►
From YouTube: Kubernetes Office Hours 20181017 (West Coast Edition)
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. We have these on the third Wednesday of every month. 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/
B
You
sure
sure
all
right,
that's
it
let's
get
started
welcome.
It
is
October
17th
welcome
to
the
West
Coast
edition
of
the
kubernetes
office
hours.
This
is
a
monthly
session
set
of
live
strands
that
we
do.
Third
Wednesday
of
every
month.
We
basically
hop
on
zoom'
I
answer
your
questions
on
slack
about
how
to
use
and
deploy
kubernetes
associated
tools
whatever
it
is,
you
need
help
with
and
then
stream.
It
live
on
the
Internet,
so
I'm
in
joined
by
our
illustrious
panel.
B
D
A
B
Awesome
great
to
have
you
Matt.
Let
me
just
go
over
the
questions.
Real,
quick
thanks,
those
of
you
that
are
joining
us
on
YouTube
during
the
livestream
I'm
gonna
go
over
how
things
work,
real,
quick
and
then
we'll
get
started.
So,
first
of
all,
just
ask
your
question
in
the
channel.
That's
hash
office
hours
on
the
kubernetes
slack.
B
So
it's
obvious
to
us
and
we
can
pick
it
out
or
use
emojis
whatever,
whatever
helps
stand,
your
question
out
a
few
gotchas
though
just
remember
that
this
is
a
judgment-free
zone,
we're
recording
on
YouTube
and
of
course
you
follow
the
code
of
conduct,
so
we
all
have
to
start
from
somewhere.
So
there
are
no
dumb
questions
asked
whatever
questions
you
feel
you
need
help
with
and
what
we
will
do
our
best
to
answer
your
questions.
B
We
don't
have
SSH
access
to
your
cluster,
so
there
may
be
situations
we
might
just
be
able
to
tell
you
how
to
debug
something
or
how
to
get
further
help,
as
opposed
to
being
able
to
to
actually
get
get
in
there.
Panelists
you're
encouraged
to
expand
on
your
answers
and
experiences,
and
pro
tips
stuff
that
you
see
in
production
stuff
that
you
might
be
deploying
it
work
that
sort
of
thing.
So,
while
we
do
try
to
answer
the
questions
real,
you
know
to
answer
the
technical
question.
The
real
value
there
for
users
is
your
expertise.
B
As
far
as
experiences
running
these
things
in
production
of
audience
members,
you
can
help
us
by
tweeting
spreading
the
word
paying
it
forward.
If
you
are
listening
to
a
question
and
you
have
URLs
to
documentation-
or
maybe
you
saw
an
interesting
blog
post-
feel
free
to
just
whack-
that
into
the
into
the
slack
Channel,
we
like
to
have
the
slack
channel
kind
of
be
a
discussion
board
for
people
to
post
different
ideas.
B
So
if
you
see
something
that
remind
you
of
something
that
you
read
feel
free
to
throw
the
URLs
in
there
remember
that
we're
trying
to
help
each
other
out
here
as
much
as
we
can
and
if
you
want
to
sit
on
this
panel.
All
these
people
here
are
volunteers.
It's
it's
it's
one
hour
a
month
you
don't
have
to
show
up
every
single
month
or
every
single
session.
If
you
don't
want
to
so
it's
a
really
good
way.
If
you
figured
out
something
extremely
difficult,
you
have
expertise.
You
want
to
help
somebody
out.
B
This
is
a
good
way
to
do
that,
and
those
of
you
whose
questions
we
do
feel
today,
I'll
be
giving
away
a
kubernetes
t-shirt
at
the
end
that
you
can
get
from
the
CNCs
store
and
we
have
continued
to
to
where
actual
kubernetes
t-shirt
that
we
give
away
but
we're
working
on
it.
So
with
that
also
feel
free
to
hang
out
in
hash
office
hours
afterwards,
you
can
feel
free.
I
know
it
could
be
intimidating
the
Hangout
and
kubernetes
users
with
45,000
other
people
so
feel
free
to
just
hang
out
in
here.
B
What
we
do
is
even
when
we're
not
doing
the
live
stream,
we're
collecting
questions
throughout
the
month
and
then,
when
you
kind
of
use
those
as
backfield
questions.
If
there's
something
really
interesting
or
a
really
hard
problem,
we
like
to
highlight
it
at
least
once
a
month
so
panel.
Are
you
ready
audience?
Are
you
ready,
how's
it
sound
them
feel
free
to
give
us
feedback
right
to
the
channel
on
how
the
live
stream
is
going.
B
Let's
see,
I
got
a
new
question
here.
So
first
things
first
is
gonna,
be
from
mati
Cruz.
Welcome
back
money
says
hello.
I
have
a
quick
question
of
a
kerbin.
A
t's
cluster
master
nodes
are
one.
Ninety
one,
six,
eight
that
one
that
0/24
all
the
containers
are
on
10.2
44.0
another
sixteen
containers
can
ping
Google
combo
campaign,
other
machines
on
the
192
168
1
that
0/24
subnet.
B
D
D
Even
having
a
thread
on
this,
so
they've
got
mixed
Linux
and
Windows
nodes.
It's
only
the
Windows
ones
that
aren't
getting
this.
My
first
suggestion
was
Windows
Firewall,
because
Windows
Firewall
breaks
everything,
but
it
looks
like
actually
he's
added
a
root
on
his
gateway,
and
that
has
fixed
it.
He
doesn't
understand
why
or
how
that
fixed.
It.
D
It
yeah
I,
still
think
it
should
be,
or
is
a
firewall
during
issue
somewhere.
Something
that's
saying.
Actually
no
traffic
from
these
IPS
shouldn't
be
coming
here,
but
I
think
has
happened
here
is
where
he's
added
a
route
on
his
gateway.
The
traffic
is
now
going
from
the
pods
to
the
gateway
and
back
to
the
nodes
and
the
nodes.
Firewall
is
actually
like
yeah,
that's
fine,
but
it
shouldn't
need
to
do
that
should
be
able
to
stay
on
the
node
itself
or
just
come
directly
from
the
node
and
not
need
to
go
via
the
Gateway.
B
C
B
Okay,
well
Maddie,
please
feel
free
to
keep
adding
information
to
your
question
and
we'll
we'll
revisit
it
as
we
can
all
right.
Moving
on
Daniel
F
asks
and
I'll
be
really
looking
forward
to
this
question,
since
he
posted
it
open
any
question
for
office
iris
today.
What
best
practices
would
you
recommend
for
developers
prototyping
on
a
shared
cluster?
How
can
engineer's
use
tools
like
telepresence
without
stepping
on
each
other's
toes
different
development
clusters?
B
For
engineer
clusters
would
like
to
later
name
spaces
for
engineer,
as
your
aks
has
dev
spaces
feature
in
preview,
which
seems
to
offer
routing
to
particular
dev
services,
while
returning
retaining
shared
services
throughout
the
rest
of
the
cluster.
Is
this
kind
of
approach
widely
use?
Are
there
alternative
tools
to
support
it?
What
do
we
think
this
is
similar,
I?
Think
to
the
question
we
had
earlier,
you
know:
is
it
better
to
have
a
bunch
of
small
clusters
or
one
big
cluster
with
a
bunch
of
namespaces?
You
had
a
similar
question
like
that
this
morning.
C
So
I
know
a
lot
of
customers
that
I
deal
with
actually
do
a
namespace
protein
because
a
lot
of
times
it's
not
just
a
single
application,
developer
working
on
one
application,
it's
more
of
a
team
working
on
an
application,
so
they'll
do
a
namespace
per
application
or
per
team
as
a
pretty
common
scenario.
And
then
you
don't
have
to
take
care
of
the
cluster
administration
and
then
certain
Ingrid
I
don't
mean
you
can
just
latched
on
to
ingress
route
matches
for
your
needs.
A
Know
I
would
agree
with
that
completely
with
that
and
I
think
I
want
to
emphasize.
Perhaps
when
Matt
said
to
expand
a
little
bit
is
if
you're,
actually
developing
stuff,
like
operators,
controller
stuff
that
actually
messes
with
the
communities
themselves
or
even
like
a
custom
scheduler
that
might
be
worse
to
have
your
own
a
you
know
a
little
cluster
for
you
as
opposed
to
trying
to
mess
with
it
somebody
else,
but
otherwise,
if
it's
just
a
normal
application
or
a
normal
like
you
know
it
something
like
harmless,
I
guess
so
does
become
less
but
info.
B
All
right
any
other
comments
on
that
one
before
we
bum
all
right,
XC
k,
uu,
1,
XC
q,
1
sorry
asks
I've
got
kind
of
a
Daffy
question.
How
often
should
I
update
kubernetes
and
how
particularly
dangerous
is
it
to
get
behind
eg
running
a
year
old
clustering
production,
the
Ralph
I
know
Ralph,
has
opinions
on
this
one.
Anybody
else
want
to
go.
First
sure,
I'll.
A
Buy
so
no
I
guess
it
really
depends.
You
know,
like
I'm,
obviously
I
would
say,
like
I'm,
a
big
proponent
of
like
three
treating
clusters
as
cattle,
not
pets
right
so
like
I'm,
a
big
fan
of
moving
things
over
as
fast
as
possible
and
discarding
bill
thing.
However,
reality,
especially
in
corporate
environments.
That
is
not
a
reality.
That's
not
a
possibility.
A
Let's
say
like
you,
just
you're
selling
1.9
so
like
in
those
cases,
is
when
I
would
further
consider
like,
instead
of
just
upgrading,
because
it's
very
painful
right,
you
cannot
leapfrog
from
1.9
to
1.12.
You
need
to
like
do
every
upgrade
to
every
single
miner
right,
so
you're
gonna
do
one
point:
10,
1,
2011
and
1.12
itself.
It's
very
hazard
all
through
all
these
particular
steps,
and
a
lot
of
things
might
have
been
like
backwards,
paddleball,
etc.
A
So
you
might
have
to
be
new
some
fixing
in
between
so
like
in
those
cases,
I
do
recommend
doing
something.
Instead,
like
a
work
load,
migration
first,
like
I
work,
will
back
up
and
blood.
You
know
shameless
plug
for
arc.
You
know
bark
art
all
you're,
like
communities,
objects
and
making
sure
that
everything
that
you
have,
whether
it's
volume
society
is
all
backed
up
and
just
stand
up,
and
you
one
point
twelve
point:
three
or
whatever
and
move
all
your
move.
All
your
work
goes
back
to
the
new
cluster.
A
That's
some
much
easier
and
like
a
lot
less
error-prone
path,
that's
what
I
would
say
so
the
risk
that
you
have
to
wait.
You
know
this
again,
it's
like,
if
you
wait
too
long,
there's
going
to
be
a
pain
to
upgrade
either
way
right
like
whether
you
do
it
the
shortcut
way
or
not.
It
there's
no
leap
frogging
and
the
other
one
is
like
going
too
soon
and
then
officer
I'm
finding
that
the
newer
version
doesn't
really
have
functionality,
functionality,
functionality
that
you
were
counting
on.
E
You're
now
out
of
the
cycle
for
even
looking
at
those
things,
I
would
also
so
I
would
suggest,
keeping
up
with
your
versions
enough
that
you're,
you
stay
in
the
supported
window
unless
we
get
to
something
like
an
LTS
which
we
don't
have
right
now.
I
would
also
suggest
because
again,
as
those
changes
come
along,
maybe
look
at
something
like
testing
when
you
upgrade,
so
maybe
you
wait
till
one
dot.
11.3
comes
out.
E
I
also
would
like
to
see
cases
where
you
actually
could
upgrade
a
cluster
and
even
do
rolling
upgrades
of
a
cluster
I.
Don't
think
we're
really
there
yet
in
terms
of
functionality,
but
when
in
public
clouds,
it's
not
so
bad.
If
you
can
adjust
your
routing
rules
and
things
like
that,
but
again
in
those
these
enterprise
and
on-premise
situations.
Sometimes
you
really
want
to
upgrade
a
cluster.
E
That's
just
a
set
of
hardware
and
you
don't
have
a
whole
other
set
of
hardware
to
move
to
and
and
one
of
those
things
that
I,
don't
think
is
there.
But
I
would
love
to
see
is
about
our
ability
to
upgrade
those
clusters
in
a
rolling
manner
live
but
again
make
sure
that
you
test
your
upgrades,
because
you
know
if
you're
in
an
enterprise
environment
or
something
like
this
and
it
falls
over
and
you're
not
ready.
C
Ralph
looks
like
you
had
an
opinion
on
this
yeah
so
with
the
N
minus
2
scenario
being
a
year
back,
not
only
security,
but
even
trying
to
find
new
and
novel
things,
depending
on
the
items
that
you're
trying
to
work
with
inside
of
your
cluster.
Some
of
those
don't
exist
so
some
of
the
new
items
support
especially
on
Krim,
some
of
the
CSI
stuff,
has
been
added
very
recently
in
newer
versions
of
kubernetes,
so
you're
not
just
getting
newer
security
fixes
you're.
C
B
And
I
did
want
to
add
that
if
you
look
on
the
kubernetes
dev
mailing
list,
there
is
a
working
group
being
formed
right
now
to
explore
is
an
LTS
version
of
thing
that
the
kubernetes
project
can
support.
So
if
you
have
opinions
on
that,
you
should
probably
check
that
out.
This
I'm
gonna
move
on
to
a
related
question
that
Nick
asks
it
goes.
B
D
E
When
it
comes
to
API
versions,
kubernetes
buy,
it
always
supports
the
V
ones
forever,
and
so
those
will
carry
over
the
betas
are
supported.
For
let's
say
it's
two
releases
or
six
months,
whichever
is
longer,
and
so
when
you're
upgrading
your
cluster,
you
know
and
the
alphas
can
change
from
one
to
another,
I
would
suggest
in
production.
Don't
use
your
alphas
unless
you
really
know
what
you're
doing
if
something's,
beta
or
released
I
would
probably
suggest
trying
to
upgrade
your
apps
and
your
cluster
separately
from
one
another.
E
So
that
way,
if
something
does
go
wrong,
you
know
where
to
look
rather
than
two
different
places.
So
I
would
suggest
that,
and
quite
frankly,
if
we're
being
honest,
there's
actually
a
bug
in
kubernetes
and
so
lots
of
old
API
versions
are
still
around.
That
haven't
been
removed
to
a
storage
bug
of
things
being
written
to
disk,
and
so
at
the
moment,
we're
generally
safe
on
things
not
disappearing
right
away.
So
I
would,
you
know,
upgrade
my
cluster
and
then
come
back
later
and
upgrade
my
stuff.
B
And
this
is
something
that
all
kubernetes
will
have
to
use.
Users
will
have
to
do
right,
Lee,
even
if
you're
on
the
public
cloud
and
the
cloud
providers
do
upgrades
for
you,
you
still
have
you
still.
You
can
just
leave
it
right.
You
still
have
plenty
plenty
of
work
to
do.
Okay,
anything
else
related
to
upgrades
before
we
move
on.
That
was
a
really
great
question.
Thank
you.
B
D
Think
90%
of
it
is
guesswork
at
first
they're
like
it
really
is
so,
whenever
I've
added
something
new
to
the
cluster
I
tend
to
give
it
some
sort
of
arbitrary
resource
request,
maybe
half
a
CPU
half
a
gig
of
memory
and
just
see
what
it
uses
over
time.
Try
and
stress
it.
So
if
it's
a
like
an
engine,
X
I'll
fire
off
a
thousand
requests
so
that
see
roughly
how
much
resource
it
uses
I
mean
request
a
lot
to
handle
and
then
be
like
okay,
well,
I'm
gonna
have
ten
replicas
on
10,000
requests
concurrently.
D
This
is
how
much
resource
I
need,
so
you
can
sort
of
get
it
a
bit
like
that.
But
at
the
end
of
the
day,
you
just
gotta
monitor
what
they
use
and
I
think
just
adjust
based
on
that
and
yeah
I.
Don't
know.
There
is
also
the
vertical
pod
on
toast
gala,
which
you
might
want
to
look
at
if
you're
doing
this,
it
kind
of
keeps
track
of
how
much
resource
you're
using
and
adjusts
the
pods.
D
B
D
D
A
Even
if
you
do
use
the
HPA,
it
is
not
without
hazard
right
like
it's
like
it's
still,
no
I
would
I
would
say
like
it
works
well,
but
it
still
has
some
kinks.
You
know
like
it
seems
to
be
over
active
when
way
or
the
other.
It
tends
to
either
over
generate
or
overkill
like
a
you
know
like.
So
it's
just
you
need
to
make
sure
that
you
monitor
your
thing
and
make
sure
they're
the
metrics
that
you
are
and
the
parameters
that
you
configure
with
makes
sense
for
your
use
case
gives
up
against.
A
A
So
like
I
would
yeah
I
would
reiterate
with
oliso
just
basically
first,
you
know
put
some
parameters.
There
see
how
it
behaves
and
then
once
you
figure
out
a
good
set
of
quotas
and
resource
limits,
then
you
can
start
adding
like
a
out
of
scale
into
that.
But
you
know
don't
try
to
go
like
auto
scaling
right
off.
The
bat
is
that's
gonna,
be
a
bad
time.
Yeah.
B
And
he
has
sort
of
a
related
follow-up
question:
what
are
some
of
the
best
ways
to
able
to
track
your
pod
resource
usage
over
time?
What
do
you?
What
do
you
guys
use?
Is
it
the
usual
mix
of
me
theists
or
whatever
monitoring
thing,
but
a?
What
are
some
techniques
we
can
recommend
for
Christian
here,
Prometheus
with
Cortana
yeah?
That's
exactly
what
makes
us
in
slack.
Okay,.
B
D
It
adds
some
more
metrics
I,
think
it's
more
maybe
on
the
node
level.
I
just
know
it's
part
of
our
monitoring
setup
that
exposes
some
metrics
around
sort
of
resource
usage.
I
think
it
does
have
some
stuff
over
on
the
containers
as
well,
adding
things
like
if
you're
requesting
GPUs
it
adds
that
as
a
metric,
so
how
many
GPUs
you're
requesting
might
be
useful
for
someone
but
I
know
that's
another
part
of
the
system,
but
yeah
mainly
prometheus
nc
advisor
you.
B
Okay
and
the
other
recommendation
I
saw
Matt
just
nodding
so
I'm,
assuming
he
likes
Prometheus
anger,
okay,
moving
on
oh
and
any
time
those
of
you
listening.
If
you
request
to
follow
up
just
feel
free
to
pop
in
another
question
and
we'll
get
to
it
even
see.
Sorry
if
I'm
mispronouncing,
that
welcome
back
I
recognize
your
name
from
before
this
question.
D
So
I've
been
having
a
discussion
there
with
a
few
in
the
channel
as
well.
Nick
who
asked
a
question
earlier
was
also
joining
in.
So
there
were
some
suggestions
for
things
like
meta
controller
and
key
builder
and
I
just
saw
it
added
there
there
more
for
custom
resource
controllers,
then
admission
controllers,
key
builder
I
know,
does
have
admission
control
of
stuff.
Coming
though
it's
in
the
works,
but
I
don't
think
it's
quite
ready
yet
there's
definitely
no
documentation
on
how
you
create
an
admission
controller
using
keyboard
moment.
D
So
I
pasted
it
in
the
open
ship
generic
admission
server,
which
I've
used
pusher
called
Kwak,
which
gives
an
example
of
how
this
works
it's
been
around
for
a
while.
Now
that
project
we
did
like
eight
or
nine
months
ago,
and
it's
pretty
mature
at
this
point
I
think
it
gives
you
all
of
the
are
back
stuff
as
well,
so
it
does
they
like
delegated
or
thing
to
check
that
the
person
requesting
the
API
endpoint
has
the
correct
are
back
to
do
so.
D
So
if
you've
got
secrets
that
are
being
injected
in
you're
mutating
webhook,
then
it's
safe
there
as
well.
So
I
think
easy
to
come
back
and
be
like
actually
no
can
go
off
with
that
and
use
the
quack
thing.
As
an
example,
the
I
found
that
I
think
I
posted
in
sick
API
machinery
and
was
like
hey
I,
want
to
build
it
admission
controllers.
Anyone
have
any
examples
and
the
open
guys
come
back.
I
think
it's
definitely
you
just
need
me.
It
was
like
hey,
go
use
this
and
it
yeah
works
very
simple.
B
E
On
top
of
that,
because
we're
starting
to
see
things
like
Etsy
D
operators,
Jenkins
operators,
I,
know
Oracle's
working
on
my
sequel
operator.
If
there's
a
whole
bunch
of
these
out
there
to
bring
in
this
business
logic
around
these
stateful
applications,
I
think
it's
worth
looking
at
any
one
of
these
because
it
really
has
to
do
with
who
built
it,
how
much
they
know
and
how
well
they
they
know
the
things
about
those
applications
and
how
they're
able
do
it.
So
it
kind
of
depends
on
the
operator
who's
building
it.
E
How
well
they
know
things
I've,
seen,
operators
that
are
very
alpha
and
some
of
them
even
say
that
so
it's
dangerous
to
play
with
those
sometimes
are
even
missing
features
and
then
others
that
are
really
well-rounded.
Like
I
think
chorus
is
Etsy
D
operator.
They
know
at
C,
D
pretty
darn.
Well,
they
created
it,
and
so
they
created
an
operator
and
it's
been
around
for
a
while
and
it's
in
a
bit
of
use.
E
So
something
like
that's
gonna,
be
a
little
bit
more
stable,
but
just
like
any
application,
you're
grabbing
off
the
shelf
with
those
operators
take
a
look
at
them
and
make
sure
that
they
are
mature
enough
well-rounded
enough.
Give
them
a
good
evaluation
before
you
use
them,
but
I
do
think
in
general.
You
can
do
stateful
stuff
in
kubernetes
now
and
that
that's
okay,
as
long
as
you
have
backups
and
other
things
in
place
like
you
should
have
anyway,.
D
Not
at
the
moment,
no
weird
relying
on
Amazon
for
all
of
our
states
at
the
moment
that
will
change
in
the
future,
though,
we've
been
looking
at
various
operators
as
well
for
these
kind
of
things
for
tests
in
particular
yeah.
That's
not
something.
We've
been
working
on
at
the
moment,
Ralph
or
Rueben.
C
We
have
customers
that
use
certain
operators.
I
know
we
have
customers
that
use
of
Prometheus
operator
from
core
OS
just
because
some
of
the
setting
up
prometheus
isn't
always
a
great
time.
So
there's
that
and
then
I've
seen
an
RDS
operator
that
one
of
our
customers
also
uses
for
RDS
databases
in
AWS.
That's
worked
pretty
well
for
them
up
to
this
point,
mm-hmm.
B
And
I'm
going
to
paste
a
discussion
that
we've
had
in
the
past
when
we've
had
Josh
burkas
on
seems,
like
someone
always
asks,
and
are
we
doing
containerized
databases
or
the
hosted
cluster
style
thing
ends
up.
He
has
a
blog
series
about
this
and
there's
this
entire
discussion
that
he
had
in
August
with
someone
where
they
really
kind
of
get
into
the
details
on
balances.
As
far
as
what
what
you
need
out
of
your
database
and
what
limitations
and
advantages
there
can
be
to
either
approach.
B
So
that's
definitely
good
background
information
I
definitely
learned
a
lot
just
by
reading
it,
even
though
I
couldn't
understand
half
of
it.
So
I
definitely
recommend
that
conversation
of
posted
that
URL
in
the
slack
and
Nick
would
like
to
addy.
He
says:
I
found
a
lots
of
operators
in
Elfa
stage
as
well.
The
only
operator
we're
using
in
production
is
core
core
OS
Prometheus,
which
is
the
one
that
you
mentioned.
B
Well,
we
are
looking
for
more
questions,
so,
while
people
are
coming
up
with
more
questions,
let
me
just
remind
everybody:
you
are
listening
to
the
kubernetes
office
hours,
we're
about
halfway
done
here.
Little
bit
over
halfway,
if
you
have
any
questions,
you
can
join
us
directly
on
hash
office
hours
on
slakoth
kubernetes
is
at
I/o.
Just
follow
the
instructions
post
it
down
here
and
let
me
see
if
I
missed
your
question
feel
free
to
repost.
It
I.
B
See
lots
of
people
typing
so
we'll
just
give
that
a
minute
anything
interesting.
You
guys
see
this
week
that
you'd
like
to
talk
about
I,
feel
like
asking
Matt
a
bunch
of
how
some
3.0
coming
questions.
Put
you
on
the
spot.
Can
you
give
us
like
a
30?
Second,
how
are
things
going
over
there
how's
house,
being
your
own
CNCs
project
now
these
days,
it's.
E
Actually
going
pretty
well
helm,
3
is
moving
along,
so
one
of
the
things
about
home,
3
I
know
a
lot
of
people
don't
like
till
her
and
how
tillers
already
gone
from
home
3.
It's
no
longer
there,
no
longer
a
thing,
we're
actually
in
the
process.
So
there
was
this
nice
document
that
we
worked
on
and
discussed.
So
we
planned
things
out
a
little
bit
of
ahead
of
time,
because
so
how
has
this
idea?
E
Where
version
2.0
was
always
going
to
be
stable,
API
compliant,
we're
not
breaking
internal
API
is
for
people
who
import
packages
or
commands,
or
any
of
that
and
of
course,
over
time
you
build
up
design.
Debt
and
kubernetes
has
changed
a
lot
since
helm
came
out,
and
so
we
wanted
to
take
advantage.
Some
features
and
things
like
that,
so
Humphreys
the
opportunity
to
to
break
api's
right.
E
Somebody
imports
how
much
a
package
api's
will
break
things
like
that,
and
so
we're
taking
that
chance
to
clean
up
a
lot
of
stuff
and
that's
where
a
lot
of
work
has
been
happening
and
tillers
gone
and
and
lots
of
underneat
stuff
has
been
happening
right
now.
We're
trying
to,
but
earlier
today
he's
working
with
a
couple
of
other
folks
to
get
everything
in
a
nice
trackable
manner
and
a
project
that
we'll
be
able
to
share
soon.
E
That
really
does
show
status,
so
we
can
see
where
are
things
at
and
how
is
it
moving
along
and
how
much
stuff
is
done,
so
we
can
see
it
happening.
We
probably
won't
get
a
release
this
year,
it'll
probably
happen
sometime
first
quarter
or
something
next
year,
but
we
are
moving
along
with
it.
In
fact,
some
of
its
actually
been
delayed
because
of
having
to
deal
with
client
go
changes
from
release
to
release
and
bringing
that
into
helm,
v2
and
those
upgrades.
B
E
A
Something
Dwarfs
no
to
like
I
think
it
was
yesterday.
The
first
v1
13
was
today.
The
first
tag
was
already
pushed,
so
you
know
like
this
will
be
and
as
a
reminder,
this
will
be
a
shortened
cycle,
so
we
want
to
get
at
1:30
now
before
Rubicon
or
ironical
become
so
yeah.
This
is
something
to
look
forward
to
we're.
Getting
there
really
really
fast.
B
Awesome
yeah,
it
looks
like
we're
expecting
quite
a
large
crowd
at
this
cube
con
cloud
native
con,
so
I'm
hearing
north
of
like
8,000,
ish
I,
don't
know
we'll
see
we're
gonna
pack
that
thing
it's
gonna,
be
great
I'd
love
to
tell
you
all
the
sign
up
for
the
new
contributor
summits,
but
I
think
we're
so
bad.
So
are
you
all
going
to
keep
con
just
those
of
you
that
are
interested
in
going?
You
can
definitely
link
up
with
us
there
and
I'll
have
swag
for
you
and
stuff.
B
Alright,
keep
the
questions
coming
and
move
gonna
move
on
to
Christians
question
that
you
all
seem
to
answer
pretty
unanimously.
So,
let's
look
Kirsten
asks.
What's
the
easy
way
to
install
Prometheus
is
cube,
Prometheus
the
way
to
go
and
and
Ralph
says
the
operator
works
so
nicely.
How
newbie
says
take
a
look
at
the
Prometheus
operator
and
then
Christian
says
cube
Hermes.
He
seems
to
be
using
that
operator.
So
are
we
do
we
have
the
exact
URL
for
the
thing
he's
he's
looking
for,
if
someone
could
just
slack
that
URL
in
I
believe.
C
B
B
Appear
so
if
not
just
whack
em
in
chat,
easy
seems
to
have
a
hum
question.
We're
gonna
go
ahead
and
get
to.
It
is
hum
three
keeping
the
same
logic
of
how
to,
and
that
upgrade
is
diff
against
the
previous
release,
to
determine
changes
to
apply
rather
than
see
what
actually
exists
in
the
cluster.
So
something
is
modified
outside
of
how
it
will
be
brought
back
in
line
with,
what's
defined
in
the
helm.
Chart
on
next
upgrade.
E
So
helm
3
is
under
development,
so
everything
is
open
to
change.
Remember
that
we
don't
even
have
an
alpha
yet
yeah,
so
anything
is
open
to
change.
I'm,
not
sure
of
the
answer
to
this,
but
I
haven't
heard
of
any
different
Adam
riess
is
working
on
that
work
right
now
and
in
that
pipeline
and
cleaning
it
up,
but
I,
don't
think
he's
actually
changed
the
logic
in
it.
One
thing
I
will
say
is:
if
you
look
at
something
like
linker
d2,
it
does
things
pipe
with
coop
control
and
would
make
out
of
out
of
sequence.
E
So
how
would
that
apply
to
working
with
helm,
we're
actually
talking
and
looking
at
how
that
situation
could
be
made
easier,
because
sometimes
there
are
reasons
to
you
know:
make
modifications
outside
of
a
chart
and
just
how
should
that
work
and
how
should
that
flow
go
in
order
to
make
it
work
and
so
we're
discussing
it?
But
right
now
I,
don't
think,
there's
a
change
from
helm
to,
but
there
may
be
some.
B
A
E
B
E
This
point
I
would
not
chase
the
moving
target
elm
tree
not
at
this
point.
B
E
B
E
B
B
Yep
so
speed
about
113
being
being
a
short
cycle,
I
volunteered
for
a
release
cycle
for
the
release
team,
so
I'm,
actually,
third,
shadow
of
the
communications
team,
so
I
think
I
get
to
help
out
with,
with
blog
sorry,
my
kids
running
around,
with
with
blog
posts
and
stuff
and
and
announcing
things
something
I
thought.
It
was
interesting.
Those
of
you
that
are
ever
interested
in
participating
in
them
and
like
the
kubernetes
release
team
is
you
have
to
be
a
shadow
before
you
do
anything
else,
so
you
can't
accidentally
walk
into
failure.
B
So
you
kind
of
start
off
and
there's
a
apprenticeship
role
where
you're
helping
a
lead
do
some
things
until
you
kind
of
understand
and
then
maybe
next
cycle
you
get
a
little
bit
more
responsibility,
so
I
was
hesitant
to
volunteer
because
I
was
like
well,
you
know
what,
if
I
met
what,
if
I
don't
have
as
much
time
as
I
thought,
I
did
and
stuff,
but
the
project,
the
team
is
really
structured
to
kind
of
help.
You
not
fail,
which
I
thought
was
really
awesome,
so
yeah
and.
A
A
B
E
Not
for
113
I
think
Mike
citement
over
new
things
mostly
went
away
with
the
workloads
API
becoming
stable.
No
because
we
didn't
have
major
changes
all
the
time
we'd.
Finally,
ferreted
things
out.
The
and
cron
jobs
going
towards
stable
is
still
several
releases
away,
but
we're
working
on
it
now,
no
there's
actually
I'm
enjoying
the
stability
and
then
the
sanity
that
brings
yeah.
B
I
feel
like
this
year
has
been
relatively
boring,
I
think,
or
am
breakfast
the
other
day,
man
I
think
this
is.
We
haven't
had
a
broken
point,
zero
release
in
like
a
while.
It
feels
like
it
feels
like
we
finally
finally
are
getting
there
easy
says
the
bug
containers,
hopefully
I'm
not
tried
that
one
use
Ralph
or
Joe
anything.
That's
particularly
you're.
Looking
forward
to
I.
D
C
B
Alright
and
with
that
last
call
for
questions
I'm
about
to
generate
the
result
of
the
t-shirt,
giveaway
app.
So
unless
we
have
any
other
questions,
we'll
go
ahead
and
wrap,
it
wrap
it
up.
I
like
to
thank
our
panel
for
volunteering,
their
time.
There's
one
more
question:
who
there
is:
oh,
yes,
how
newbie
go
ahead?
Yes,
I
always
like
to
get
one
in
there.
B
B
A
He's
it
can
be
rough,
I
would
say
so,
like
there's.
Actually
some
frameworks
out
there
to
help
provisioning
a
eks
clusters,
because
there's
a
quite
a
bit
of
like
overhead
and
you're
starting
charlie,
clear
and
it
particularly
as
it
surrounds,
like
the
I,
am
rules
that
you
need
to
have
and
the
roles
that
you
need
to
have
for
that.
A
D
This
is
Ilias
project.
At
the
moment,
it's
meant
to
make
making
eks
clusters
a
lot
easier.
I've
not
used
it
personally,
but
I
saw
earlier
do
a
demo
of
it.
On
the
day
that
eks
was
sort
of
announced,
he
sort
came
up
on
the
stage
and
was
doing
this
and
then
debugged
and
released
the
entire
new
dot
release
during
the
course
of
this
demo.
D
When
there's
a
hundred
people
debugging
one
application,
it
was
a
bit
weird,
but
the
theory
is
that
you
can
use
this
to
just
go
eks.
Make
me
any
case.
I,
don't
know
exactly
how
far
it's
gone.
I
keep
seeing
him
tweets
about
it
and
the
leaf
guys
tweet
about
it.
Yeah,
so
I
guess
they're
working
on
it
pretty
hard
at
the
moment.
Actually.
B
D
B
C
Caps,
a
lot
for
deaf
clusters
on
AWS,
because
they're
really
fast
to
spin
up
once
you
have
your
little
tiny
template
just
creating
and
deleting
that
cluster
list
of
clusters,
pretty
quick.
It
takes
me
less
time
to
do
that.
Then,
if
I
were
to
manually,
do
an
e
KS
cluster
to
be
honest
and
even
the
eks
CTL.
Some
of
the
time
frames
aren't
super
fast.
B
D
B
All
right
and
with
that,
we
will
wrap
it
up.
Thank
you
very
much
to
our
contributors
for
volunteering.
Your
time
I'd
like
to
thank
the
following
companies
for
supporting
the
community
by
allowing
their
developer
volunteers
to
join
us
giant,
swarm
hefty
Oh
stock
ex
packet,
botnet
pusher,
comm,
Red,
Hat
Samsung.
Yes,
yes,
we
works,
VMware,
Xing,
Huawei,
the
University
of
Michigan
and
special
thanks
to
Google
for
sponsoring
the
t-shirt.
B
Giveaway
will
be
posting
this
on
Google
and
we
will
see
everybody
in
a
month
in
the
meantime
feel
free
to
hang
out
in
the
channel
thanks
everybody
for
listening
in,
and
we
always
appreciate
your
questions
and
feel
free
to
give
me
any
feedback.
If
you
feel
like
we
can
make
the
show
better
and
with
that.
Thank
you.
Thank
you.
Everybody
all
right.