►
From YouTube: Migrating Configuration to Kubernetes with Container-Transform by Micah Hausler, Skuid
Description
Migrating Configuration to Kubernetes with Container-Transform - Micah Hausler, Skuid
"Kubernetes has accelerated application development time for many organizations but one of the most tedious aspects of moving from application prototypes to running pods on Kubernetes is the repetitive task writing pod configuration files. A common workflow for many teams is to write development configurations in docker-compose before running a pod in Kubernetes. After using various container clustering systems, Micah recognized the need to be able to quickly interchange between formats. In late 2014 Micah open-sourced container-transform for interchanging docker-compose and Amazon's EC2 Container Service tasks, and has since added several other application formats including Marathon, Chronos, and more recently Kubernetes! In this talk Micah will demonstrate how developers can convert an app from docker-compose to Kubernetes, or even migrate from another clustering systems to Kubernetes. https://github.com/micahhausler/container-transform
About Micah Hausler
Micah Hausler is a SRE at Skuid and enjoys building tools in addition to operating sites. Originally from Michigan, Micah has lived the last 9 years in Chattanooga, TN and loves kayaking, climbing, and biking in the area.
Join us for KubeCon + CloudNativeCon in Barcelona May 20 - 23, Shanghai June 24 - 26, and San Diego November 18 - 21! Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.
A
Kind
of
give
it
a
little
bit
of
background
first
for
me.
Personally,
it's
been
a
long
road
to
get
to
q.
Brunette
ease,
I've
tried
a
lot
tried
a
lot
of
different
systems.
It's
probably
a
lot
of
people
have
and
I
kind
of
wanted
to
give
a
little
bit
of
history
in
context
for
kind
of
how
everything
is
developed,
so
try
to
remember
way
back
into
2013
in
March
at
a
five-minute
lightning
talk
by
a
CTO
of
a
small
past
provider
called
doc
cloud.
A
A
A
The
next
thing
that
I
kinda
want
to
talk
about
is
November
2014.
What
happened
then
well
I
Amazon
get
some
skin
in
the
game.
They
launched.
Amazon
ecs
their
ec2
container
service,
which
again
was
a
really
great
even
at
the
time,
is
a
really
great
step
up
where
dr.
compose
was
great
for
prototyping
and
kuber
Nettie's
was
pretty
nascent.
A
A
It
was
a
simple,
simple
start
so
again
for
context
november
2014,
that's
like
a
year
and
a
half
after
docker,
so
before
we
get
too
excited
I
kind
of
want
to
back
things
up
to
two
thousand
eleven
five
years
old
right,
really
old
meses
comes
on
the
scene
and
has
initially
just
a
few
schedulers
Aurora
being
the
chief
scheduler
and
originally
since
then,
we've
seen
things
like
marathon
and
Chronos
by
mesosphere
and
one
of
the
great
things
about
missus
if
you've
ever
used.
It
is
that
you
can
even
write
your
own
scheduler.
A
So
if
you're
into
really
need
to
be
into
resource
management
for
a
long
time,
mesos
has
been
kind
of
a
key
player
in
in
in
that
space,
so
fast
forwarding
again
September
2015
anyone
familiar
with
hashing
Corp
there,
the
company
behind
vagrant
console
vault.
They
release
Nomad
as
kind
of
a
again
another
scheduler
cluster
management
system.
A
So
again
that's
a
year
ago.
That's
that's
really
new.
So,
along
with
all
these,
we
start
to
kind
of
see
some
diverging
standards,
the
everyone
kind
of
has
their
own.
They
call
things
sort
of
the
same
parameters
slightly
different
things,
whether
it's
mem
or
memory
or
capital
m
and
mem
they're,
all
kind
of
have
the
same
functionality
or
similar
functionality
may
be
slightly
different,
but
they
they're
all
generally
the
same,
so
the
next
thing
I
kind
of
want
to
dig
into
is
so
we've
said
we
have
there's
all
these
different,
these
for
schedulers.
A
You
can
run
things
on
cluster
management
systems,
I
kind
of
want
to
look
specifically
at
the
configuration
formats,
because
I
think
those
are
interesting
case.
One
being
done
so
your
place,
doctors
composed,
was
believed
way
to
get
some
of
it,
and
it
was
just
kind
of
a
demonstration
of
example
of
file
this
one
we
have
an
engine
X
container,
that's
exposing
some
ports
containers
playing
a
web
app,
it
has
a
certain
environment
variables
and
then
we
have
a
database
container.
A
So
this
is
really
integrate,
like
this
I
really
do
like
it's,
my
go-to,
for
what
I
want
to
prototype
something
start
getting
something
to
work.
I
have
numerous
may
be
at
least
20
in
a
project
that
just
like
okay
I
want
to
try
this
on
the
try.
Great
log
I
want
to
try
console,
try
bolt
great
doctrine
pose
as
a
configuration
file
to
it
and
you've
got
something
really
awesome
and
it
just
works.
A
A
Essentially,
dr.
commodes,
you
don't
have
a
top-level
name
for
everything.
It's
just
you're,
my
container,
the
top
of
only
having
it's
assumed,
based
on
your
directory
on
your
on
your
laptop
or
where
and
CSU
family
name
was
volumes.
Dr.
proposed
Alex
you
that
and
then
you
just
have
essentially
the
same
parameters
so
yeah,
that's
cool
like
there's
some
that
aren't
there.
Let
you
go
see,
links
and.
A
But
this
is
kind
of
compact
Jason
of
a
marathon
that
Jason
there's
a
lot
more
extra
parameters
for
footage.
You
specify
the
type
of
container
doctor
because
marathon
has
other
container
coins
at
I.
Think
it's
now
having
rocket
CPUs
specifies
never
there
to
give
every
everything
an
ID.
You
have
to
give
some
some
memory
limits
to
everything
and
you
can
specify
the
network.
A
It
does
have
some
additional
parameters
if
you,
if
you
able
to
see
health
checks-
and
that
is
actually
just
actually
got
burned
into
dr.
act
like
a
new
health
care
for
ammeter
for
an
actual
retainer
but
hope
checks
have
been
around
there,
so
to
kind
of
get
the
idea,
it's
all
the
same
or
it's
very
similar,
and
then
the
last
case
will
look
out
early
to
mid
80s.
This
is
the
animal
for
this.
A
You
know
that
there's
yeah
no
config
and
Jason
the
same
thing,
but
they
named
personally
for
any
and
it
so.
This
shows
a
deployment.
It
could
there's
a
pod,
but
deployment
shows
just
a
little
better.
Some
more
configurability
labels
apply
to
every
reply
to
the
deployment
object,
not
just
in
tears
and
like
CSC,
going
to
the
bottom
and
there's
a
lot
you
incur
you
be
any
of
this
factors,
there's
a
lot
of
really
great,
not
objects
rating
strength,
but
so
it's
kind
of
stepping
backwards.
Again,
there's
always
these
similar
formats.
But
what
are
we?
A
So
what
is
an
engineer
to
do
automate
so
couple
years
ago,
when,
when
ecs
first
came
out,
I
was
said
frequently
use
dr.
Dafoe
has
really
enjoyed
it
and
wanting
to
run
these
same
things.
I
was
running
an
Amazon
really
quickly
and
so
the
sort
of
the
first
format
it
was
fine,
was
dr.
Campos
to
Amazon
nice
gifts,
and
this
is
a
kind
of
compact
version,
but,
as
you
can
see
like,
there
are
required.
Parameters
for
ucsf
compose,
doesn't
support.
A
A
The
most
recent
addition
to
all
this
being
driven
any
as
well
as
a
couple
other
formats.
One
is
system
mean
it's
ever
used
to
like
or
OS
call,
please
it's
very
similar
system,
the
units
kind
of
been
retired
at
this
point,
but
back
in
today
for
Humanity's
really
weak
was
with
it's
one
of
the
early
ways
to
try
to
try
to
run
containers
on
quarterbacks.
A
So
we
have
a
daughter,
composer
and
basically
what
what
is
what
I
showed
a
really
simple
setup
where
a
very
standard
web
app
you'd
have
your
engine
X
container,
you
have
a
web
app
behind
it
and
you
would
have
a
Postgres
database
behind
it.
So
you're
prototyping
your
web
app,
and
you
want
to
run
it
in
on
your
local
doctor,
but
all
of
a
sudden,
you
need
to
run
it
in.
A
A
We
can
just
run
container
transform
and
the
default
input
is
composed
just
because
that
is
the
most
common,
workflow
and
I
think
a
lot
of
developers.
Most
common
workflow
is
to
prototype
something
running
elsewhere,
and
so
you
can
really
quickly
see
the
output
formats
ecs
composed
systemd
marathon
Chronos,
which
is
another
scheduler
for
mesos,
that
does
cron
background
jobs
and
in
Coober
Nettie's,
so
we'll
just
try
to
will
do
ecs
first.
A
So
if
we
container
and
transform,
we
don't
specify
the
input
type
the
default
is
proposed
and
then
output
to
ecs
for
verbosity,
we'll
just
say
that
dark
composed,
oh
yeah
mo.
We
now
have
our
almost
ruin
bowl
ecs
file.
There's
again
a
few
things
missing
memory
memory
we
didn't
give
it
a
family
name,
so
we'll
have
to
manually
put
that
in
but
again
say
we
want
to
run
that
okay,
we're
in
that
in
the
Amazon.
The
next
thing
we
want
to
do
is
say:
okay,
let's
look
at
that.
A
A
Docker
compose
file,
and
it's
worth
noting
that
in
Marathon
you
actually
can't
run
groups
of
containers
like
you
do,
you
can't
occur
at
ease
with
pods.
You
actually
have
a
separate
service
per
container
I
believe
that's
something
that
they're
going
to
be
adding
in
the
near
future.
But
for
now
you
can
just
run
a
single.
A
single
service
has
a
single
container,
so
this
actually
makes
three
different
services,
and,
what's
that
you
with
the
individual
service,
you
can
adjust
the
number
of
containers,
yep
and
so
for
one
last
example:
with
this
stock
compose.
A
So
now
we
have
our
urban
indies
deployment
and
it's
worth
I'll
also
say
so
right
now
I
just
output,
Q,
brunette
ease,
but
container
transform
also
works
in
the
opposite
direction,
with
any
format,
almost
any
format.
If
I'll
run
that
help
again.
So
you
can
kind
of
see
the
input,
types
and
output
types,
there's
right
now,
I
think
yeah,
there's
just
one
output
type,
that's
not
supported
as
input
type
and
that's
a
system
d
so
say:
I
wanted
to
run
those
on
like
a
core
OS
box
and
I
dropped
it
in
on
the
cloud
config.
A
It
will
remove
any
existing
in
genetics,
pull
the
Krait,
correct
image
and
then
run
it
and
then
have
the
correct
stop
command
so
that
again,
that's
the
only
an
output
type
but
say
we
just
for
getting
a
little
meta
say
we
have
a
crew,
brunette
ease,
config
right
so
say
this
is
our
Cooper
neji's
fig
that
we
generated,
and
we
want
to
convert
that
to
some
like
you
have
an
existing
Cooper,
Nettie's
config.
You
want
to
debug.
A
And
that
is
because
the
cuban
at
ease
is
released
in
alpha
and
I
did
not
totally
bet
that
yet
okay,
so
in
this
case
we
output
is
ecs
without
put
a
doctor
classes,
ecs
rien
jested
as
ecs
and
put
it
to
say,
kuber
Nettie's.
So
let's
do
that
so
say
you
had
an
ecs
config
that
you
wanted
to
migrate
right.
So
here's
your
ecs
config,
let's
in
just
that
and
send
that
to
cube
Randy's.
A
Ok,
we
piped
that
in
and
now
we
have
a
deployment
for
Cooper
Nettie's,
it's
not
totally
good
to
go,
as
you
can
see
like
there's
some
labels
that
are
no.
We
don't
have
a
name
for
the
deployment,
that's
kind
of
something
that
I'm
not
sure
I'd
like
I,
don't
want
to
overload
and
add
extra
parameters
to
dr.
compose.
It's
not
not
compatible
with
our
compose.
A
I
might
add
some
extra
command-line
arguments
for
that,
so
you
can
say
a
name
for
your
your
output
of
whether
that's
your
family
for
ecs
or
your
your
deployment
name
for
for
Cooper
Nettie's.
But
that's
that
kind
of
gives
you
an
idea
of
the
capability
of
this.
So,
let's
jump
back
to
presentation
really
quick.
A
Alright,
so
I
talk
really
quick
about
the
internals
and
the
code,
the
the
project
written
in
Python
and
the
basic
idea
is
just
that:
there's
input
formats
and
there's
output
formats,
and
then
we
also
have
kind
of
an
intermediary
format,
and
this
either
intermediate
format
has
there's
a
lot
of
parameters
that
we
saw
that
are
or
not
a
lot,
but
a
number
of
parameters
that
are
unique
to
specific
formats,
so
say
for
marathon.
It's
fetch.
It
has
a
fetch
parameter.
A
So
when
you
boot
a
container
it
always
fetches
some
file
down
into
the
container
for
carew
Nettie's.
It
might
be
like
a
secret
that
you're
mounting
in
or
a
a
readiness
check
like
there
might
be
a
health
check
in
other
formats.
But
a
readiness
check
is
is
a
different
concept.
So
what
we
essentially
have
is
in
the
intermediary
concept
or
in
your
intermediary
format.
That's
not
ever
output.
A
It
just
has
a
representation
of
all
all
possible
parameters,
and
so,
if
there
are,
is
an
equivalent
on
the
output
format
based
on
that
input,
format
will
try
to
omit
it
if
it.
If
the
output
format
has
a
certain
requirement
that
the
input
format
doesn't
provide,
will
send
out
an
alert
and
say
hey,
you
need
an
image,
you
need
memory,
you
need
something
and
with
that,
as
are
there
any
questions,
you
guys
have
yeah.
A
A
With
with
this,
it's
really
just
doctrine
pose
or
container
transform
is
really
just
meant
to
be
a
developer
side
tool.
There
are
like
I,
said:
you'll,
see
no
no
values
and
things
so
I
I.
Try
not
try
to
do
the
least
amount
of
surprising
things
and
tell
you
if
something's
wrong,
because
it's
not
meant
to
okay
run
this
actually
in
production
like
like
there's
always
some
fine
tweaking
that
you
need
to
do,
and
so
I
believe
k
compose
will
try
to
like
run
things
in
communities
for
you.
That
might
be
wrong,
but
ok,
yeah.
A
It's
yeah,
it
isn't
its
own
thing.
It
has
similarities
to
dr.
Campos,
but
it
is
its
own
format
and
they
have
they
actually
open
source
their
agent
that
runs
on
so
with
with
ecs.
You
have
like
a
local
agent
that
runs
on
your
node
that
talks
to
your
local
docker
damon
and
runs
the
containers
and
checks
back
in
with
the
amazon
service
and
says
I
can
run
tasks
like
tell
me
what
to
run
it's.
A
It's
own
service,
yeah,
it's
not
even
Kuber
Nettie's,
it's
it's
own
thing,
their
own
implementation
and
so
similar
to
how
you
have
pods.
You
have
these
task
definitions
which
get
run,
and
the
most
of
these
different
formats
have
the
word
service
in
it
like
whether
that's
marathon
or
ecs
or
kuru
Nettie's,
and
so
that
actually
leads
to
a
lot
of
confusion.
My
kuber
Nettie's
is
the
last
of
these
things
that
I
came
to
the
most
recent,
and
so
there
was
a
big
mental
hurdle
for
me
to
be
like
certain
pledges
of
service.
A
A
Yeah,
the
the
original
output
format,
output
format
I
did
for
Cooper
Nettie's-
was
just
a
deployment,
but
one
of
the
things
I've
thought
about
and
haven't
implemented
yet,
but
would
just
be
to
say:
okay
have
a
different
output
types
for
Cooper
Nettie's
kubernetes
deployment.
You
bring
a
pod
urban
areas,
Damon
set
because
they're
there
are
different
you're
right.
A
There
are
different
types
of
urban
eighties
objects
that
you
would
like
to
run,
whether
it's
a
replica
set
or
whatever,
and
for
me
most
of
my
my
kuber
Nettie's
things
that
I
run
our
deployments
and
so
that
that
made
the
most
sense.
But
it's
all
like
the
spec
is
actually
almost
the
same.
A
crop
go
back
really
quick.
A
A
A
Yep
yeah
we'll
give
it
a
whirl
here's
the
links,
I'll
post,
this
set
of
the
slides
on
the
schedule,
but
I
have
documentation
for
it
as
well
as
the
project
itself,
and
if
you
find
bugs
file
an
issue
or
if
you
have
just
questions
about
adding
new
features,
please
please,
let
me
know
last
thing:
I
want
to
say
is
my
import
thanks.
My
employer
squid
for
sending
me
here
and
squid
is
hiring.