►
From YouTube: Kubernetes Office Hours 20220216 (SIG ContribEx 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. They are great for answering questions, getting feedback on how you’re using Kubernetes, or to just passively learn by following along.
For more info: https://k8s.dev/events/office-hours
A
B
B
Okay,
yeah
we're
live.
Where
is
a
window
to
the
introductions.
C
So,
let's
get
started,
welcome
everyone
to
today's
kubernetes
office
hours.
We
are
where
we
answered
your
questions
live
on
the
air
with
our
steam
panel
from
experts.
You
can
find
us
in
the
office
hour
slack
channel
and
check
the
topic
for
the
url
for
the
information
before
we
begin.
Let's
start
introduce
introducing
ourselves
and
we'll
then
we'll
continue
so
who
wants
to
start.
D
C
And
david
said:
a
story
like
this
david
said:
who
goes
next
all
right.
A
I'm
guy
templeton,
I'm
a
principal
software
engineer
at
skyscanner,
where
I
am
a
member
of
the
team
that
takes
care
of
our
kubernetes
clusters
as
well
as
other
container
platforms.
I'm
also
a
co-chair
of
kubernetes
psychoscale.
A
E
E
I
also
am
an
emeritus
chair
for
contributor
experience,
but
I
also
still
do
a
lot
of
contributor
experience
stuff,
as
I
am
a
maintainer
for
a
lot
of
the
sub
projects
which
we'll
get
into
in
a
second
super
excited
to
be
here
and
thanks
y'all
for
inviting
me
to
get
the
word
out.
Next
is
mario
hello,
mario,
who.
F
Are
you
tell
the
world
it's
good
to
talk
to
you
again?
I
am
nothing
really,
but
I
am
part
of
the
michigan
crew,
the
renowned
michigan
crew,
some
names
you
might
have
heard
of
george
castro,
bob
killen,
mr
bobby
tables,
jeffrey
cecca
gifi.
Chris
short,
I
could
keep
going.
Michigan
is
great.
I
am
currently
a
site
reliability
engineer
for
a
financial
services
company
named
carta.
F
If
you
have
stock
options
with
any
organization
you
might
have
used
carta
and
so
I'm
kind
of
solving
cloud
native
in
the
financial
world,
which
is
incredibly
stressful.
So
I
don't
have
much
hair.
That's
why
I'm
wearing
a
hat
really
happy
to
be
here.
I've
been
using
kubernetes
since,
like
1-6
or
something
crazy,
I
love
the
networking
side
of
the
table
and
scalability
pieces
so
welcoming
your
questions
that
hopefully
stump
us
a
little
bit,
I'm
going
to
pick
on
archie
next,
thanks
mario.
G
Hello:
everyone,
my
name,
is
archie.
I
am
cncf
ambassador
from
canada,
happy
to
be
here
with
this
amazing
crew,
I'm
organizing
the
meetups
in
canada
and
also
teaching
kubernetes
in
the
university
now
and
in
the
free
time.
I
work
at
google
as
an
application,
modernization
specialist.
So
looking
forward
to
this
event,.
H
What
do
you
want
to
pass
it?
I
think
we
have
chris
yeah.
My
name
is
chris
priveter.
I
work
over
at
equinix
metal.
We
do
bare
metal
in
the
cloud
I've
done
kubernetes
for
a
few
years
before
that,
and
now
I
get
to
trip
over
stuff
that
david
left
behind
when
he
left
and
one
of
his
co-workers
kind
of
pointed
me
at
this,
and
it's
my
first
time
so
glad
to
join
you
all.
Thanks
for
having
me
on.
C
Good
yeah
perfectly,
I
need
to
get
a
new
mic.
I
think
I'm
last
my
name
is
carlos
santana.
I
was
in
a
band
for
six
years
playing
the
guitar
and
not
not
really.
I
am
a
senior
architect
working
for
ibm
cloud.
I
help
clients
an
enterprise,
software
multi-cloud,
mostly
openshift,
but
I
do
community
work
and
one
of
them
is
I'm
a
steering
committee
member
for
k
native,
that's
an
open
source
project
that
little
secret
is
becoming
cncf
this
week
and
the
other
thing
that
I
do.
C
I
run
a
book
club
on
on
a
weekly
basis,
the
kubernetes
book
club-
it's
not
official,
so
I
need
to
get
with
contrib
contributor
x6
to
see
if
that
could
be
a
program
that
they
want
to
sponsor.
And
what
is
what
I
do.
I
I
run
a
newsletter
and
I
help
people.
You
know,
learn,
kubernetes
and
and
learn
about
how
to
get
started.
C
C
So
with
that
introductions,
let's
get
a
little
bit
of
the
house
in
order
also.
The
other
thing
that
I
wanted
to
help
was,
and
I'm
learning
about
the
release
of
kubernetes.
So
I'm
a
release,
shadow
and
release
team
for
1.24,
so
I'm
doing
prs
on
release,
notes
and
learning
the
progress
process
and
meeting
a
lot
of
folks.
So
that's
another
thing
that
people
can
do
if
they
join
one
of
the
six.
So
I'm
in
sick
release.
C
Right
now
you
can
say
that's
a
shadow
and
also
in
helping
david
and
david
have
been
here.
I
think
every
episode
we
we
shout
out
to
one
of
the
one
of
the
members.
I
think
we
missed
it
last
last
month,
but
this
month
I
think
I
would
choose
david
as
our
member
of
the
month
on
the
office
hour.
C
He
have
been
doing
it
for
a
year
hosting
this,
so
I
and
he
had
a
little
kit
the
other
day,
and
I
wanted
to
help
out,
like
with
the
community
and
help
hosting
these
these
office
hours
to
to
continue
the
the
tradition.
C
So
with
that,
this
is
a
kubernetes
event,
so
the
code
account
is
an
effect.
Be
please
be
excellent
to
each
other.
This
is
a
judgment-free
zone.
Everyone
had
to
start
some
somewhere
like
myself
and
others
that
didn't
know
what
a
container
was
right
a
few
years
ago,
many
years
ago
for
some
of
us,
so
please
help
out
your
body
having
a
supporting
environment
in
the
channel
in
the
office
hours
channel.
C
While
we
do
other
our
best
to
answer
your
questions,
the
panel
doesn't
have
access
to
your
cluster
would
love
to
debug
your
cluster,
but
you
can,
but
we'll
do
our
best
to
get
moving
down
the
next
steps.
Normally
we
provide
shirts,
but
I
think
that's
you
know
on
hold
right
now.
Panelists
you're
encouraged
to
expand
on
answers
with
your
experience
and
pro
tips.
The
audience
can
help
by
pasting
urls
in
the
office
hours
channel
to
the
official
docs
blogs
anything
that
is
helpful
to
answer
the
questions.
C
If
you
have
questions
post
them
in
discuss,
kubernetes
dot
io,
if
they'll
get
answer
answered,
we'll
pick
them
up
for
the
next
next
month,
but
usually
it's
very
active.
That's
where
we
share
a
lot
of
information
and
get
your
your
questions
answered
and
for
the
people
watching
the
live
stream.
You
can
tweet
this
spread
the
world
work
paying
forward.
The
panel
is
made
of
volunteers.
C
So
with
that,
let's
get
started
and
one
of
the
things
that
we're
doing
with
the
office
hours
this
year
is
having
a
member
of
the
sieg
coming
for
the
few
minutes
at
the
beginning
and
tell
us
about
the
sick
and
what's
going
on
with
your
sick,
and
today
we
have
paris
from
the
contributor
experience,
paris.
E
Hey
all
I'm
so
happy
to
be
here
thanks
so
much
so
what
I
wanted
to
do
today
was
just
talk
a
little
bit
about
contributor
experience.
I
think
that
a
lot
of
folks
think
that
we
just
run
events
and
contributor
summits
and
that's
definitely
a
part
of
it,
but
it's
actually
a
lot
more,
especially
needing
of
engineering
talent
to
come
and
contribute
to
our
sag,
because
we've
run
highly
trusted.
E
Roles
like
github
management
and
those
folks
are
have
keys
to
the
entire
car,
and
when
I
need
car,
I
mean
like
over
300
repositories
that
that
we
manage-
and
we
also
do
a
lot
of
what
I
dub
community
get
ups.
E
So
we
have
a
lot
of
yaml
that
powers
all
of
our
community
groups
and
that's
really
how
we
stay
organized
and
our
crew
with
contrib
x,
takes
care
of
all
of
this
and
maintains
this
and
is
always
looking
for
not
only
new
contributors
but
new
maintainers
we're
actually
in
the
process
of
trying
to
grow
more
leaders
in
our
group
and
these
leaders
manage
over
60
000
contributors
like
I
mentioned
across
300
repositories.
E
E
So
one
of
our
top
level-
things
that
we
manage
is
this
community
repo
and
obviously
the
term
community
is
very
overloaded
in
2022
like
what
does
that
mean
here
and
for
kubernetes
in
this
repository?
E
It
really
means
upstream
contributors
and
maintainers
and
people
who
are
contributing
code,
documentation,
process,
programs
etc
to
the
closest
to
the
maintainer,
which
is
what
we
dub
upstream
and
in
this
is
you'll,
find
all
kinds
of
things
so,
for
instance,
you'll
find
our
contributor
documentation
and
our
contributor
guide,
so
this
houses
so
much
of
kind
of
how
we
are
what
we
do,
why
we
do
it,
including
at
the
root
with
governance
where
cigs
are,
who
cigs
are
how
you
can
find
things
within
the
community
and
all
of
this
means
maintenance.
E
So
you
can
actually
go
in
and
see
in
each
sig.
So
I'm
going
to
pick
contributor
experience
just
because
I'm
selfish
here,
but
each
sig
has
a
readme
and
it
tells
you
what
exactly
they
own
and
in
this
case
they
own
either,
like
I
say,
code,
docs,
process
procedure,
something
or
another
right,
and
these
are
also
projects
that
we
as
community
as
community
experience
run.
E
Oh
and
by
the
way
we
are
run
by
two
awesome
folks
that
share
us,
that's
bob
and
allison,
and
then
we
do
have
two
awesome.
Tech
leads
kristoff
and
nikita.
E
So
shout
out
to
those
folks
when
you
see
them,
but
as
far
as
our
sub
projects
are
concerned,
some
of
our
biggest
ones
have
to
do
with
community
management
and
what
is
it?
What
does
community
management
entail
for
us
here,
pretty
much
how
we
communicate
so
things
like
slack?
E
That
has
a
lot
of
slack
infra
slack
and
for
being
like
customizations,
that
we've
made
to
slack
to
help
us
moderate
and
all
kinds
of
things
that
we
need
engineering,
engineering,
support
for
and
then
all
kinds
of
other
stuff.
So,
like
contributor
documentation,
like
I
mentioned
before,
events
which
I've
I've
said
that
a
lot
of
people
already
know
us
about,
because
so
many
of
our
contributors
have
come
to
contributor
summits
throughout
the
years
we've
had
so
many.
E
I
think
it's
over
nine
now,
but
the
really
cool
stuff,
I
think,
is
really
with
github
management
and
what
we're
doing
there
and
the
automation
of
a
lot
of
processes
that
we
have
and
the
collaboration
that
we
do
with
other
things
like
say,
testing,
sig,
infra
and
trying
to
automate
a
lot
of
the
toil
that
we
have
with
things
like
our
communication
pipelines.
E
So
the
other
thing
that
we're
working
on
this
year
is
this
contributor
site
we
launched
this
a
while
ago,
and
it
has
it's
pretty
much
trying
to
raise
all
of
the
documentation
that
you
find
in
this
repository,
because,
as
you
can
see,
this
is
a
lot
right.
So
how
can
we
organize
all
of
this
information
to
a
visible
place
right?
So
this
is
where
we
need
a
lot
of
ux
support
a
lot
of
engineers,
so,
if
you're
out
there
like,
how
can
I
help
kubernetes?
E
E
Another
thing
that
we're
working
on
is
the
communication
of
how
we
communicate
changes
with
with
the
contributor
experience
and
code
of
conduct,
and
things
like
that
right.
So
another
thing
we've
been
looking
into
is
automation
of
some
of
this
stuff,
so
this
is
actually
one
of
the
first
things
that
we
that
are
excuse
me.
Our
marketing
team
has
been
working
on
with.
How
can
we
automate
tweets
right
like
how
can
our
community
like?
How
can
our
community
support,
like
the
things
that
we're
communicating
out
right
instead
of
just
having
one
person?
E
Do
it
like,
for
instance,
like
one
social
media
manager?
How
can
we
get
out
of
that
out
of
that,
like
one
person
doing
it
paradigm?
So
this
is
the
kind
of
stuff
that
we're
working
on
all
year
and
we
would
definitely
love
to
have
you
all
come
out
and
help
us.
The
cool
thing
is
contributor.
Experience
is
actually
meeting
at
the
top
right
after
this
right
after
office
hours.
E
So,
if
y'all
are
like
hey,
I
just
want
to
drop
into
a
meeting.
That's
totally
cool.
We
do
introductions
for
new
contributors
all
the
time
and
we
would
love
to
have
you
so
we're
doing
so
much
stuff
this
year.
I'm
really
trying
to
keep
the
project
sustainable
and
make
sure
that
we're
running
and
growing
in
the
right
areas.
E
C
You
paris,
one
thing
that
I
would
say
is:
if
you're
not
aware
your
tooling,
that
you're
managing
is
used
by
other
people,
so
it's
open
source
is
in
github
and
people
can
leverage
it.
So
one
example
is
in
k,
native
being
selfish
and
talking
about
always
about
decay
native,
but
the
canadian
project
I
mean
steering-
and
I
didn't
want
to
invent
a
new
tool
to
manage
g
suite
and
to
manage
snack.
C
So
we
are
currently
integrating
your
the
kubernetes
tooling
to
adjust
it
and
configure
it
and
and
use
it
in
k
native.
So
as
an
open
source
project,
you
can
leverage
these
tools
and
we
find
things
that
are
quite
hard-coded
for
kubernetes
things
like
that,
we'll
assume
it
pull
requests
or
issues
saying
if
you
change
this
variable
or
that
variable
will
make
it
easier
for
the
next
project
in
cncf
to
leverage
it
right
and
that
way
you
get
one
way
of
getting
maintainers
right
to
fix
a
bug
or
pr's.
C
So
we
just
wanted
to
shout
out
to
contributes
that
those
tools.
Actually,
you
are
used
beyond.
E
Current
yeah,
they
I,
I
think
our
folks
love
to
hear
that
stuff,
honestly
so
yeah
and
any
any
pull
requests
that
you
have
from
features
that
y'all
have
worked
on,
submit
them
up
to
see
if
it
works.
That's
the
whole
beauty
of
open
source,
so
yeah,
I'm
so
pumped
to
hear
that.
C
I'm
working
on
g
suite
because
yeah
we
don't
want
to
write
that
code.
E
Yeah,
and
so
we've
got
like
I
mean
when
I
tell
you
hundreds
of
mailing
lists
y'all,
I
mean
hundreds
like
we
actually
kind
of
stopped
counting
at
certain
at
certain
points,
and
these
are
all
hosts
on
google
groups
and
they're
all
free
groups
right
and
we're
trying
to
take
all
of
this
stuff
and
put
it
on
the
kubernetes
domain.
So
the
first
thing
that
we've
recently
done
is
is
taken
kubernetes-dev
at
google
groups,
which
was
about
6
000
contributors,
and
we
migrated
them
to
dev
at
kubernetes.io.
E
So
that's
just
the
first
thing
on
our
on
our
purview
there
for
like
automating
of
all
of
google
groups,
but
we
have
a
lot
more
work
to
do
and,
like
I
said,
we
could
definitely
use
the
support.
So
this
is
like
great
learning
too,
if,
if
you're
working
for
an
employer
who's,
you
know
talking
about
automation-
and
you
know
automating
away
kind
of
that
toil.
This
is
a
great
place
to
learn
so
definitely
come
on
down.
C
Thank
you
so
much
for.
For
speaking.
I
know
you
need
to
drop,
but
anyone
has
any
any
questions
to
shout
out
for
paris
before
she.
She
leaves
from
the
panel.
D
E
D
D
C
Thank
you
so
much,
and
if
you
want
to
learn
more
all
the
k
k,
it's
k,
a
s
dot
dev
is
the
hub
that
I
send
everyone
through
and
you
will
find
all
the
information
there
on.
All
the
sikhs
and
country
backs
is
one
of
them.
C
So
thank
you,
paris.
So
how
are
we
doing
with
time?
So,
let's
go
to
the
questions
david
capture,
a
few
of
them
and
go
for
the
first
one
and
the
first
one
is
from
when
I
pronounce
the
name
so
the
the
first
one
is
put
to
paraphrase:
how
do
how
do
we
downgrade
or
maybe
install
a
specific
version
of
kubernetes
with
keep
adm.
C
D
I
could
do
this
one
unless
someone
else
is
really
keen
all
right,
so
qbm
and
cubecontrol
they're
all
installed
with
the
versions
that
they
provision
themselves
like.
So
when
you
do
an
apt,
install
a
yum
install
a
zipper,
install
pacman,
install
whatever
your
your
flavor
of
linux
is
the
version
of
qbdm
that
you
have
on
the
machine
is
the
version
of
kubernetes
that
will
be
provisioned
when
you
run
the
cube
admin
command.
C
And
one
one:
one
thing
with
this
question
is
installing
a
specific
version.
If
the
person
doesn't
know
right,
if
you're
using
apt,
you
will
pass
the
specific
version
like
keep.
Adm
equals
1.23.2.
C
And
you
will
get
that
version
and
get
the
cubelet
that
matches
that
version
and
you
install
a
specific
version,
but
a
question
for
you
is
so
maybe
a
trick
question
is:
why
would
you
downgrade
and
and
if
that
is
common
thing
right
and
would
you
do
it
in
production
and
that's
first
question?
The
second
question
is:
can
I
upgrade
or
downgrade
like
multiple
major
minor
versions
like
going
from
1.18
to
1.23
or
the
proper
ways
to
do
it
step
by
step?
C
D
So
you
can
danger
it.
I
wouldn't
say
it's
encouraged.
Of
course
you
know.
Hopefully
you
should
always.
D
Well,
if
any
of
the
databases,
stuff
changes,
schemas
paths,
nfm
of
ncd
and
you're
downgrading,
then
the
chances
are
that
there
may
not
be
a
rollback
option
for
that
or
you.
If
there
is,
you
potentially
lose
any
data,
it's
the
same
problems.
You
would
have
with
your
own
applications
when
downgrading
typically
it's
best
to
try
and
avoid
that
as
much
as
possible.
D
D
E
D
Previous
versions,
I
believe,
like
you,
could
in
theory,
you
will
be
getting
security
apaches
up
to
three
versions
back.
G
Sometimes
there
are
deprecations
something
we're
trying
to
notify
two
three
releases:
I
had
that
this
feature
will
be
deprecated
or,
like
some
major
changes
in
apis.
G
This
is
usually
reflected
in
the
release
notes
and
for
the
people
who
missed
those
and
they
didn't
read
it
they
get
in
trouble
during
during
the
upgrade
is
always
recommended
to
start
with
the
dev
clusters
upgrade
c.
You
know
if
all
apis
are
working
and
then
go
to
the
staging
and
production
clusters
to
avoid
the
issues
like
downgrading
stuff.
Well,.
D
C
So
I
mean
1.24
and
that
will
be
the
famous
version
that
docker
stream
will
be
removed
for,
and
people
will
remember,
people
are
not
talking
about
now,
but
they
will
be
talking
about
after
april.
Next
next
question:
I
think
that's
that's
a
good
one.
I
think
the
person
doesn't
know
that
you
can
specify
specific
versions
to
adm
and
also
it's
good
to
know
for
the
certification
exams.
There's
questions
around
upgrading
and
putting
specific
versions
of
qa
dm
you.
You
can
get
specific
versions.
C
The
next
question
is
this
is
an
interesting
one.
How
can
I
get
the
host
name
that
my
pod
is
deployed
to?
I
think
this
is
more
on
the
node
how
to
get
the
hostname
of
the
machine
or
node
in
which
the
node
at
the
bot
are
deploying
kubernetes
host
ip
we're
able
to
retrieve
that
by
using
the
environment
variables
in
the
ammo.
But
how
do
I
get
the
hostname
of
the
node?
C
G
G
C
I
think
I
think
the
question
is
more
of,
for
example,
if
you
run
and
again
I
call
it
kip
cto,
you
call
it
banana
cube.
Ctl
get
part
dash
or
wide
that
will
give
you
the
parts.
There
will
be
a
column
with
the
names
of
the
nodes.
It
depends
on
the
kubernetes
cluster,
how
it
was
configured.
Sometimes
the
the
noted
name
will
be
an
ip
address.
C
Sometimes
the
the
admin
can
configure
the
node
name
to
be
a
host
name
that
matches
the
the
operating
system
host
name
and
some
others
will
be
like
a
long
domain
name
that
distillate
on
the
known
name.
So
at
least
I
know
it's
few
providers
that
you
do
keep
ctl
get
but
touch
a
white
and
you
get
ip
addresses
and
you
will
know
this
is
like
if
you're
troubleshooting,
like
where's,
my
body
is
located
like
if
it's
going
to
the
right
one
or
not
the
right
one
and
you're
trying
to
throw
it
shoot.
C
But
I
think
this
this
person
is
like
that
column
is
displaying
an
ip
address
and
they
want
to
get
the
host
name.
I
think
that's
kind
of
like
a
corner
case.
Environ
scenario
will
be
like
you
will
need
to
like
run
something
on
the
host
node
like
debug
or
like
run
a
pod
that
has,
you
know,
privilege
and
host
and
get
the
host
name
configured
in
the
operating
system.
If
you
really
want
it
or
they
can
just
do
a
dns,
you
know
a
reverse
lookup
of
that
ip
to
a
domain
name.
Well,.
D
Is
at
all
time
low
and
try
not
to
use
them,
and
but
kubernetes
has
a
concept
of
a
downward
api
which
will
allow
metadata
from
them
the
pod
and
the
node
on
which
you're
running
to
be
injected
into
the
container
via
fails
or
environment
variables,
which
will
allow
you
to
get
the
node
name,
not
necessarily
the
host
name.
But
you
will
get
what
kubernetes
understands
to
be
the
name
of
that
node.
C
Yeah
exactly
that's
what
I
said
like
it
depends
how
you
configure
that
node
name
and
people
by
chance
can
be
the
host
name
or
it
could
be
the
ip
address,
or
it
could
be
something
else,
but
if,
if
they
have
control
over
their
kubernetes
environment,
the
the
answer
could
be
like
configured
in
a
way
that
the
known
name
is
the
host
name.
C
If
you
care
about
that
that
thing,
but
with
no
more
details
on
like
the,
why
right,
why
would
you
need
that
information
and
also
archie
said
like
tried
not
to
care
where
that
part
is
sitting
today,
because
he
might
be
sitting
somewhere
else
and
and
as
such,
but
if
you're
troubleshooting
something
might
may
be
useful,
that's
that
that
was
a
good
one.
Thank
you,
archie
for
and
others
to
contribute
to
that
one.
The
other
one
is
what
okay.
So
this
one
I
it's
also
recently.
C
We
did
some
changes
in
kubernetes
about
this.
One
is
what
is
the
difference
between
server
side
and
client-side
apply?
C
If
I'm
wondering
why
there's
a
difference
between
the
usage
of
dry
dry
run
in
the
client
and
the
api
using
the
cli
have
to
dry
round
scratch
creation
modes,
you
can
do
a
client
or
server
must
be
non-server
or
client.
If
client
strategy
only
prints
an
object
that
will
be
sent
without
sending
it
and
if
server
strategy,
it
will
submit
the
server
side
request
without
persisting
the
resource,
which
I
think
the
person
is
answering
the
question.
C
On
the
other
hand,
when
sending
a
drive
run
request
to
the
api
we-
and
this
is
where
I'm
sure
api
it
only
supports
all.
I
would
expect
that
both
client
and
api
will
support
the
same
dry,
run
creation
modes,
client
server
and
all
I'm
not.
I
understood
the
first
part.
I
don't
somebody
has
a
clue
about
this.
All
in
dry,
run
and
server
side
apply.
C
I'm
I'm
familiar
with
okay,
this
one
in
terms
of
troubleshooting
web
hooks
like
if
you
do,
the
server
side
apply
to
like
if
you're
I'm
coding,
a
mutation,
webhook
validation,
workbook.
I
want
like
to
test
my
code.
I
will
do
a
driver
on
server
side
to
to
exercise
and
see
that
debug
in
my
my
web
hook.
If
I
don't
do
server
and
I
do
client,
which
are
like
all
the
tutorials
and
everywhere
it
says
you
always
use
client,
nothing
goes
to
the
api
server.
A
And
people
can
shout
out
when
I
inevitably
get
it
wrong
again.
So
I
think,
from
from
my
view,
like
the
key
benefit
of
of
server
side
over
client
side
apply.
Is
that
you,
as
an
end
user,
trying
to
apply
the
yaml
files
to
the
cluster
with
client
side,
can
end
up
with
your
files,
not
actually
matching
the
the
desired
stat,
the
state
that
is
in
the
cluster?
So
you
can
maybe
be
debugging
a
problem
and
going
what?
What?
A
Rather
than
previously
you
just
getting
not
being
informed
of
that.
Instead,
the
server
will
come
back
and
tell
you
you
can
then
or
you'll
get
a
conflict
as
and
and
you
can
choose
conflict
right.
There's
a
number.
I
can't
remember
how
many
different
conflict
resolution
methods
are
supported
these
days,
but
you
can
choose
a
resolution
method
for
the
server
to
to
deal
with
those
conflicts
then,
but
like
back
in
client
side
days,
instead,
you
just
get
that
applied
and
you'd
get
you'd,
be
overwriting.
A
C
Yeah
yeah,
it's
very
and
that's
why
it's
very
common
in
like
tutorials
and
things
like
that.
You
want
like
the
minimum
yammo
right,
you
do
the
client
side
of
a
pod
that
you
put
like
this
is
the
container
that
I
want
to
run,
but
actually
the
server
when
you
send
it
to
the
server
it
would
start
inserting
all
these
defaults,
like
the
service,
account
the
security
group.
Everything
else-
and
maybe
you
don't
want
the
hard-coded
thing
in
your
your
get
in
your
git
right
and
configuration
so.
D
C
Yeah
and
for
people
that
are
like
having
long
history
with
kubernetes,
it
used
to
be
a
time
that
when
you
did,
cuba
city
apply
that
information.
The
client
did
all
the
things.
That
would
be
the
differences
I
put
it
into
an
annotation,
then
later,
I'm
going
to
say
maybe
1.18
or
1.18,
it
was
a
server
side
apply,
and
then
it
was
the
manage
field,
because
if
you
had
like
a
controller,
let's
say
something
like
argo
or
somebody
that
is
managing
that
resource
and
also
keep
ctl,
does
a
change
to
it.
C
They're
fighting
over
a
field
like
number
of
replicas,
so
the
managed
field
was
a
way
of
delineating.
I
own
the
replica
count,
and
then
this
other
controller
owns
the
another
piece
of
the
spec
and
and
that
information
is
managed
field
and
the
feedback
from
the
community.
Some
feedback
was,
like:
I
hate
vanish
view.
It's
like
all
this
clutter.
C
I
have
the
managed
fields
and
of
course,
we
created
plugins
right
to
delete
it
because
we
didn't
want
to
see
it,
but
it
was
useful
information
in
terms
of
of
managing
having
like
two
controllers
or
two
things.
Controlling
one
resource
and
different
parts
controlled
by
two
controllers
and
cubesat
is
a
controller
right.
It's
like
controlling
it
with,
which
is
a
client,
I'm
still
lost
on
the
all,
and
I'm
not.
I
don't
know
what
driver
on.
But
I
don't
know
I
don't
know
what
all
means.
H
Well,
I
was
wondering
there's
another
aspect
of
the
client
server
differences.
I
was
wondering
it
comes
into
play
or
not
when
you
have
a
client-side
apply
and
you
have
like
web
hooks
that
trigger
about
admissions
or
certain
things
that
are
allowed.
The
client
can't
check
those
right.
You'd
have
to
have
a
servant
side
to
even
check
those,
so
I
think
some
of
your
depending
what
you're
submitting,
if
you
don't
do,
a
server-side
dry
run.
It's
not
valid
at
all.
C
Exactly
that's
why
I
said
that
my
example
of
like
when
I
was
writing
a
little
mutation.
Webhook
actually
was
the
one
for
cluster
that
I
created
for
david.
The
way
to
debug.
It
was
like
you
know,
force
it
to
go
to
the
server
side,
but
I
didn't
want
to
create
the
resource.
C
I
just
wanted
to
check
my
code
to
exercise
it,
so
it
was
a
dry
run,
hit
the
server,
do
all
the
validation,
authentication
mutation
and
give
me
back
the
resultant
thing
of
like
all
these
mutation
web
hooks
doing
stuff
to
the
cr,
and
then
I
see,
like
my
my
code,
actually
did
the
mutation
that
I
wanted
to
debunk
david.
Yes,.
G
Yeah,
I
think
if
there's
any,
you
know
follow-ups
on
that
question.
I
need
to
please
let
us
know
and
we
can
try
to
connect
you
with
the
right
people
as
well
to
talk.
C
And
then
we
put
a
link
to
the
to
the
docs
saying
when
you
when
you
set
the
drive
run,
equals
all,
and
it
goes
with
with
a
long
explanation
on
that,
so
we
can
put
that
link
into
into
there.
I
don't
I
don't
want
to
read
this
live
here,
understand
it,
but
the
driver
on
all
this
side.
You
say
we'll
we'll
be
more
on
that
so
moving
along.
C
Let's
take
another
quick
question
from
jorge:
do
you
use
any
any
more
thoughts
on
dragon?
It's
very
useful
to
get
the
ammo
back
in
the
certification
exams,
you're
studying
for
your
cka
dck,
you
become
friends
with
the
kip,
ctrl
and
dry
round.
C
And
and
it's
a
it's
a
good
way
to
get
the
right
version
of
the
ammo
that
is
matches
your
version
of
kubernetes
right.
Some
people
ask
me
like
what
emma
do.
I
use
like
I'm
not
trying
to
get
documentation.
I
just
want
to
use
the
one
that
the
cluster
gives
me
back.
So
the
next
one
is.
Do
you
split?
This
is
a
good
one.
C
Why
and
when?
So
the
actual
question
is
hi
all
new
to
kubernetes
here
and
still
working
on
learning
structuring,
architecting
it.
One
of
the
one
of
the
reasons
we
are
interested
in
kubernetes
is
not
just
as
a
great
way
to
scale
our
custom
applications,
but
also
because
we
feel
this
would
be
a
great
way
to
help
us
scale
maintain
some
portions
of
application
infrastructure
right
use
case
for
kubernetes
by
the
way
things
like
rabbit
and
q,
kafka,
etc.
I'm
guessing
that
it's
like
stateful
things,
right,
databases
for
those
type
of
components.
C
There
are
no
strict
applications
or
services.
Do
you
deploy
them
to
their
own
server
cluster,
or
do
you
mix
them
in
as
just
delineate
among
namespaces,
or
something
like
that?
So
I
think
it's
around
guidance
when,
if
you
have
like
gravity,
let's
take
driving
queue
and
kafka,
do
you
deploy
it
in
a
separate
cluster
or
on
the
same
cluster,
with
the
app
when
why
and
when.
G
I
think
the
other
he
means
also
like
in
general,
like
multi-tenancy
question.
Like
would
you
do
this?
You
know
kubernetes
multi-tenant.
Why
and
when?
So
I
think
it's
a
very
big
topic
but,
like
I
don't
know
like
I,
I
think
it
might
be.
Also.
The
answer
depends
where
the
the
person
who
is
trying
to
run
applications
sitting,
if
he's
sitting
on
prem
on
bare
metal
or,
if
he's
sitting
in
the
cloud,
I
think
the
answer
could
be
slightly
different
as
well.
C
It's
more
on
databases
right
like
to
to
pinpoint
on
the
he's
worried
about.
Do
I
put
my
rivalry
in
queue,
my
kafka
or
my
sequel.
H
Do
go
ahead
chris,
my
my
first
thought
is:
always
don't
run
the
database
in
the
kubernetes,
if
you
don't
have
to
honestly,
but
I've
done
this.
Both
ways
like
I
I've
had
the
multi-tenancy
thing
with
the
app
and
the
database
and
the
cluster
I've
had
it
where
they're
they're
split
out.
It's
really
down
to
how
you
manage
things.
What's
your
get
up
strategy?
Your
are
people
developing
in
this
cluster,
because,
if
they
are,
you
probably
don't
want
production
apps,
also
in
the
same
cluster,
it's
it's
so
situational.
E
F
E
F
So
I
come
at
this
from
kind
of
a
network
performance
perspective.
There's
latency
to
be
considered
here
right.
If
you
have
something
in
another
cluster,
you're
traversing
from
one
cluster
to
another,
maybe
hopping
multiple
times,
maybe
you're
in
the
same
region.
Maybe
you're,
not
availability
zones
right,
there's
there
might
be
latency
involved.
So
all
the
other
answers
are
correct,
but
this
is
another
factor
that
you
need
to
consider
in
for
your
application.
I've
worked
in
environments
with
like
graphql
and
with
payment
processing
we're
like
that.
F
They
actually
are
very
sensitive
to
latency,
and
so
that's
something
that
has
to
be
evaluated.
I
will
say
as
well-
and
I
think
we
all
know
this.
Any
pattern
is
basically
supported.
You're
supported
whether
or
not
you
want
to
run
it
as
a
side
car,
whether
or
not
you
want
to
run
its
own
name
space,
whether
or
not
you
want
to
run
in
a
different
cluster.
You
can
use
link
or
d.
You
can
use
psyllium
to
connect
those
clusters
to
get
like
there's,
there's
so
many
different
options
now.
F
So
I
think
a
lot
of
it
is
what
is
what
works
for
you?
What
is
easy
to
manage
long
term?
How
do
you
measure
and
what
does
your
observability
look
like
and
then
the
the
long-term,
like
maintenance
security
support?
You
know,
is
your
team
ready
to
take
this
on
in
terms
of
managing
multi-cluster
communication?
That's
a
whole
nother
journey
for
a
lot
of
organizations
right
so
like
it's
such
a
multi-passive
answer
and
there's
no
right
or
wrong
way.
F
It
really
depends
on
your
organization,
so
I
I
can
say
for
the
most
part,
like
in
development
environments,
we
actually
like
to
keep
for
what
we
do
in
our
use
case,
we
like
to
keep
things
close
with
the
workload
and
let
developers
actually
in
their
own
ephemeral
environments,
and
so
we
like
to
use
v-cluster
and
from
loft,
which
is
a
great
tool
and
use
our
clusters
a
little
bit
more
densely
right
and
so
we're
not
worried
about
spawning
clusters
and
low-level
infrastructure
pieces.
F
We
have
a
few
clusters
that
do
many
things
and
are
nimble
in
that
way
and
our
developers
we
give
them
that
that
tool
set
and
then
for
production
workloads,
we'll
have
a
more
secure,
definitive
template
focused.
You
know
maybe
an
rds
instance
or
elastic
cache,
whatever
the
database
type
of
thing
might
be
with
the
infrastructure
component,
and
we
we
do
some
testing
of
that
as
well,
because
there
again
there's
latency,
that's
a
different
sort
of
thing
than
a
redis
instance
in
your
local
ephemeral
environment
right
in
kubernetes.
F
C
I
love
this
question
yeah,
so
I
I
I
I
agree
with
you
the
first.
The
first
thing
to
think
about
is
it's
their
requirements
and
the
the
thing
that
that's
why
I
was
saying,
like
the
two
clusters
like
right
to
each
other
right,
if
you
have
two
nodes
that
are
like
sitting
to
each
other,
it
just
happened
to
be
belong
to
different
clusters.
C
That'll
be
the
clusters
that
you
will
get
to
that
communication
between
the
app
talking
to
its
performance
is
the
first
one
like.
If
that
doesn't
work
for
you,
because
it's
too
slow,
then
maybe
you
go
and
have
the
pods
both
parts
sitting
on
the
same
node
that
you
need
that
performance
like
ipc
communication,
that
that
the
the
your
web
app
is
talking
to
the
database.
C
Both
spots
are
sitting
on
the
same
node,
so
it's
almost
ipc
communication
and
all
the
data
is
almost
cash
in
memory
right.
That
would
be
like
the
ultimate
ultimate
requirement,
like
this
person
needs
the
highest
performance
possible.
But
then
you
pay
the
price
like
that
type
binding
over
there.
If
you
don't,
you
don't
have
that
requirement
like
the
latency,
the
performance
and
these
things
things
be
together,
you
can
have
them
together
in
the
cluster
that
is
going
to
be
the
product.
One
prop
two
part
three,
but
the
other
one
staging
dev
right.
C
You
can
have
them
separately.
I
would.
I
would
tend
to
say
that
the
next
level
between
performance
is
organization,
people
and
people
roles.
So
I've
talked
to
people
rose
that
it's
like
this
team
over
there
with
that
vp
and
that
director
they
manage
databases
and
they're
dbas
and
they're
experts
on
databases
and
perform.
C
You
know
performance
of
databases,
tuning
the
storage,
the
backups
and
dr's,
and
all
that
and-
and
I
have
come
to
situations
where
I
say
like
well-
maybe
that
team
should
have
a
cluster
themselves,
manage
the
data,
the
kafka
cluster,
the
ravencue
cluster
and
and
offer
the
apps
team
as
a
service,
a
database
asset
service,
a
rav
mq
as
a
service,
and
then
the
app
stream
says
like
I
need
a
kafka
server.
I
don't
know
where
it
is
so
this
other
team
is
going
to
offer
it.
C
They
just
happen
to
have
that
cluster
and
then
the
apps
team
have
their
cluster,
so
both
can
have
different
versions
of
kubernetes.
Both
can
have
like
their
own
life
cycle
of
updating
things,
because
that
where
it
gets
into
trouble
like
this
thing
wants
to
update
kubernetes
version,
and
this
team
is
like
we're
not
ready
yet.
So,
if
you
have
the
different
versions
of
kubernetes
like
122
versus
123,
then
it's
an
org
and
the
the
people
responsibilities.
C
At
the
end
of
the
day,
maybe
the
two
clusters
are
sitting
in
the
bpc,
the
same
bpc
and
they're
like
very
connected,
but
they're
managed
differently
by
different
teams
and
that,
I
would
say,
put
them
in
different
in
different
clusters,
not
because
of
the
cluster,
but
because
of
the
people
and
the
managing
like
that
team
is
going
to
offer
you
a
sas
solution,
but
it's
maybe
sitting
in
bare
metal,
but
it's
often
going
to
offer
you
a
sas
solution
of
a
database
of
that
stateful
thing
and
the
apps
thing
is
like.
I
just
need
a.
C
C
Any
thoughts
on
that
because
I
like
I
like
clusters
to
be
throwaway,
I
want
to
treat
them
like
throw
away
like
even
for
that
team
that
that
managed
the
driver
in
queues
like
yeah.
You
just
draw
away
the
cluster
and
get
a
new
one.
You
should
have
like
that
storage
somewhere
right
and
get
a
new
version.
Everything
should
work
infrastructure
code,
everything
defined
as
infrastructure
as
code,
but
not
all
things
have
that
level
of
expertise
right.
C
So
hopefully,
hopefully
that
that
helps,
but
for
the
dev
and
the
mini
cube,
you
don't
have
to
have
two
clusters,
just
put
them
together
to
develop
your
app.
A
I
don't,
I
don't
think
I've
really
got
much
more
time
on
the
in
in
terms
of
the
like
making
choices
about
how
to
divide
applications.
But
I
think
one
thing
that's
worth
thinking
about
and
obviously
far
harder
for
for
staple
stuff
is
like
that,
where
your
fault
zones
are
and
how
tolerant
you
are
to
failures,
or
you
know,
if
you're,
if
you're
working
on
a
platform
team
like
I
do,
how
do
you?
A
How
do
you
ensure
that
you
can
take
clusters
down
for
downtime,
still
have
resiliency
to
like
failures,
whether
it's
hardware
or
cloud
provider
or
networking,
and
still
still
have
you
know
enough
of
an
application
able
to
serve
data
or
serve
requests,
etc?
And
that's
that's
obviously
really
hard
to
do
for
databases
for
stateful
microservices
generally
pretty
easy,
but
then
maybe
that
feeds
into
your
decision
about
where,
where
you're,
locating
your
databases,
if
you've
got
your
application,
spread
across
three
clusters
already.
C
Yeah
and
then
for
the
apps
team
or
the
stateless
apps
team.
I
I
always
encourage
if,
if
that,
that
cluster
of
kafka
or
raring
queue-
or
you
know,
database
is,
is
in
the
same
happen
to
be
in
the
same
cluster
three
tweeted
as
is
not
in
the
same
cluster.
Don't
assume
that
is.
You
can
use
the
the
dns
resolution
as
a
service
local
right,
if
you
treat
it
as
yeah.
My
app
just
needs
a
secret
and
it
has
the
connection
information
in
there.
C
It
might
be
a
a
service
name
or
service
ip
and
it
may
be
in
the
same
name
space,
but
maybe
in
a
few
months
it
could
be
a
sas
container
right.
Serverless
containers,
that's
a
thing,
and
it
just
have
the
connection
information
to
to
a
database
or
router
and
queue,
and
that
app
may
not
be
running
kubernetes.
It
may
be
running
something
on
top
of
kubernetes
as
a
sas,
so
it's
easier
to
move
that
container
around.
C
If
you
treat
your
connection
information
as
as
configuration
information,
it
could
be
secret,
config
or
config
maps,
and
not
assume
that
it's
going
to
be
like
mine,
my
namespace.service.cluster.local
right
or
dbdb.that,
using
that
dns
resolution
is
to
to
to
so.
You
have
the
option
later
of
running
the
database
in
dev
in
the
same
cluster
or
running
in
production.
A
separate
cluster,
or
maybe
the
app
ends
up
running
in
something
that
is
on
top
of
kubernetes
as
a
sas.
D
At
this
time
slot-
yes,
I
I
think
this
works
well
for
for
europe
and
most
of
the
us
and
we'll
look
at
alternative
options,
because
we're
supposed
to
be
running
a
us
one,
eu
one,
but
you
know
it's
difficult-
to
manage
and
operate.
So
this
is
what
works
well
for
both
and
if
we
get
some
any
interest
from
apac
and
other
areas,
then
we'll
do
our
best
to
try
and
get
something
scheduled
for
there.
But
this
is
their
our
new
time.
Slot
yeah.
C
Cool,
so
with
that,
I
think
we
end
and
feel
free
to
hang
out
in
our
office
hours
afterwards
in
the
slack
channel.
There's
always
people
there
and
we'll
be
back
at
the
same
time
next
month,
and
if
you
were
in
pacific
time-
and
you
were
complaining
that
this
session
was
too
early
for
you
now-
you
can
be
at
it
will
be
7
a.m
during
time.
So
you
can
bring
your
coffee
and
breakfast
and
watch
the
office
hours.
You
cannot
complain
now
right.
C
So
with
that
we'll
end,
thank
you
for
for
coming
and
thank
you
for
for
the
panel.
As
always
all
right,
hi.