have
access
to
your
cluster,
so
live
debugging
is
kind
of
off
topic.
We're
not
gonna,
be
able
to
like
SSH
into
your
thing
and
figure
out
where
your
control
plane
is
broken
or
whatever,
but
we
will
do
our
best.
A
However,
to
at
least
let
you
know
where
to
look
to
at
least
get
you
moving
in
in
a
place
if
you're
stuck
so
please
be
cognizant
of
that
panelist
you're
encouraged
to
expand
on
your
answers
with
their
experiences
and
pro
tips.
We
definitely
want
to
spread
from
information
from
people
who
are
running
kubernetes
in
production
audience.
You
can
help
us
out
by
pasting
URLs
to
official
documentation,
blogs
projects
at
some
point,
someone's
gonna
ask
hey
what
tool
do
I
use
for
this
feel
free
to
just
whack,
all
those
URLs
into
slack.
A
We
do
stream
the
slack
Channel
here
on
this
here.
It
is
on
the
side
during
the
broadcast
and
what
I
do
is
at
the
end,
I
grab
all
the
URLs
and
publish
them
in
the
YouTube
description
as
a
sort
of
show
notes.
So
if
you
see
us
mention
something-
and
you
can't
quite
remember
or
remember
the
name
or
whatever
we'll
always
publish
the
URLs
at
the
end,
so
feel
free
to
post
your
questions
in
this
channel
and
if
not,
we
have
a
thread
on
discuss
kubernetes,
which
I
will
post
here
in
a
minute.
A
Now
that
I
don't
have
Bob
I,
don't
have
anyone
to
post
here.
If
you
have
a
long-form
question
that
you
want
to
toss
into
this
thread
here
or
give
us
a
link
to
a
stack
overflow
question
or
something
that
you
might
be
stuck
on,
we'd
be
happy
to
look
at
those.
You
can
also
help
us
out
by
tweeting
spreading
the
word
paying
it
forward.
Stuff
like
that.
So
with
that
we
are
about
to
get
started
before
we
get
started.
A
Just
a
quick
thanks
out
the
giant
swarm
stock,
X
packet,
pusher
com,
we'd
works,
VMware
lifts
the
University
of
Michigan
Utility
Warehouse
and
the
City
of
Ottawa
Ontario
Canada
for
donating
engine,
their
engineering
time
to
have
these
folks
here,
help
everyone
out
and,
of
course,
a
special
thanks
to
the
CCF
for
sponsoring
the
t-shirt
giveaway.
So
if
we
ask
your
question
on
the
air,
we
go
away
to
t-shirts
at
the
end
of
the
show,
so
ask
a
questions.
A
Stick
around
with
the
little
raffle
is
guaranteed
to
be
random,
not
really,
and
we
will
give
out
as
many
kubernetes
t-shirts
as
we
can,
because
we
only
have
three
more
of
these.
Before
cube
cons,
so
with
that
panel,
how
are
we
doing?
How
do
we
sound
out
there
audience
feel
free
to
start
typing
in
your
questions?
Ooh
looks
like
Miguel's
discovered
prowl,
so
we're
gonna
talk
about
that
in
a
little
bit
all
right,
we're
gonna
start
with
the
question.
A
Cue
audience
feel
free
to
just
keep
slapping
the
questions
in
the
channel
and
we'll
go
all
right.
Y'all
ready
give
me
thumbs
up
all
that
good
stuff.
All
right,
Alice
asks
hi
folks
curious.
If
anyone
has
seen
a
tool
that
can
make
recommendations
for
limits
on
pod
resource
limits
based
on
running
history,
I
know
about
the
vertical
pod
scaler,
but
as
I
understand,
it
does
not
play
well
with
the
cluster
autoscaler,
which
is
what
we
use
thanks
in
advance.
Any
tips
here.
D
So
you
can
use
the
first
cool
puddle
together
in
a
like
recommendation
mode,
rather
than
actually
doing
anything,
which
is
probably
what
I
would
suggest
you
do.
So,
there's
like
I
think
three
ways
you
can
use
the
BPA.
You
can
either
have
it
say
just
like
log
what
it
would
recommend
for
the
requests
and
a
CPU
memory,
in
which
case
you
can
then
read
those
and
kind
of
use,
those
as
recommendations
or
you
can
have.
D
It
start
actually
like
intercepting
when
you
pods
are
created
or
just
actually
start
killing
pods
and
like
changing
the
request,
and
then
it's
sort
of
on
the
fly.
I
think
it's
those
latter
to
where
it
won't
play
very
well
with
the
horizontal
autoscaler,
where
that
first
one
there
shouldn't
be
any
problems
so
give
that
a
try
and
see
what
it
recommends.
C
We
use
BPA,
together
with
the
SCADA
and
haven't
run
into
big
issues,
but
we
also
only
use
it
on
like
one
or
two
parts.
Only
so
I
would
try
that
out
also
be
careful.
I.
Think
vertical
part
of
the
scaler
is
doing
is
setting
requests
and
not
limits.
When
you
set
limits,
be
careful
just
in
general.
Cpu
limits
might
not
be
what
you
think
them
to
be.
There
is
a
very
good
article
and
threads
about
it
by
hemming
from
zalando
you'll
find
those
out
in
a
bit.
C
Tl
DR
usually
see
few
limits,
don't
make
sense,
don't
do
them.
We
for
a
lot
of
things,
actually
even
remove
limits
on
memory
and
things
like
an
engine
X,
for
example,
because
they
have
a
strange
way
of
reloading,
for
example,
where
they
double
the
memory
for
a
very
short
time,
and
then
you
might
just
want
to
have
alerting
on
extended
use
of
resources
instead
of
something
that
OMS
your
parts.
C
A
Have
a
follow-up
question
for
both
of
you
since
sure
any
special
considerations
for
like
things
like
Java
that
might
get
unkilled
is
fine,
so
I
know
in
older
versions
of
the
JVM.
That
was
like
an
issue
but
I
know
that
in
newer
versions
or
whatever
they're
becoming
more
container
friendly.
Anyone
have
any
insights
on
on
things
like
that,
because
I
know
we've
had
questions
in
the
past
about
you
know
what
is
my
stuff
always
getting.
Unkilled
yeah.
C
I,
remember,
I,
think
anything
below
Java,
10
or
Java.
9
was
not
very
easy
to
behave
inside
a
container
and
whatnot
like,
even
if
you
give
it
limits,
it
would
just
not
adhere
to
them.
I
think
10
has
fixed
that
and
11
even
more,
but
there's
workarounds,
I,
think
redhead
and
some
others
had
very
good
blog
posts
about
how
to
work
around
that.
C
D
I
miss
hi,
yes,
oh
sorry,
I
was
just
going
to
say
on
the
sort
of
memory
and
stuff
so
from
actually
the
fear
and
Iroh
Meetup
last
week
in
London,
and
there
was
a
mention
there
that
go.
112
isn't
playing
very
well
with
the
vertical
models
of
scalar
and
there's
like
a
debug
flag
that
you
can
turn
off
to
give
it
the
old
sort
of
memory.
Behavior
again
I
will
try
and
find
what
that
was
and
post
it,
but
yeah.
If
you
use
the
VP
a
would
go
112
only
it
can
cause
issues.
A
Good
to
know
I
see
several
people
typing
in
the
channel
feel
free
to.
Let
us
know
your
experience
in
whatever
topic,
we're
discussing
and
keep
the
questions
coming,
we're
queuing
them
up
all
right
anything
else.
I'm,
pod
resource
limits,
okay,
moving
on
an
assignee
asks
how
do
I
use
as
successfully
to
reference
a
service
account
in
another
namespace
than
when
I'm
testing
against
contacts
I'm,
trying
to
make
a
kate
service
account
and
trying
to
scope
what
permissions
it
has
access
to
using
off.
A
D
So
I
believe
you
have
to
use
the
same
sort
of
reference
format
that
there
is
for
our
back
row
bindings
when
you're
doing
service
counts,
so
service
counts,
while
they
have
a
name
when
they
referred
to
in
sort
of
permissions.
They're
normally
system
code
belongs
service,
count
colon,
then
the
name
space
than
a
colon
and
then
the
name
space.
Sorry,
the
service
count
name
so
I
think
that's
probably
what
you
need
to
do.
There.
No
I
haven't
tested
that.
A
D
Yeah,
okay,
so
prowl
is
a
bit
of
a
beast.
There's
a
lot
of
stuff
to
it.
It's
made
of
lots
of
different
individual
controllers,
so
they're,
the
best
way
to
just
get
started
is
is
read
through
the
docs
there's,
also
a
bunch
of
talks
linked
on
the
rupees,
so
maybe
check
out
a
couple
of
those
I
watched
one
while
I
was
in
Seattle
I.
Think
it's
linked
by
common
bore
titles.
D
I
would
also
recommend
checking
out
some
of
the
repositories
so
I'll
link
ours
as
well,
but
a
bunch
of
the
so
on
the
readme
on
prowl.
There's
a
section
that's
like
who
uses
us
a
few
people
there
have
actually
linked
their
configuration,
which,
if
you're
gonna
start
using
proud,
is
like
Goldust,
so
I
spent
most
of
like
ours,
sort
of
getting
started
period.
Looking
at
jets
tax
repo
and
seeing
what
was
going
on
there,
how
they're
configuring
it
yeah.
D
Okay,
where
you've
got
I,
don't
know
if
you
ever
opened
a
POS
kubernetes
you'll
get
this
little
bot
saying:
hey
I
need
someone
to
approve
running
the
test.
That's
part
of
proud,
okay,
there's
the
slash
looks
good
to
me.
Start
approve,
that's
also
part
of
prowl
they've
got
this
thing.
That
does
the
labels
for
you.
They've
got
something
called
tied
which
watches
these
labels
and
then
it's
like
automation,
configuration
for
how
many
approvals
do
I
need
how
many
like
reviewers
do
I
need
before
I
can
get
merged.
D
That's
all
part
of
tied,
there's
something
called
trigger
that
starts
running
tests,
so
you
have
this
concept
of
pre
submits
and
post
submits
and
periodic
jobs.
Pre
submits,
run
on
PRS
post
submits,
run
on
pushes
to
branches,
they're
typically
use
when
you
merge
into
like
master
on
your
repo
or
something
like
that,
there's
I
say:
there's
a
whole
load
to
it.
I
could
probably
talk
about
it
for
a
narrative
and.
A
B
D
Yeah,
that's
one
of
the
reasons
we
actually
started.
Looking
at
this,
so
when
I
was
in
Seattle
for
coop
con
last
year,
I
was
doing
the
research
for
this
project
to
move
over
to
prowl
and
I
thought.
It
was
talking
to
lots
of
different
companies
about
what
the
CI
systems
look
like
and
then
had
a
demo
from
Jenkins
X
guys.
I
was
just
like
I
recognized
a
lot
of
this
stuff
from
working
on
keeping
their
season.
They
were
like
oh
yeah.
It
was
what
we
used
prowl
underneath
and
it's
like
the
discussion.
D
I
had
with
them
at
the
time
and
I,
don't
know
if
I
should
be
saying
this
or
not.
Is
that
like,
if
you're
in
a
Jenkins
ecosystem,
Jenkins
X
is
really
good,
because
you
helps
you
to
solve
my
great
as
it
were:
cuz
it's
or
wraps
the
Jenkins
conflict
and
stuff
I
think,
but
he
said,
given
we
have
no
Jenkins
just
use
prowl.
There's
one
answer
to
me,
but
yeah
that
if
you've
seen
Jenkins
X
demos,
all
that
cool,
like
automation,
stuff,
is
its
prowl.
Underneath
awesome.
A
Alright,
we're
gonna
move
into
some
beginner
questions.
Trisha,
you
have
a
bunch
of
questions,
so
we're
gonna
split
them
up
here.
A
little
bit
and
I'm
gonna
take
one
out
of
order
first,
so
the
first
one
she
asked
is
how
do
I
get
started
with
kubernetes
as
a
beginner
and
she
some
people
started
responding
in
a
thread
here.
E
First,
step
and
kind
of
just
understanding
the
main
concepts,
after
that
it
was
getting
mini,
cubed
playing
around
doing
doing
demos,
eventually
jumping
into
Kate's
the
hard
way
for
the
real
brain,
scratcher
and
just
kind
of
keep
building
and
tinkering.
It's
really
helpful.
If
you
have
a
project
or
something
you
want
to
transition
to
containers
and
cubes.
A
A
E
C
Think
I
got
started
the
hard
way,
I
went
into
Docs
and
back
then
I
think
Doc's
were
really
subpar
way
and
then
I
went
to
tutorials
online
that
were
all
outdated,
because
so
it
was
really
tough,
but
maybe
that
that
made
made
me
learn
better
because
then
I
had
to
try
it
myself,
I
think.
Nowadays
it's
a
lot
nicer.
When
we
get
people
joining
the
don't
know
when
it
is
very
well
I.
Think
Kelsey
has
an
online
course
on.
Is
it
Udacity
or
the
other
video
online
learning
platform
is
for
free?
C
A
C
A
I
started
I
literally
grabbed
a
Linux
sister
and
I'm
like
let
me
install
this,
and
it
was
absolutely
terrible,
terrible
idea
to
do
that.
I
think
so.
Now,
in
hindsight,
cuz
I
hung
around
people
they're
like
do
Kelsie's
the
hard
way
thing
and
all
that
did
was
like
confused
me
and
it
was
like
terrible
like
so
when
I
tell
people
these
days
is
start
with
the
book,
actually
the
book
that
they
give
a
Q
Khan,
it's
the
O'reilly
book
and
the
very
first
two
chapters
kind
of
explains
the.
C
C
B
Yeah
I've
actually
started
with,
like
Udacity
Co
course
from
like
Kelsey
Hightower,
and
if
you
don't
know
anything
I
think
it's
like
just
like
the
best
resource,
because
it
sort
of
approaches
you
in
a
like
really
simple
way
and
like
you,
if
you
don't
know
anything,
you
just
sort
of
sort
of
understand
the
basics
and
then
just
yeah.
If
you
have
any
like
project
and
you
can
just
like
use,
mini
mini
cube
up
your
competitors
and
play
with
it.
It's
like
really
easy
to
do.
Yeah.
A
Yeah
and
Kelsey's
chapters
and
book
always
explain
the
why
they
ended
up
at
that
kind
of
design
right.
It's
like
I,
don't
understand!
Why
have
to
do
it?
This
way,
why
can't
I
move
something?
Why
do
I
have
to
destroy
something
and
recreate
it?
Oh,
okay,
these
are
kind
of
the
tenants
they
kind
of
tell
you
the
rules
that
you're
playing
in
and
then
then
a
bunch
of
things
start
to
make
sense.
So,
thanks
to
everybody
who
is
putting
the
links
to
all
these
courses
here
in
the.
D
A
D
Just
going
to
mention
also
cats,
coders,
don't
think
anyone
much
to
do
yet,
but
they've
got
a
really
good
course
on
there
and
it's
kind
of
cool,
because
it's
like
interactive,
so
you
get
a
terminal
and
they
bring
up
the
cube
lettuce
custard
on
their
server
for
you
and
then
you
can
like
play
with
it
and
install
stuff
and
play
around
and
they
go
like
fifteen.
Sixteen
really
goods
like
getting
started.
A
Yeah,
so
hopefully
that
will
help
get.
You
started
I've
heard
of
a
bunch
of
these.
Of
course,
a
VMware
we
just
helped
launch
kubernetes
dot,
Academy
check
that
out.
That's
like
a
brand
new
one
and
let's
see
Oh
Bobbie
tables.
Has
this
Kate's
intro
tutorials.
Those
are
really
good.
What
like
I
think
if
you're
a
technical
person,
something
like
kind
or
if
you're,
on
a
Mac
doing
like
docker
with
kubernetes
I,
help
some
students
doing
that
they
just
right-click
they're
like
turn
on
kubernetes,
and
then
it
like.
A
Does
this
little
thing
and
they
were
able
to
deploy
on
their
Mac
right
away
that
was
kind
of
useful,
so
that
was
really
easy
for
them.
They
just
grabbed
tutorials
on
the
net
and
actually
do
it
on
a
live
cluster
on
their
thing,
and
we
have
to
follow
up
questions
on
kind
of
moving
things
off
of
mini
Kuban
into
real
clusters.
So
I
think
we
will
get
there
all
right,
so
that
is
a
ton
of
resources,
Trisha
I
hope
that
gets
you
started
and
we
will
come
back
to
your
other
questions
here
in
a
minute.
A
Next,
questions
from
in
ROM
validating
an
admission
webhook,
those
I
found
three
ways
to
deploy:
validating
admission,
webhook
deploys
an
HTTP
server
and
others
to
deploy
the
web
hook.
Server
is
a
kubernetes,
aggregated,
API
server
and
another
a
server
lists,
our
repo
from
Kelsey
Hightower
on
that
and
then
he's
got
links
to
that
which
one
is
the
best
way
to
go
about
doing
this
advantages
or
disadvantages
over
the
other
I
think.
C
There
is
another
way
that
is
a
bit
different.
At
least
I'm,
not
I'm,
not
sure
what
the
the
server
this
way
is,
but
even
like
just
integrating
it.
You
can
now
use
dynamic,
validating
admission
that
folks,
so
you
can
just
no
matter
how
you've
deployed
the
beat
as
a
service
inside
your
community
server
or
outside
as
an
HTTP
server
that
is
completely
external
or
as
I
guess,
also
as
a
service
platform.
C
Of
communities
as
for
where
to
run
it,
that
depends
a
lot
on
what
you're
running
I
guess,
like
any
other
ideas,
I
mean
if
you're
running
something
like
gatekeeper,
which
is
the
open
policy
agent
admission,
bad
book
implementation
that
runs
perfectly
well
inside
of
communities,
something
else
that
you
developed
might
be
easier
to
to
run
outside,
or
if
you
have
something
that
could
act
as
validating
admission
red
book,
then
I
wouldn't
just
change.
That
I
was.
E
D
A
A
Yeah.
Okay,
so
let
me
let
me
catch
up
everyone
on
chat
here.
We've
got
a
ton
of
links
there
to
courses
including
Bobbie
tables,
so
Bob
and
Jeff
actually
put
their
courseware
into
github
and
I've
actually
sat
through
that
course.
Well,
we've
run
workshops
here
locally
and
stuff
and
that
one's
pretty
good,
Omicron
Ian
wants
to
say
I.
Think
George
is
correct.
A
I've
never
done
the
hard
way
and
then
goes
on
to
say:
I
wouldn't
tell
the
new
Linux
user
to
learn
how
to
compile
a
kernel
and
canoe
utils
to
get
started
right
say
for
kubernetes
I.
Think
that's
an
excellent
point
in
Ramah
I've
got
your
follow
up
question
they're
going
looks
like
the
questions
are
keeping
on
coming
so
keep
them
coming.
Here
we
are
going
to
move
on
to
Venkat
questions
as
I've
been
using
kind
for
my
day-to-day
learning
and
development
purpose.
A
A
A
A
A
C
E
A
And
they
run
kubernetes
like
the
channel
is
hash
kind
which
I
walk
there
in
the
chat,
if
you
pop
in
through
there
and
either
ask
them
in
the
channel
or
file
an
issue,
I'm
sure
someone
can
help
there.
I
macro
me
is
saying
that
the
name
equals
is
like
the
prefix
for
all
the
things
and
just
not
the
cluster
name.
So
let's
see
what
they're
chatting
about
there.
A
So
sorry,
we
can't
help
you
there,
but
the
kind
of
folks
are
really
kind
and
they
should
be
able
to
help
you
out
so
kind
of
kind
of
changed.
The
way
I
like
between
kind
and
micro,
Kate's,
I
I,
feel
terrible
that
I
wasted
all
that
time.
Learning
all
that
by
hand
I,
remember
like
grabbing
machines
and
plug
them
all
together,
and
that
was
just
do.
C
A
C
C
C
Do
you
no
I
just
look
at
some
best
practices
that
are
like,
maybe
some
of
the
very
top
or
very
often
updated,
helm,
stable,
charts,
those
might
be
best
practices
or
show
you
some
best
practices.
Some
best
practices
that
I
tend
to
use
is
value
all
the
things,
the
template,
actually
anything
that
you
can
template
and
including
a
registry,
because
if
you
retake
images
like
we
do,
for
example,
or
like
a
lot
of
companies,
do
it
helps
a
lot
if
you
have
templated
the
registry
separately,
okay,
there's!
C
A
E
A
Okay,
all
right
moving
on
so
we're
about
halfway
through
everyone,
if
you're
just
joining
us
as
the
kubernetes
office
hours,
feel
free
to
join
the
slack
channel
that
you
see
below
and
post
a
question.
If
I've
gone
over,
your
question
I
might
have
not
seen
it
feel
free
to
repeat
it
or
if
you
want
us
to
follow
up
on
any
of
the
answers,
feel
free
to
type
if
you're
just
joining
us
just
say
hi
on
the
channel.
A
B
C
You
want
to
see
actual
usage,
you
might
need
to
install
metric
server
or
something
else
that
acts
as
the
metrics
API
back-end,
because
by
default
a
lot
of
distributions
of
communities
might
not
be
running
any
metrics.
So
you
get
only
requests
and
a
node
information
right.
So
even
sure
if
node
information
is
included
without
metric
server.
B
A
So
yeah
any
other
ideas
here.
I
have
a
follow-up
question
because
I
was
I
was
messing
around
part
of
the
reason
I
started.
This
show
is
so
I
can
ask
questions
and
not
like
so
down
these
CPU
measurements.
What
is
an
M
like
and
are
they
are
they
like
consistent,
like
if
I
say,
500
M,
on
the
CPU
limit,
but
then
I
move
over
to
an
Amazon
cluster?
Is
that
the
same
thing.
C
A
C
D
B
A
C
A
So
I
was
gonna,
ask
is
a
milic
or
the
same
on
an
epic
CPU
versus
Xeon
CPU
but
of
course,
of
course,
they're
all
right,
well,
everyone's
plopped
in
a
bunch
of
links.
Thank
you.
Thank
you.
So
much
for
clearing
that
up.
For
me,
but
music,
you
burn
edge
for
like
three
years,
I
just
learned
that
today
see
we
can
all
learn.
Something
today.
A
Keep
me
away
from
your
production
servers,
alright,
next
moving
on
Trisha
how
to
move
away,
how
to
move
away
from
running
and
deploying
a
service
from
mini
cube,
which
runs
locally
to
an
on-prem
cluster.
Now,
theoretically,
I
want
to
say
there
should
be
no
difference
but
felt
filming
in
here
on
some
of
the
actual
practical
things
that
you
have
to
do.
C
One
thing
you
would
need
to
know
is
like
who's
running,
that
on
pro
placer
and
what
is
it
like?
If
it's
really
a
very
vanilla
cone,
I'd
say
if
the
eye
that
you're
getting
you
shouldn't
have
too
many
problems,
it
might
be
on
frame
that
you
don't
have
any
dynamic
load
balancers
or
you
don't
have
an
ingress
controller
running
which
I
think
Nick.
You
PI
now
comes
by
standard
and
storage,
I.
C
Think
persistent
storage
is
now
a
mini
cube
by
enabled
as
a
default,
but
on
Prem
you
might
not
have
that
or
it
might
be
a
different
kind
of
storage.
So
there's
that's
the
kind
of
hurdle.
Sometimes
when
you're
moving
to
on
Prem
clusters
in
the
cloud
you
usually
have
less
of
those,
it
depends
a
lot
who
is
running
it
and
who
has
set
it
up
and
how
its
set
up
like
which
features
are
enabled.
B
Mean
like
in
a
decent
kubernetes
cluster
like
if
it
has
configured
like
ingress
controllers
and
things
like
that,
it
should
be
just
like
take
my
manifests,
keep
TTL
apply
on
the
new
cluster
and
it
should
it
should
be
good
to
go.
Another
thing
is
like
you
should.
If
you
are,
if
your
thing
is
actually
an
HTTP
server
or
something
like
that,
maybe
you
want
to
introduce
some
kind
of
network
policies
and
yeah
basically
limit
access
to
your
deployment.
C
E
E
A
Yes,
definitely
a
good
point:
you
don't
want
to
get
our
back,
we're
not
expecting
it
all
right,
any
other.
Any
other
tips,
any
tips
here
from
the
audience.
Let
me
look
here
about
is
moving
from
local
development
to
a
cluster.
Let's
see
thanks
everyone
for
the
information
on
Mille
cores,
just
question:
that's
what
we're
talking
about
now,
I'm
a
Kearney
wants
to
say:
I,
think
it
expressed
related
to
CPU
time,
so
the
CPU
speed
affects
the
actual
impact
setting
good
to
know
so
more
says.
A
The
core
is
a
core
doesn't
mean
that
performance
won't
be
better
awesome.
Alright,
so
keep
the
questions
coming.
While
we
move
on
Venkat,
says,
I
would
like
to
try
different
GUI
tools
to
manage
the
kubernetes
cluster
I've
tried
the
official
dashboard
UI
Ranger,
we've
scope.
Is
there
any
place
that
lists
all
dashboard
management
projects
for
kubernetes?
Or
do
you
have
any
more
suggestions,
different
solutions
for
learning
purposes?
We
got
any
links
for
them.
Awesome.
C
A
C
A
D
A
C
A
A
Yes,
so
that
will
help
you
out
another
week:
I
plopped
in
there
is
awesome
kubernetes,
that's
probably
if
you're
that
that's
not
a
bad
getting
started
one
as
well,
because
that's
a
curated
list
of
awesome
links
across
the
web
of
kubernetes
stuff.
It's
like
the
kind
of
thing
that
you
want
handy
for
all
the
people
on
your
team
kind
of
thing,
and
you
could
just
make
your
way
through
it.
So
that's
really
that's
really
handy
to
have
all
right
anything
else.
On
dashboards.
A
Alright,
don't
use
them.
Yep
Trisha
was
mentioning
stencil
before
and
scaffold
llamo
I
just
learned
it
helps
makes
a
container
will
be
the
same
while
deploy
build
cold
and
humus
and
helm
helps
us
to
maintain
the
rollback
in
deploy
so
hold
on.
Is
there
a
question
in
here
somewhere?
Let
me
look
any
of
y'all
using
scaffold
or
any
of
those
similar
tools.
A
C
Currently,
at
least
for
the
communities
part,
you
should
definitely
check
out
the
CIS
benchmark.
Okay,
it's
a
bunch
of
community
people
and
also
a
lot
of
other
security
people
working
on
it.
I
think
our
current
version
out
there
is
well
for
113,
but
it
still
applies
and
we
try
to
get
up
get
one.
A
lot
I
think
about
every
six
months
up
looks
like
Kristen
Cubans.
Exactly
Cuban
is
an
implementation
of
that
benchmark.
That
is
a
bit
easier
to
run.
There
is
also
a
sauna
boy
plug-in
that
is
running
I,
think
you
bench
inside.
B
A
Una
thing
I
want
to
point
out,
as
well
as
the
work
that
cluster
lifecycle
say.
Cluster
lifecycle
has
been
doing
to
make
that
the
clusters
kind
of
more
secure
by
default
when
you
use
in
cube
admin
right
like
when
they
turn
on
are
back
by
default.
It's
like
well
we're
not
gonna.
Let
you
kind
of
accidentally
set
up
a
cluster
without
our
bag,
so
they've
been
slowly
over.
A
You
know
over
years
or
whatever,
to
kind
of
tightening
it
up
to
make
it
more
secure
out
of
the
box
and,
of
course,
if
you're
using
a
hosted,
kubernetes
platform,
then
you
should
definitely
check
out
whatever
features
a
lot
of
them
have
like
a
features
page
of
all
the
things
that
they
do
to
make
your
thing
production
great,
ready,
any
more
comments
on
this
one
BBS.
How
do
you
guys
keep
track
of
so
much
stuff?
We
actually
use
a
great
tool
called
hack
MDI.
A
Oh
it's
my
favorite
thing
ever,
it's
like
Google
Docs,
but
with
markdown,
and
you
can
connect
it
to
github
I
just
find
out
there's
a
new
feature.
We
can
actually
keep
all
our
notes
and
github
and
like
push
and
pull
and
stuff
so
I'm
gonna
be
investigating
that
right
after
the
show.
Alright,
so
Tricia's
follow-up
question
is
how
do
I
write
a
scaffold,
dot,
yeah,
no.
B
This
one
I'm
not
sure
if
I
think
you're,
just
like
copy
pasted
from
existing.
Yes,
this
is
what
I
do
like
a
turbine
at
ease.
Yes,
yes,
it's
anybody
like
I've
used
it
and
I
think
we
have
an
example
where
you
just
like
sort
of
start
up
scaffold
and
it
starts
watching
the
code
then
actually
deploys
the
manifests
to
be
shot
in
investing
folder
yeah.
So
yeah
just
try
the
demo.
It's
like
it's
I
think
it's
like
pretty
straightforward
to
figure
it
out.
So.
A
So,
let's
see
she
goes
on
to
say:
I've
gone
through
the
whole
course
of
Pluralsight
half
way,
and
then
she
puts
the
link
the
course
explains
about
the
command.
But
now
how
to
create
chart
so
I
asked
the
same
for
scaffold,
llamó
and
home
church.
Are
those
Holloman
charts
different
as
I
am
completely
new
and
trying
to
learn
something?
Ok,
so
first
things:
first,
we
need
to
determine
what's
a
honey
chart
and
what's
what's
a
scaffold
thing
and
what's
a
what's
a
hum
thing,
scaffold.
E
A
E
A
A
Probably
wouldn't
hurt
it
I'm,
just
thinking
in
my
head
to
go
back
and
maybe
ask
Joe
or
someone
on
TGI
K
to
go,
take
a
look
at
scaffold
and
draft
and
all
of
these
kind
of
like
ancillary
tools
that
we
probably
haven't,
checked
out
in
a
while.
So
sorry
about
that,
we'll
keep
the
channel
open,
though
so,
if
anybody
has
any
tips
for
Trisha,
please
let
us
know
George,
yes,
TJ.
B
A
A
Moving
on
venkat
asks.
I
would
like
to
provision
a
kubernetes
cluster
with
separate
etsy
D
cluster.
That's
already
set
up
and
running.
Can
someone
point
me
to
any
documentation
article
on
getting
started?
Please
Thanks
and
then
Jill
said
you
replied
to
him
in
a
thread,
but
I
do
have
a
follow-up
question,
because
I've
heard
this
ass
before
when
you
say
already
a
standing,
separate,
Etsy
d
cluster,
we
definitely
don't
want
to
share.
A
D
C
A
C
A
And
that
gets
as
a
renew
separate
CD
cluster
I'm
acromion
recommends
yeah
don't
share
at
CD
with
anything
else,
which
is
that
I've
been
told
like
do
not
keep
it
yeah.
Well,
we've
been
running
a
net
CD
thing
internally
at
work
for
other
things.
Can't
we
just
like
connect
like
don't
do
that,
so
that's
good.
So.
D
Yeah
they
just
for
the
context
there.
If
you
haven't
read,
seen
the
fret
sorry
Frank
at
this
standing
up,
this
cost
is
using
qad,
so
he
can
pass
in
a
configuration
file
there,
which
then
allows
him
to
set
where
his
external
HDD
endpoints
are
and
the
credentials
to
connect
to
it.
So
that's
just
part
of
the
cube
again
in
it
phase.
B
C
C
Know
how
to
run
at
city
clusters
if
you're
running
and
yourself,
yes,
yeah,
trying
to
scale
them
up
and
down
trying
what
happens
if
one
of
the
nodes
goes
down
if
you're
running
a
multi,
node
yeah,
it's
definitely
something
that
you
don't
want
to
learn
on
the
ropes.
While
production
goes
down
yes
and.
A
I
think
everyone,
hopefully
by
now,
is
my
fully
migrated
to
exit
III
and,
if
you
have
done
not
done
so
already,
definitely
definitely
investigate
that.
So
people
are
still
gonna
talk
about
encryption.
It
looks
like
Trisha
has
a
great
question:
I
want
to
follow
up
on,
but
let's
get
Guatemala
the
way
in
a
situation
if
we
have
multiple
pods
belonging
to
one
user
and
a
consumer
facing
products,
so
basically
thousands
of
users
and
each
have
three
to
four
pods
running.
A
C
D
C
And
different
drinks
to
to
isolate
that
said,
you
will
still
have
just
a
certain
level
of
isolation.
It
depends
a
lot
on
what
you
need.
If
you
really
want
to
separate
per
machine
in,
you
might
even
put
the
namespace
on
with
note
selectors
on
certain
machines
yeah
or
you
might
even
at
some
point
run.
One
separate
I
know
virtual
or
actual
cubelets
for
four
different
yep.
A
And
I'm
recur
me
and
points
out
namespaces
a
good
way
for
a
soft
multi-tenancy,
whatever
that
really
means
so
we're
getting
into.
We
got
to
be
careful
here
right
because
a
lot
of
people
say
well
isolated
and
then
they
think
oh,
like
alright.
This
thing
is
running
and
this
thing
is
running
but
like
in
this
contest
with
multi-tenancy
we
mean
you
know,
can
my
credit
card
data
get
to
your
thing.
You
know,
like
kind
of
like
real
separation
like
in
a
separate
machine
or
what
a
VM
gives
you.
A
A
This
little
heads
up
there
on
that
one,
but
I
see
lots
of
lots
of
people
typing
so
I've
seen
this
question
asked
of
a
lot
of
like
senior
developers
like
Tim,
Hawking
and
stuff,
and
it's
phew
feels
like
a
lot
of.
It
depends
on
what
you
are
looking
for
when
you
say
multi-tenant
right
like
are
you
looking
for
that
soft
multi-tenant
as
he
mentioned,
or
are
you
looking
for
like
a
real?
No,
no
there
there
needs
to
be
a.
A
You
know,
really
hard
line
separation
between
this
workload
and
this
other
workload
so
yeah
he
mentions
like
credit
card
level.
Isolation.
Is
that
even
possible,
so
ya
know
McCartney
points
out.
G
Visor
is
a
good
run
time
to
use
your
multi-tenancy
to
I
know
that
the
clip
not
clear
cotta
containers
also
has
features
in
that
space,
and
it
feels
like
there
is
a
lot
of
work
going
into
this
area,
because
it's
not
as
awesome
as
it
could
be.
As
far
as
it
comes
to
containers,
separation
and
isolation
like
real
isolation,.
A
So
related
Reggie,
oh
cool,
my
pronounce
that
right
and
can
you
recommend
resources
for
getting
started
on
multi-tenancy
multi
cluster
management,
so
especially
multi
cluster
management
I
feel
like
is
the
next
frontier.
Now
that
cube
admins
kind
of
solved
the
set
up
one
cluster
thing:
I,
don't
think
we've
gotten
to
the
point
yet,
where
you're
kind
of
making
disposable
clusters,
like
people
often
talk
about
so
do
we
have
any
tips
here
as
far
as
getting
started
on
multi
cluster
management,
multi-tenancy
and
stuff
I
think.
C
C
E
A
A
Tricia
asks
if
your
since
y'all
are
Cuban
Haiti's
experts,
if
you
run
performance
tests,
an
application
code
in
the
same
cluster,
does
it
impact?
How
does
that
impact
from
the
infrastructure
point
of
view
I'm
assuming
a
bunch
of
you
when
you
set
up
your
cluster,
you
test
and
kind
of
benchmark?
First,
before
putting
workloads
on
it,
any
yeah.
C
A
Okay,
we
got
about
five
minutes
left
and
then
I
want
to
do
the
raffle,
so
we're
gonna
try
to
power
through
these
last
questions.
Here,
venkat
ass,
sorry
about
the
dog,
everyone
I
learned
about
cute
kernel,
plugins
to
extend
the
functionality
and
also
the
crew
plug
Lyn
to
easily
install
additional
plugins.
What
sort
of
use
cases
are
there
for
using
cute
cuddle
plugins?
What
sort
of
custom
things
can
you
do
with
cube?
Cuddle
plugins
thanks.
C
D
Yes,
you
validating
what
it
thinks
the
state
of
the
world
is,
so
you
want
to
see
what
endpoints
it
thinks
it's
going
to
route
to,
for
instance,
but
that's
kind
of
where
I
see
it
going.
It's
lots
of
like
extensions
to
kubernetes,
we'll
start
building
these,
so
things
like
manager
might
make
themselves
a
CLI
for
like
doing
some
extra
stuff
or
things
like
that
see
our
DS,
maybe
we'll
have.
C
A
Like
people
are
tossing
in
their
favorite
plugins
in
chat
as
well,
Rakus
are
a
cake,
8
ESS
I'm,
never
heard
of
this
one
before,
but
that's
always
good
to
know.
I
was
learning
something
in
the
show,
I
loved
it
all
right.
We
are
running
out
of
time,
so
we've
got
two
more
Beebe
says:
I
have
a
four
node
cluster.
One
no
got
rebooted
pods,
not
rescheduled
to
other
nodes.
Pods
got
rescheduled
to
other
nodes.
Now
that
the
node
that
was
rebooted
is
up
again,
but
it
doesn't
have
any
spot
it's
empty.
A
B
A
Any
curiosity
in
in
a
production,
cluster
or
whatever
that's
actually
live
and
doing
a
bunch
of
this
stuff.
Do
any
of
you
worry
in
real
time
about
how
many
pods
are
in
each
node?
Or
do
you
just
let
the
scheduler
kind
of
do
its
thing,
cuz,
I
kind
of
see
what
they're
coming
from
right.
It's
like
well
I
have
a
pot,
anode
and
empty
I
kind
of
want.
You
know,
I
wanna
use
the
cluster
for
the
cluster
right,
keep
it
all.
Even
things
like
that,
yeah
I
mean.
D
D
The
idea
is
that
they'll
always
be
empty
in
case
any
of
the
spot
nodes
start
getting
terminated
so
that
we've
got
like
basically
empty
nodes,
ready
sure
the
idea
here
being
that,
if
ever
all
of
the
spot
instances
were
to
go
away,
the
cost
water
scale
will
bring
up
more
on-demand
nodes
and
then
eventually,
when
the
spa
instances
come
back
into
the
cluster,
everything
will
start
scheduling
back
from
the
on
demand
to
the
spa
instances
and
then
the
on-demand
those
would
be
scheduled
down
again.
So
we'd
go
back
to
using
our
like,
cheaper
compute
right.
D
A
D
C
D
C
D
Bought
the
way
we
figured
out
the
best,
what
we
decided,
the
best
way
to
bid
for
sponsors
was
just
a
bit
the
on-demand
price,
yeah
cuz.
It
basically
never
gets
there,
which
means
you're,
always
paying
less
for
the
computer
than
you
would
be
paying
otherwise,
and
if
it
ever
gets
to
that
price,
then
you
just
start
paying
the
optimum
price.
So
right.
A
So
you're
actually
living
the
dream
and
then
somewhere
says
I
use
cluster
autoscaler
and
that
node
just
goes
away,
since
it
has
nothing
on
it,
which
is
also
nice.
Okay,
last
question
goes
to
Alan.
What
do
you
guys
recommend
for
packaging
and
deploying
applications
to
kubernetes
right
now,
I've
a
CBC
I
pipeline
and
get
lab
where
I
use
cube
cuddle
for
that
should
I
change
it
and
look
at
other
options.
This
is
an
open-ended
question,
but
let's
just
see
what
each
of
you
and
the
channel
has
to
say.
Well,
I
will
work
on
the
vessel.
B
A
All
right
going
a
little
bit
time.
We
appreciate
everyone.
Listening
up,
I
appreciate
all
the
help
for
the
panel.
It
always
makes
this
a
lot
more
fun.
When
we
have
your
expertise
there
storm
more
than
cat
I'm
a
crown.
We
always
appreciates
any
information
that
you
can
share
among
the
community.
That
always
makes
the
show
a
lot
more
fun
and
easier
and
more
diverse
as
well
too
right
because
makes
it
easier
for
us
to
find
out
what's
out
there
and
help
each
other
out.
So
we
hope
you
have
enjoyed
this.
A
We
will
be
scheduling
the
next
one
here,
probably
a
few
hours
after
this,
we
always
go
on
the
third
Wednesday
of
every
month,
except
for
today,
and
the
winners
of
the
raffle
are
one.
Our
Venkat
and
Bibi.
You've
won
a
kubernetes
t-shirt.
I
will
PM
you
after
the
show
and
give
you
a
little
code
for
the
CNC
F
store.
A
Pretty
cool
we
had
161
people
join
us
today.
So
thank
you
very
much
everybody
all
right
and
with
that
we're
gonna
close
it
out
panel.
Everyone
waves
say
goodbye
thanks.
Everybody
we'll
see
everybody
third
Wednesday
of
every
month,
right
here
on
YouTube
and
keep
on
kubernetes
ameen,
keep
on
clustering,
I,
don't
know
whatever
keep
on
copying
and
pasting
from
Stack
Overflow.
How
about
that
we'll
leave
it
at
that
all
right
cheers!
Everyone.