►
From YouTube: TGI Kubernetes 147: CoreDNS
Description
Come hang out with Luke Short (@ekultails) and learn about CoreDNS. Luke works with a lot of customers directly to make Kubernetes work for them. Over the past week he has been going down a deep rabbit hole around CoreDNS and will share those experiences and learnings with you! Come along for the ride!
Show notes: https://github.com/vmware-tanzu/tgik/tree/master/episodes/147
Timestamps:
- 00:00:40 - Welcome to TGIK!
- 00:06:03 - Week in Review
- 00:24:55 - Katacoda Kubernetes Playground
- 00:26:33 - CoreDNS Introduction
- 00:30:56 - CoreDNS Plugins
- 00:32:38 - CoreDNS Internal Plugin: File
- 00:34:28 - CoreDNS DNS Records
- 00:37:19 - Story time! Luke's first container was a DNS server
- 00:38:44 - Kubernetes at Scale: https://openai.com/blog/scaling-kubernetes-to-7500-nodes/
- 00:39:34 - Chat - What DNS servers do you use?
- 00:41:38 - CoreDNS Pods on Kubernetes
- 00:43:57 - Luke making spelling mistakes :-)
- Luke thought he was being trolled by him resizing the window. Sometimes that will move the cursor around in the Kubernetes Playground in unpredictable ways.
- 00:45:10 - Moving to Luke's home Kubernetes cluster
- Configuring CoreDNS in Kubernetes
- 00:49:28 - Load balancing with DNS
- 00:51:48 - View all CoreDNS plugins
- Preview the "records" and "unbound" plugins
- Authoritative vs recursive
- 00:54:32 - CoreDNS external plugins require recompilation
- 00:55:55 coredns-unbound project
- 00:56:24 Public container registries
- 00:59:29 How to compile CoreDNS with plugins
- 01:01:22 CoreDNS demo of customizing Corefile on Kubernetes
- 01:21:18 CoreDNS demo of adding customized DNS records
- 01:40:09 Documentation for customizing CoreDNS in Kubernetes
- 01:43:29 Outro
A
Hey
everyone
welcome
to
the
live
stream.
Everyone's
welcome
doesn't
matter
what
shape
size,
gender,
sexuality,
race.
You
are
everyone's
welcome
here
on
tgik
hope,
everyone's
having
a
great
friday,
I
see
we
got
a
few
people
on
chat
already
hello
from
israel,
scottsdale
netherland,
brazil,
wow,
hey
david.
I
see
you
there
thanks
for
joining
in.
A
A
A
Thanks
for
the
best
wishes
sevi,
you
know
it's
pretty
cool
to
be
here
before
I
joined
vmware,
I
used
to
watch
gg
ik
all
the
time
and
it
was
such
a
great
place
to
get
tribal
knowledge
not
just
from
the
host,
but
also
from
like
all
the
people
in
the
live
chat.
All
these
different
experiences
and
if
there's
one
thing,
I've
really
learned
it's
not
about
a
single
person,
learning
from
a
single
person.
It's
really
about
the
hive
mind
the
collective
consciousness
of
a
group
of
really
smart
people.
A
A
A
A
A
A
All
right,
let
me
I'm
going
to
share
out
the
notes.
We
normally
use
a
hackmd,
great
collaboration
tool
for
for
teams
just
to
put
something
down
on
paper.
We
so
we
have
a
hack
md.
That
kind
of
has
different
show
notes
about
different
things
we'll
be
talking
about.
I
normally
see
a
lot
of
people
ask
about
that
in
the
chat,
so
I
want
to
make
sure
everyone
is
kind
of
aware
of
what
I'll
be
following
for.
A
A
A
Yeah,
just
some
basic
things
here:
twitter
linkedin,
I
have
a
blog.
I
started
a
couple
months
ago.
I've
been
really
excited
about
that.
I
have
a
lot
more
blog
ideas
coming
up
if
you're
interested
in
kind
of
seeing
different
projects,
I
work
on
feel
free
to
check
out
my
github,
I'm
very
active
on
there
in
the
open
source
community,
very
passionate
about
open
source
actually
and-
and
also
I
I
open
source,
the
majority
of
my
technical
notes.
A
A
These
stable
notes,
I
publish
those
out
every
three
months,
so
bookmarks
are
less
likely
to
break
and
then
my
latest
notes
I
have
these
automatically
deployed
using
a
ci
cd
pipeline.
It
goes
straight
to
my
kubernetes
cluster.
So
if
you're
looking
for
the
latest
and
greatest
notes
about
what
I'm
working
on
and
looking
into
lately,
you'll
want
to
check
out
the
latest
link.
A
A
Some
notable
things
with
that
we
have.
The
cluster
api
is
being
bumped
up
a
version
from
alpha
three
to
alpha
four
and
the
summary
of
what's
going
on
at
the
back
end
is
just
more
improvements
to
user
experience
and
as
well
as
just
cleaning
up
additional
bugs
and
such
something
I
found
very
fascinating,
was
that
about
a
year
ago,
kubernetes
had
completely
redid
its
policies.
A
A
A
Which
sig
covers
ingress
sig
networking?
I
would
guess
so
I
I'm
not
100
sure
on
that
yeah
there's,
I
I
don't
know
if
it's
the
the
same
sig
or
not,
but
there's
there's
a
kind
of
an
ingress
2.0
in
the
works
called
the
gateway
api
that
I'm
I'm
very
excited
for
I'm
keeping
an
eye
on
that
ingress
has
some
limitations
about
exposing
your
application
out
to
the
internet,
mainly
that
it's
one
of
the
big
fallbacks
for
me
is
that
it
is
very
biased
towards
http
sites.
A
So,
if
you're
trying
to
like
in
our
instance,
if
you're
trying
to
set
up
like
dns
under
kubernetes
cluster,
it's
a
little
bit
more
complicated
to
get
that
exposed
all
the
way
out
to
the
public
internet
for
end
users.
But
there
are
workarounds
for
for
today
and
we'll
get
into
a
little
bit
of
that
later
as
well.
A
My
first
introduction
to
kubernetes,
at
least
on
a
more
very
experienced
level,
was
through
cold
cloud,
which
I
found
them
through
udemy,
and
you
can,
I
think,
codecloud.com
code
with
a
k
cloud
with
a
k.
You
can
also
find
them
directly
through
their
website,
really
awesome
content
to
help
get
you
into
kubernetes.
A
A
A
A
Oh
and
here's
another
interesting
tidbit,
something
I
actually
learned
about
just
today,
even
as
it
is
right
now,
kubernetes
1.20,
you
can
only
have
one
backend
for
the
load,
balancer
type
for
the
service
api.
So
if
you
want
to
mix
and
match
different
load,
balancers
based
off
of
your
use
cases-
or
perhaps
maybe
there's
some
security
implications
to
that-
you
can't
do
that,
but
it
looks
like
in
1.21
it's
already
merged
in
you
will
be
able
to
now
in
the
spec
of
your
manifest,
say,
hey.
A
A
So
moving
on
talking
more
about
you
know,
what's
going
on
in
the
general
kubernetes
ecosystem,
we've
got
a
few
things
to
note
here.
For
one
thing,
google
has
a
new
autopilot
offering
for
kubernetes.
So
obviously
google
is
is
no
stranger
to
kubernetes
they're,
one
of
the
original
creators
of
it,
in
fact,
and
so
they
have
had
a
kubernetes
offering
since
day,
one
and
or
day
zero,
even
and
with
autopilot
they've,
finally
provide
predicting
a
service.
A
That's
a
fully
managed
private
cloud
for
you,
so
you
never
have
to
worry
about
getting
into
the
internals
of
kubernetes
they'll,
make
sure
to
manage
the
entire
lifecycle
of
the
nodes
and,
of
course,
with
convenience.
There
comes
the
cost.
So
this
is
a
commercial
offering,
and
you
know
it
should
be
interesting
to
see
how
it
is
I'd,
love
to
play
around
with
that
sometime.
A
A
One
of
the
main
focuses
of
it
is
not
running
a
service
until
it's
needed,
so
you
can
keep
your
database
off
store
the
data
in
a
persistent
volume
and
as
soon
as
there
is
a
connection
from
a
user
or
or
some
kind
of
application
to
the
database,
then
it
instantly
spins
up
one
of
the
kind
of
hard
things
about
developers.
Architecting
for
serverless
architecture
is
that
you
need
to
really
have
speed,
because
you
don't
want
the
end
user.
You
don't
want
the
application
just
sitting.
A
A
Redis,
this
is
kind
of
cool.
Redis
is
a
very
popular
kind
of
database,
back-end,
for
especially
key
value
stores,
use
case
in
applications
and
with
their
new
redis
search,
2.0
release,
which
is
an
offering
on
top
of
redis
itself,
they're
able
they're
claiming
over
two
times,
speed
actually
and
some
of
the
benchmarks.
They've
shown
it's,
I
believe,
2.4
times
the
improvement
at
times
yeah,
so
that's
pretty
exciting
for
anyone
who's
using
redis
out.
There
might
be
something
to
look
at
to
adding
to
your
application.
A
A
Yeah
someone's
asking
a
check
for
for
clarification
about
the
node
port
type
of
service.
So
with
that,
let's
say
your
application
is
running
on
worker
node
one,
but
you
also
have
a
worker
node
two
and
a
three.
If
you
expose
a
port
through
node
port,
that
port
will
be
open
on
one
two
and
three,
regardless
of
whether
the
application
is
running
there
or
not.
A
A
A
They
are
on
this
journey
of
moving
away
from
opensnack
to
kubernetes
and
they're
they're
talking
about
a
lot
of
their
challenges.
They
were
having
with
it
so
with
openstack.
One
of
the
challenges
was
scale
as
they
were
trying
to
scale
out
to
bigger
clouds.
They
kept
running
into
a
split
brain
issue
with
rabbit
mq,
and
I
can,
I
can
totally
relate
rabbit.
Mbq
is
a
monster
to
kind
of
deal
with
so
when
it
breaks
it
breaks
horribly,
it
is
a
pain
in
the
butt
to
try
and
shovel
troubleshoot
it.
A
We
probably
don't
have
a
lot
of
good
documentation
up
there
in
the
upstream
community
for
that,
so
there
there's
a
lot
of
struggle.
Probably
part
of
it
is
honestly
just
kind
of
configuration
issues
that
you
know
could
be
dealt
with.
You
know
at
red
hat
we
were
able
to
verify
a
scale
of
the
700
worker
notes.
This
was
really
exciting
for
us
and
I've.
A
A
A
Sepi
says
that
everyone
in
turkey
is
looking
for
some
rabbit,
mq
support.
I
wish
you
the
best
best
of
luck
out
there
yeah
last
thing
to
note
about
sivio.
They
have
a
beta,
that's
ending
soonish,
so
you
know
get
on
this
offer.
While
you
can
it's,
they
have
a
region
in
the
united
kingdom
and
now
in
new
york
city
in
the
united
states
as
well.
A
You
get
70
70
us
dollars
of
free
credit
every
month,
which
basically
equates
to
you
can
get
a
three
three-node
kerbin
index
cluster
with
medium
specs
for
free.
So
that's
what
I've
been
doing
and
it's
fantastic
in
the
background.
It's
using
rancher's
k3s
implementation,
just
just
to
make
things
very
simple:
you
know
they.
They
say
you
know
their
their
startup
time,
for
that
is
pretty
insane,
and
actually
I
spun
up
a
new
cluster.
Just
the
other
day
had
a
had
a
working
kubernetes
cluster.
A
A
So,
for
those
who
don't
know
kind
is
actually
used
by
the
majority
of
upstream
developers
of
kubernetes,
you
get
an
instant
cluster
on
your
develop
machine.
Your
work
laptop
work,
computer
whatever
it
may
be.
It
just
spins
up
a
bunch
of
containers,
containers
that
are
running
the
kubernetes
services
and
then
you
can
then
run
containers
on
top
of
those
containers
using
the
the
integration
that
that
provides
so
yeah
as
a
getting
started
into
kubernetes,
mini
cube
kind,
k3s,
all
great
options
for
just
kind
of
an
introductory.
Getting
your
feet
wet
with
with
kubernetes.
A
A
10
lessons
hour
and
a
half
all
right,
yeah
cluster
api
is
a
very
fascinating
thing,
so
going
back
to
kind.
Actually,
it
kind
uses
the
cluster
api
in
the
background
to
do
its
initial
deployment
of
resources,
and
it
supports
many
different
kind
of
cloud
providers
as
well.
So
you
don't
have
to
just
use
containers.
It
also
has
integration
with
aws
azure,
google.
A
Basically,
all
the
public
cloud
platforms
out
there,
you
can
just
very
easily
integrate
with
and
deploy
kubernetes
cluster,
and
here
at
vmware.
This
is
actually
what
our
kubernetes
deployment
tool
uses
as
well.
So
this
helps
to
set
up
and
manage
infrastructure
and
then,
for
all
those
day,
two
operations
such
as
upgrading
we
just
we
use
cube
admin
kind
of
a
wrapper
around
cube
admin,
which
is
now
the
official
supported
deployment
tool
of
kubernetes
and
the
cn
cf
foundation.
A
A
So
also
you
know
I
I
want
to
take
a
moment
to
just
kind
of
tell
you
all
about.
If
you
don't
know
about
the
katakode
kubernetes
playground.
This
is
phenomenal
whenever
I
just
want
to
just
just
throw
something
up
on
the
wall.
Try
something
out!
This
is
a
fantastic
and
free
place
for
you
to
get
around
and
tinker.
However,
you
want
with
kubernetes,
so
it
gives
you
one
control
play
node
and
one
worker
node
as
well.
A
It
takes
a
few
minutes
to
load
up,
but
when
it
does
sky's
the
limit
and
in
fact
I'd
love
to
just
stay
in
here-
and
you
know
hack
away
at
some
some
core
dns
stuff,
so
that
you
know
you
can
follow
along
at
home.
If
you
want
to-
and
you
know,
try
it
out
on
your
own
time
as
well-
which
is
one
of
the
great
things
about
this-
is
that
it's
just
it's
instant,
it's
free
and
it's
just
so
easy
to
collaborate
with.
A
A
A
Percent
so
backtracking
a
little
bit
yeah.
Welcome
to
today's
episode.
I
mean
we're
gonna
be
talking
about
core
dns
and
you
know
why.
Why
are
we
even
talking
about
coordinating
us?
Why
am
I
talking
about
it?
So
as
part
of
my
journey
towards
getting
you
know,
kubernetes
certification,
specifically
looking
to
get
the
certified
kubernetes
administrator,
the
cka
one
of
the
topics
on
there
is
dns.
So
this
is
a
very
important
thing
to
understand
and
you
know
going
through
online
courses.
I
didn't
want
to
just
understand
the
bare
basics.
A
I
really
want
to
know
the
how
and
the
why
it
works.
You
know
I
want
to
get
into
the
nitty-gritty
detail.
So
that
way
you
know
when
it
comes
to
you
know
not
just
test
time
but
like
real
world
as
well.
I
want
to
be
able
to
understand
what's
going
on
behind
the
scenes,
so
core
dns
is
built
from
the
ground
up
to
be
a
cloud
native
dns
server.
A
A
I
don't
even
know
what
other
ones
there
are
honestly.
If
you're
interested
in
the
cni
certification,
a
container
networking
interface,
calico
well
taijira
the
company
behind
calico.
They
provide
a
free
course
and
a
free
certification.
So
I've
gone
through
that
a
lot
of
it
is
kind
of
the
fundamentals,
but
still
a
great
refresher.
A
great
course
I'll,
probably
end
up
just
finishing
it
and
get
certification
because
hey
it's
free.
Why
not.
A
A
All
right
so
core
dns,
you
know
at
first
I
was
actually
having
a
hard
time
figuring
out
like
where
do
you
even
find
information
about
coordinates
like
like
step?
One
like
I
want
to
know
how
do
you
configure
it
like
what
what
makes
this
thing
tick
and
after
a
little
bit
of
digging
core
dns
manual
here
it
is
boom
everything.
A
Yeah
so
going
in
here-
let's,
let's
see
so
oh
yeah,
plug-ins
plug-ins
are
a
pretty
important
concept
for
core
dns
that
I've
actually
come
to
find
out.
Core
dns
revolves
around
plugins.
So
when
you
compile
core
dns,
it
has
a
set
of
default
internal
plug-ins
that
you
can
use
to
customize
and
add
additional
features
on
top
of
core
dns
to
fit
whatever
your
needs
may
be.
A
A
It's
got
a
plugin
directory.
Let's
see
file's
a
pretty
important
one.
So
file
allows
you
to
have
parts
of
the
configuration
elsewhere
in
a
different
file.
Normally
in
core
dns
there's
a
single
file.
The
the
core
file
is
its
name
with
a
capital
c,
no
space
core
file,
and
that's
where
you
set
all
of
your
configurations
now,
especially
when
you're
getting
into
finding
dns
zones
for
for
different
domain
names
that
you're
going
to
be
hosting.
A
A
A
They
don't
they
don't
actually
have
an
example
of
what
the
actual
dns
record
file
would
be,
but
we'll
see
that
on
the
manual
in
just
just
a
little
bit,
let
me
actually
let
me
jump
to
that
yeah.
So
here's
here's
a
one
of
the
main
things
here.
So
when
you
specify
file
example.org,
you
will
then
literally
have
a
file
next
to
your
core
file
that
says
example.org,
and
in
this
you
can
specify
all
your
records
and,
first
of
all,
the
most
important
one.
Is
the
state
of
authority
record
the
soa?
A
This
says:
hey,
I'm
a
name
server
and
I
am
serving
up
dns
records
for
this
domain
right
here
and
you
can
optionally
set
different.
You
know
timeout
and
refresh
settings
for
for
the
dns
right.
There
won't
go
too
deep
into
that
though,
but
you
can
just.
We
will
do
this
in
just
a
little
bit.
We
just
copy
and
paste
this
example
and
just
use
it
as
a
starting
point
for
making
our
own
dns.
A
And
here
we
see,
this
is
what
it
looks
like
in
the
core
file
itself.
It'll
have
the
domain
name,
let's
say:
hey,
I'm
going
to
use
the
file
plugin,
I'm
also
going
to
use
the
log
plugin,
because
you
want
some
logs
about
your
dns
records.
I'd
hope
if
you
want
to
be
able
to
troubleshoot
some
of
these
things,
there's
also
another
related
logging,
plugin
called
errors.
A
A
Looks
like
mind
yeah,
you
know
what
I
I
must
say.
I've
in
you
know
a
previous
previous
life.
I've
wore
many
hats,
including
being
a
dns
administrator,
and
I
can
tell
you
a
lot
of
the
different
dns
servers
out.
There
are
very
similar.
I
mean
it's
the
same
fundamentals
for
these
kind
of
things.
It's
just
slightly
different
santex,
for
you
know
the
configurations
and
slightly
different
kind
of
backend
implementations
of
it.
A
That
came
out
we're
upgrading
to
and
it
just
didn't
work
on
the
new
version
of
rel
and
obviously
the
optimal
solution
was
to
just
fix
it
to
update
whatever
needed
to
be.
However,
it
was
low
priority
compared
to
other
things
we
had
on
our
plate,
so
by
containerizing
it
into
an
older
version
of
rel
we're
actually
able
to
run
it
without
any
problems.
A
A
How
do
I
orchestrate
it
and
that's
that's
ultimately
where
kubernetes
came
in
and
that's
where
we
started
down
this
journey
of
like
okay,
you
know
we're
a
red
hat
shop,
let's
see
what
they
have,
they
have
open
shift,
and
so
that's
when
we
started
exploring
into
that
and
yeah
sky's
the
limit
for
for
scalability
in
terms
of
kubernetes
man.
I
actually
I
read
an
article
a
few
months
ago
about
some
company
that
reached
some
crazy
scale
with
with
kubernetes
I
don't
for
for
worker
node
count.
A
A
Systemd
resolve
all
the
way.
Oh
yeah,
you
know
many
people
out,
there
have
have
been
system
d
haters
and
I
I've
loved
systemd
from
the
beginning.
I
was
like
you
know
it's
it's
not
too
bad.
You
give
me
a
shell,
you
give
me
system
d
and
you
give
me
a
linux
kernel
and
I
got
an
operating
system
that
I
that
is
usable
like
that's,
that's
not
too
bad,
although
I
I
do
agree,
it
doesn't
do
everything.
A
A
In
fact,
I
have
I've
actually
taken
this
and
implemented
my
own
home
dns
server
that
I
hope
to
kind
of
take
some
of
those
learnings
and
and
share
with
you
all
as
well.
It's
all
pretty
exciting!
So,
let's,
let's
actually
jump
on
a
control,
plane,
node
and
let's
see
what
this
means
in
like
a
kubernetes
contacts,
so
I'm
going
to
use
a
cube,
ctl
command,
we're
going
to
check
out
the
namespace
cube
system.
A
A
We
have
like
the
cube
scheduler
for
helping
tell
which
worker
nodes
to
start
containers
on
cue
proxy,
for
helping
out
with
networking
keep
a
live
d.
I
assume
this
is
probably
part
of
maybe
a
service
load.
Balancer
flannel
is
the
c
and
I
plug
in
the
the
networking
plugin
here,
but
what
we
really
care
about
is
core
dns.
A
So
here
we
got
two
pawns,
which
is
the
default
replicas
for
for
coordinates.
Two
pawns
are
running
and
we
can
tell
by
this
random
gibberish
at
the
end
of
it.
It's
probably
actually
a
deployment,
and
not
just
a
container
so
deployments
are
cool.
I
prefer
to
use
deployments
whenever
possible,
as
opposed
to
a
pod,
because
that
way
you
get
some
life
cycle
management
with
it.
You
can
say
hey.
A
I
want
to
deploy
a
new
version
of
this,
and
then
it
will
take
down
one
of
those
pods
take
down
one
of
those
core
dns
pods
in
this
example
still
running
the
old
version,
loads
up
a
new
version
and
then
takes
down
the
old
version
and
loads
up
a
new
version.
The
end
users
who
are
using
this
as
a
dnf
server.
A
A
A
A
A
A
A
A
And
now
you
can
actually
you
can
do
operations,
not
not
that
I'm
going
to
do
this,
but
you
could
do
either
an
apply
on
it
to
you
know.
If,
for
some
reason
you
delete
it,
you
can
just
apply
it,
you
can
delete,
you
can
apply
if
you
modify
this
file
in
most
fields,
if
you,
if
you
modify
it
for
a
deployment
and
if
you
apply
it
again,
it'll
save
it
on
top
of
what's
already
there.
A
A
A
Integration
will
send
logs
to
your
prometheus
server,
and
this
is
also
this
is
how
pods
get
their
their
networking,
not
just
between
services
with
each
other,
but
also
external
dns,
so
by
default,
in
this
case
the
accentureresolve.conf,
your
default
dns
resolvers
on
your
worker
nodes,
whatever
those
are,
those
are
what
your
container
is
going
to
be
using
for
lookups,
and
maybe
that's
not
what
you
want.
You
know
this,
that's
something
we
can.
We
can
try
changing
in
just
a
second
maximum
of
a
thousand
connections,
cache
queries
for
for
30
seconds
loop
and
reload.
A
A
A
A
A
Oh
man,
yep
the
yep,
the
the
cube
system
had
a
had
a
little
dash.
I
couldn't
backspace
it
I
I
tried
to
it.
Wouldn't
let
me
the
the
katakota
playground
it's
great.
When
it
works,
it's
not
great.
When
it
doesn't,
maybe
a
reset
command
would
have
fixed
that
I'm
not
sure,
but
something
something's,
always
wonky
with
the
shell.
There.
A
A
So
someone
some
that
I
have
found
to
be
kind
of
interesting
and
have
used
there's
the
records
plugin
so
instead
of
having
normally,
you
are
actually
forced
to
put
all
of
your
dns
settings
for
a
domain
into
another
file
using
that
file
plugin.
So
this
actually
makes
it.
So
you
don't
have
to
do
that.
You
can
put
it
right
into
the
core
file,
so
it
makes
it
a
little
bit
easier.
You
know,
especially
for
instances
like
right
here
with
kubernetes.
A
A
One
file
something
I've
also
been
doing
coordinates,
supports
forwarding
requests
to
a
different
dns
server,
but
unbound
actually
takes
that
to
the
next
kind
of
the
next
level
of
it,
and
enabling
this
enables
coordinates
to
be
what
is
known
as
a
recursive
dns
server,
and
what
that
really
means
is
the
whole
purpose
of
the
server.
Is
that
any
any
kind
of
queries
it
doesn't
know
about
anything?
A
It's
not
authoritative
with
such
as
google.com
you're
not
going
to
be
the
authority
of
google.com,
it
will
reach
out
to
whatever
servers
you
define,
it'll,
grab
those
records
and
cache
those,
and
it
will
kind
of
pretend
to
be
kind
of
an
authoritative
figure
for
some
of
those
dns
zones,
even
though
it's
not
really
so.
This
is.
A
This
is
just
great
just
give
you
another
layer
of
cash,
but
one
of
the
problems
I
ran
into
was
actually
the
fact
that
a
it's
an
external
plug-in,
meaning
you,
you
can't
just
use
it
right
off
the
bat
b
you
had
to
compile
it
and
specifically
for
unbound,
you
have
to
compile
it
against
unbound
libraries
on
your
system.
So
you
know
if
you're,
compiling
this
for
debian
or
red
hat
ubuntu,
it
doesn't
matter
you're,
actually
going
to
be
tied
to
the
the
actual
operating
system
you've
compiled
on.
A
So
this
has
a
little
bit
of
a
disadvantage,
because
you
can't
use
one
of
those
generic
container
images
like
alpine,
which
alpine's
amazing.
If
you
haven't
used
it
they've
managed
to
get
the
base
operating
system
down
to
megabytes
or
mebibytes.
I
should
say-
and
you
know
if,
if
you
have
like
a
like
a
small
application,
that's
like
a
binary
from
go
and
you
stick
it
in
an
alpine
container,
like
your
application.
Could
in
theory,
like
total
container
image
size
could
be
like
5
10
megabytes,
which
is
just
like
ridiculous
to
think
about.
A
A
A
However,
me
being
frugal
me,
I
need
some
place
that
I
can
guarantee
that
it
will
always
be
available,
no
matter
what
so,
I've
also
put
it
up
on
red
hat
equata
io,
which
also
has
kind
of
cool
benefit
of
doing
container
scanning
for
you.
So
I
know
docker
hub
recently
added
that
ability
too,
but
by
default
it's
enabled
by
quay
and
completely
free
as
long
as
your
projects
are
free
and
open
source.
A
Yeah
one
thing
I
was
wondering
too:
you
know
what
what
are.
Are
you
folks
using
for
container
registries,
because
it
kind
of
hit
me
the
other
day
I
was
like
you
know.
Besides,
like
a
year
ago,
I
didn't
know
red
hat
quay.
Existed,
I
mean
red
hat
quay
didn't
exist.
It
was
just
quay
before
red
hat
bought
it,
but
like
what?
What
else
is
there
out
there?
Because
I
feel,
like
the
the
options
for
container
registries,
are
very
limited.
A
A
A
A
So
let
me
let
me
show
you
how
to
how
to
actually
compile
core
dns
with
whatever
plugins
it's
it's
actually
fairly
straightforward.
You
probably
don't
actually
need
to
pay
attention
to
that
file
right
there.
Let's,
let's
go
into
dockerfire
build
yeah.
This
is
this
is
literally
it
go
generate,
which
I
believe
is
generating
kind
of
the
list
of
dependencies,
and
then
you
build
it
and
once
you've
done
that
you
you
did
it.
You
have
core
dns,
it's
a
single
binary.
A
A
A
Oh
and
let
me
let
me
just
say
too,
I
absolutely
love,
z,
z,
sh,
okay,
it's
a
kind
of
a
more
a
newer,
more
kind
of
advanced
shell
and
using
I'm
using
oh,
my
zsh,
which
is
kind
of
just
a
layer
of
just
like
different
plug-ins
to
make
it
a
nicer,
smoother
experience.
So
you
can
just
do
something
like
this.
I
know
the
directory
I
want
to
go
into
is
slash,
get
slash,
t
slash,
kubernetes
templates.
So
if
I
just
tab
this
out,
oh
magic,
it's
absolute
magic.
A
A
So
in
here
I
must
say
not
all
of
these.
These
are
my.
This
is
a
private
repository.
These
are
some
sample
applications
I
have
running.
I
gotta
say
they.
Don't
all
look
the
prettiest.
I
need
to
go,
go
through
and
groom
these.
I
made
them
a
long
time
ago
and
I
deployed
them,
and
I
never
looked
at
them
again,
but
corey
ness
though
this
is
something
I've
I've
actually
just
put
together,
so
them.
A
A
So
the
interesting
things
from
this
are,
I
start
off
with
an
initialization
container,
so
just
a
simple
busy
box
image,
and
then
I
have
in
my
report.
In
my
repository
I
have
my
core
file
and
I
have
a
zone
file
as
well
that
I
serve
for
my
own
home
network.
So
I
download
these
into
what
is
a
volume
known
as
empty
dir?
It's
basically
just
a
temporary
file
temporary
directory,
which
you
can
then
share
after
the
initialization.
All
I
have
to
do
is
download
two
files.
A
A
A
So
I'm
gonna
switch
over
so
for
for
any
of
those
who
are
curious,
this
is
actually,
I
think,
a
pretty
interesting
way
to
do
it.
I
actually
have
two
kubernetes
clusters
running
at
home,
so
I
have
an
infrastructure
cluster,
I'm
just
using
k3s.
I
don't.
I
don't
need
any
fanciness.
I
don't
need
to
deal
with
the
cube
admin.
I
just
need
a
basic
curvatures
cluster
that
just
works
so
on
here
I
have
some
important
applications.
A
I
have
coordinates
for
my
home
network
and
also
serves
records
that
I
need
for
my
production,
kubernetes
cluster,
I'm
also
using
nfs.
So
this
is
my
main
file
server,
I'm
actually
using
like
zfs
in
the
back
end,
which
is
really
cool.
I
can't
believe
how
much
I
love
zfs.
I
used
to
be
a
btrfs
level
lover
and
I
still
am,
but
zfs
just
has
a
few
little
extra
features
on
top
that
are
just
ooey,
gooey
goodness.
A
So
nfs
I
actually
use
as
the
storage
backend
for
my
production
cluster,
so
any
kind
of
stateful
application,
some
not
not
literally
stateful
api
and
kubernetes,
but
that's
something
that
needs
storage
in
the
back
end.
I
use
an
nfs
mount
from
this.
This
cluster
right
here,
which
next
up
on
my
to-do
list,
I
want
to
make
it
a
dynamic
storage
class.
A
I
want
to
just
say,
like
hey
just
automatically,
you
know
give
me
a
pod
and
just
just
just
grab
some
storage
off
of
here,
auto
magically,
but
for
the
time
being,
I'm
just
manually
defining
persistent
volumes
and
perfect
persistent
volume.
Claims
that
say
hey
these
are
explicitly
to
be
used
as
a
storage
back-end
for
this
application.
A
A
A
Let's
check
out
cordiness
yeah
there
we
go
so
as
we
can
see
we
have.
This
is
my
this
is
my
mount.
I
wonder
if
I
don't
even
know
if
we
can
see
the
mount
itself
in
the
container.
Let's,
let's
just
see,
whoa
that's
cool,
okay
yeah,
so
you
can
see.
Kubernetes
has
this
like
fake
device
for
us
and
it's
it's
mounted
the
volume
volumes
mounted
at
etcetera,
accordion
s
and
it's
mounted
as
read,
write.
If
there's
any
errors
remain,
remount
says
read
only.
A
A
A
So,
looking
into
this
core
file,
you
can
see
it's
it's
pretty
basic,
so
I
have
my
unbound
plugin
that
I
I've
compiled
and
I'm
amusing.
I'm
such
a
I'm
caching,
the
records,
and
this
this
is
awesome.
So
you
can
core
dns
for
forwarding
requests
supports
many
different
backends
and
so,
for
example,
it
supports
dns
over
tls.
A
So
secure
dns
queries.
That's
you
know.
No
one's
going
to
be
able
to
snoop
on
which
is
is
pretty
cool,
so
I
have
that
set
up
in
there.
I
want
to
see
the
logs
see
the
errors
and,
of
course,
for
my
dns
zone,
I'm
I'm
saying
hey.
I
want
to
I'm
going
to
load
balance
this
because
I'm
actually,
I'm
literally
using
this
dns
server
as
a
load
balancer
for
at
least
one
of
my
domains.
A
A
Yeah-
and
so
just
just
you
know,
you
can
just
literally
copy
and
paste
some
of
some
of
these
things
just
straight
out
of
the
the
core
dns
manual
here.
The
the
biggest
takeaways
here
is
just
hey.
I
just
I
have
a
state
of
authority
record
just
says:
I'm
going
to
be
managing
this,
this
domain
home.short.
A
My
last
name
is
short.
This
is
my
home
network,
so
I'm
trying
to
give
it
some
logical
hierarchy
layers.
Here
I
got
the
k3s
cluster
for
my
infrastructure
and
then
these
are
my
main
production
nodes
which
you
can
see
here.
I
just
have
the
same
a
record
on
the
same
subdomain
for
these
a
records
just
defined
three
times
and
that
right
there,
along
with
using
the
load
balance
plugin,
you
got
a
load,
balancer,
dns
load,
balancer
now
granted,
there's
a
reason.
A
Dns
isn't
always
used
for
load
balancer
and
why
you
don't
hear
about
it
too
much
is
because,
if
you
run
into
a
situation
where
one
of
those
nodes
goes
down,
it
doesn't
matter,
dns
is
still
going
to
say:
hey,
you
know
you
can
go
to
that
node
and
then
it
gives
it
the
ip.
But
the
node
may
be
down,
and
then
the
user
gets
some
like
404
or
just
like
hey.
You
know
this
connection
refused
and
they
just
wouldn't
know
why?
A
A
A
A
A
Let
me
just
make
sure
I'm
definitely
on
the
right
cluster
here.
Okay,
so
we'll
make
note
of
of
this
pod
name.
We
want
to
make
sure
that
this
gets
updated
to
the
newer
version.
Once
we
apply
the
configuration
file
so
because
the
manifest
actually
has
the
same
exact
name,
because
we
literally
we
just
we
saved
what
was
in
kubernetes,
we
should
be
able
to
just
apply
this
right
now
with
the
latest.
A
A
A
I
would
never
recommend
straight
up
editing
an
actual
manifest
through
the
command
line,
because
you
will
run
into
configuration
drift.
I've
run
into
problems
into
this,
and
you
will
inevitably
too
so
just
be
very
mindful
of
that
so
yeah,
instead
of
using
like
a
yaml
file
to
define
your
stuff
once
it's
actually
in
excess
d
or
whatever
your
database
back
in
for
kubernetes,
is
your
source
of
truth
with
all
of
those
manifests
saved.
A
You
can
just
straight
up
edit
this
and
then
you
can
just
using
some
normal
linux,
shell
stuff,
it
actually
pulls
your
editor.
So
for
me
vim,
that's
my
go-to.
I
believe
that's
the
default
too,
but
you
know,
maybe
you
feel
more
comfortable
with
nano
or
emax,
perhaps
so
just
set
that
so
yeah,
let's,
let's
edit
the
config
map
object
named,
coordinates.
A
A
A
A
So
when
I
delete
this,
it's
the
cubelet
service
on
the
actual
worker
node
is
gonna,
or
in
this
case
it's
actually
control,
plane.
Node,
it's
gonna,
say
hey.
This
pod
is
missing.
Oh,
let's
recreate
it
because
there
needs
to
always
be
at
least
one
running.
There
needs
to
be
exactly
one
running
at
any
given
time.
A
Yeah
and
check
that
out
got
a
brand
new
pod,
different,
unique
identifier.
Here
it's
been
up
for
about
probably
30
seconds
now
and
magic.
It
should
yeah.
Now
now
it's
been
recreated,
it
should
have
the
latest
config
map
applied
to
it,
because
the
old
config
map
doesn't
even
exist.
The
whole
configuration
file
doesn't
exist.
All
it
knows
about
is
our
new,
updated
one
that
has
forwarding,
with
tls
secured
to
cloudflare
cloudfare
servers,
one
one
one
one
and
one
zero
zero.
A
A
A
A
A
A
You
know
there
isn't
a
tgik,
I
don't
know
why
there
should
be
so.
The
fact
is
because
we
are
using
an
authoritative
server
right
here.
We
can
define
it,
we
can
create
whatever
we
want
and
it'll
just
work
as
long
as
you
know,
whatever
computer
system
is
using
our
resolver,
we
can
actually
resolve
directly
to
that
ip.
A
A
A
A
A
A
A
A
A
You
can
revert
back
changes
easily,
but
the
problem
was:
is
that
ingress
doesn't
work
for
services
such
as
dns,
so
I
actually
had
to
use
an
option
and
let's
see,
let's,
let's
hop
into
it,
look
at
the
spec
file
an
option
called
host
network.
You
set
that
to
true.
So
when
you
set
host
network
to
true
in
your
manifest
cni
takes
over
zero
management.
Does
not
do
any
fancy
networking
overlay
for
your
containers,
it's
just
whatever
the
port
is
on
your
containers.
A
They
will
be
exposed
on
that
exact
worker
node
that
it
is
on
it's
very,
very
different
from
node
port
by
the
way
so
node
port,
like
I
said
earlier,
opens
up
a
port
on
every
worker
node,
regardless
it
is
managed
by
cni
and
node
port
uses.
It
assigns
a
random
port
somewhere
in
the
range
between
like
three
thousand
and
like
almost
thirty,
three
thousand
three
thousand
three
hundred
something
there's
a
couple
hundred
ports
in.
A
A
Yeah,
so
this
is
my
work
around,
because
this
is
not
an
http
service.
Ingress
doesn't
know
how
to
handle
it.
So
I
say:
hey
you
know
what
just
expose
it
don't
even
use
cni.
I
know
what
the
ip
address
is
of
my
node
I'll
use
that
as
my
resolver,
that's
totally
fine,
there's
no
other
services
listing
on
that
port,
and
I
I
couldn't
actually
make
this
a
deployment
because
the
deployment
it
tries
to
set
up
a
secondary
pod.
A
You
know
whether
you,
the
main
problem
actually
is
when
I
rebooted
my
server
is:
there's
still
the
existing
pod
and
the
deployment
would
try
to
get
rid
of
that
and
then
spin
up
a
new
pod,
but
the
first
pod,
because
it's
using
host
network
sets
it
true.
It's
listening
on
port
53.,
so
this
new
pod
is
trying
to
spin
up
and
it's
like
hey.
I
can't
listen
on
port
53
because
something
else
is
listening
on
it.
A
A
And
so
in
my
initialization
container,
I
just
pull
down
whatever
the
latest
configuration
files
are
for
the
core
file
straight
from
the
master
branch.
So
as
soon
as
we
delete
and
recreate
this
pawn,
we
will
see
that
it'll
just
it'll
have
our
latest
changes,
because
we
just
committed
those
not
too
long
ago.
A
A
A
Now
how
how
you
name
your
kubernetes
objects
is
totally
up
to
you,
something
I
like
to
do.
I
I
really
like
to
prefix
it
with
whatever
the
object
type.
Is
I
just
for
organizational
purposes?
I
just
I
don't
know
part
of
my
ocd
helps
me
better
understand.
Like
okay,
like
you
know,
I
can
ls
a
directory
and
I
can
be
like
oh
right
like
this.
Is
you
know
obviously
supposed
to
be?
You
know
that
that's
a
deployment,
that's
the
deploys,
deploys
the
shorthand
for
that.
A
A
A
A
A
A
You
know
what
I
don't
think
I
added
a
w
get.
I
didn't
I'm
not
downloading
the
the
file,
the
the
actual
dns
zone
file.
I
bet
you
when
we
look
into
the
core
file
right
here.
It's
gonna
say
yep
yep,
so
our
changes
are
being
updated.
I
just
forgot
to
update
the
initialization
container
to
download
that
in
easy
fix.
A
Yeah
marcos
asked:
why
am
I
not
using
node
port
instead
of
host
network
yeah,
really
great
question?
The
problem
I
have
is
that
I
specifically
need
to
be
listening
on
port
53..
If
I
do
a
node
port
it'll
force
me
to
use
a
port
over
3000.
I
I
need
there's
so
many
assumptions
and
like
resolvers
and
different
services
that
your
dns
server
is
always
going
to
be
listing
on
port
53
and
it's
very
difficult
to
force
them
to
use
anything
else.
A
A
A
A
A
A
So
it's
going
to
continue
just
to
serve
those,
but
because
we
didn't
specify
a
load
balancing
one
so
for,
for
example,
if
this
was
111
at
the
end,
without
specifying
the
load
balancing
plug-in
for
this
domain,
it
will
always
return,
always
return
110
as
the
actual
ip
address
which,
if
you're,
defining
multiple
of
the
same
a
records
or
quadruple
a
records
for
ipv6,
probably
not
really
what
you
want
so
cool.
I
am
I'm
now
the
proud
owner
technically
of
courdiness.tgik,
and
I
might
keep
it
there
for
for
months
to
come.
A
A
A
Yeah
so
gosh
there's
a
lot
of
good
guides
out
there
on
making
your
own
plugins.
I
doubt
anyone
here
would
really
care
about
that
too
much,
but
a
lot
of
great
documentation
out
there.
It's
the
this.
This
isn't
specifically
it
but
yeah
if
you're
interested
it,
at
least
from
what
I've
seen
very
easy
to
get
into
lots
of
great
examples
out
there
as
well
for
making
your
own
plugins
they've
they've
really
made
it
from
the
ground
up
to
you,
know
plug-ins
or
a
first-class
citizen.
A
So
this
is
the
official
kind
of
kubernetes
documentation
as
well
about
just
deep
diving
into.
You
know
how
you
can
configure
this.
So,
as
we
saw
earlier,
you
know
the
different,
the
core
file,
which
you
can
you
can
update
yourself.
You
can
put
whatever
you
want
in
here.
You
know,
if
you
you
know
for,
for
my
instance,
for
for
my
situation.
I
use
my
own
core
dns
server
with
the
unbound
plugin,
but
for
personal
uses,
not
not
for
kubernetes
necessarily
itself,
not
for
internal
pod
networking,
but
more
as
an
end
user.
A
On
my
home
network,
I
just
like
to
generally
have
that.
So
if
you
want
to
do
something
like
that,
they'll
build
your
own
core
dns
container,
you
can
totally
do
that
and
just
go
into
the
the
deployment.
If
we
edits
deployment
coordinates,
you
just
go
down
image
image,
so
this
is
k3s
made
by
rancher.
So
of
of
course,
they
have
their
own
kind
of
custom
image
here,
but
yeah
I
mean
you
can
just
need
to
plug
and
play
this
out.
A
However,
you
want
switch
configuration
file,
update,
config
maps,
add
additional
files
that
way
too
yeah,
which
also
note
me
doing.
Wget
was
just
like
a
quick
and
and
cheesy
hacky
way
of
of
getting
files
in.
You
really
want
to
use
a
config
map
when
you're
like
doing
this
in
kind
of
more
of
a
production.
A
Setting
yeah,
so
I
mean
this,
is
I
mean
this
goes
well
beyond
anything.
You'd
have
to
worry
about
on
like
the
the
cka,
but
I
just
I
thought
it
was
real
interesting
to
kind
of
look
into.
You
know
how
these
different
pieces
fit
together,
like
how
you
can
use
chordiness
yourself,
because
understanding
how
to
use
it
out
of
kubernetes
as
well.
Like
that
that's
kind
of
step,
one
is
just
being
being
able
to
figure
out
like
okay.
You
know
this
configuration.
A
Yeah
so
wow,
that's
that's
kind
of
kind
of
a
lot.
I
hope
everyone
learned
a
little
bit
about
core
dns
today.
You
know
I
I
feel
like
I
know
just
enough
to
be
dangerous,
so
you
know
sky's
the
limit
for
a
lot
of
this
stuff,
and
you
know
I've.
I've
had
just
a
blast.
It's
really
just
been
the
past
week
that
I've
been
digging.
You
know
deep
into
this.
I
I
knew
cordiness
existed.
I
knew
it
was
a
thing,
but
I
was
like
you
know
I
just
I
just
wish.
A
Yeah
closing
off,
you
know
I'm
kind
of
curious.
I
have
a
few
other
ideas
for
different
tgik
sessions
in
the
future.
One
is
less
of
an
idea
and
more
about
presentation
style.
I've
actually
given
a
few
presentations
in
virtual
reality
before
so.
I
have
this.
This
awesome
unlimited
whiteboard
app
that
allows
me
to
just
like
pick
up
markers
and
just
draw
concepts
up
and
just
walk
through
some
things.
A
A
I've
also
for
a
while
I've
been
wanting
to
do
a
serverless,
minecraft
server,
so
I've
talked
to
our
k-native
engineers
at
vmware
about
the
the
concept
I've
pitched
it
to
them.
I'm
like
hey
is
this:
is
this
possible,
like
from
what
I've
read
and
seen
like
theoretically,
they're
like
yeah
totally
so
k
native
has
been
something
I've
been
super
interested
in
for
a
long
time.
I
picked
up
a
book
on
it
recently.