►
Description
Don’t miss out! Join us at our upcoming events: EnvoyCon Virtual on October 15 and KubeCon + CloudNativeCon North America 2020 Virtual from November 17-20. Learn more at https://kubecon.io. The conferences feature presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.
Minikube - Rohit Anand, NEC Corporation & Medya Ghazizadeh, Google
Join minikube maintainers for a deep dive into how minikube runs Kubernetes on your local machine. We'll cover how to get started with minikube and what customizations are available so you can have the best possible local development experience. We'll also go over how you can contribute to the project, new features we've recently added, and what you can expect from minikube in 2020. This session is intended for both new and seasoned minikube users.
https://sched.co/Zexq
A
B
A
A
So
whenever
I
go
to
conferences
and
then
talk
about
any
cube
with
tech
in
the
tech
events,
the
first
thing
they
bring
up
is
the
mini
cube
emojis
and
how
many
you
expresses
itself
in
in
emojis
and
who
is
behind
these
emojis.
I
would
like
to
take
one
slide
to
acknowledge
the
hard
work
and
the
passion
that
I
have
witnessed
from
from
these
names
myself.
A
People
like
thomas
stromberg
myself
and
under
spirikland,
shari
valgamal,
priyawadawa,
jose
donizetji,
previous
maintainers
of
mini
cube
like
don
lauren,
matriakart
iron
prindle
and
the
newer
contributors
are
mini
cubes
such
as
gold,
vargo,
urban
bias
and
more
than
500
minutes.
You
contribute.
Thank
you
very
much.
A
B
A
What
minicube
provides
a
very
fast
way
to
build
images
with
docker
and
and
five
other
ways
and
also
access
your
app
mount
your
files
into
the
cluster
feature
gates
and
more
than
25,
but
what's
new
mini
q?
You
gotta
listen
to
this
joke
to
to
like
get
the
rhyme
of
this
title
by
the
way,
one
secret
that
I'm
going
to
share
with
you
with
that
we
mini
q
maintainers,
we
never
never
type
mini
cube
with
capital
m,
that's
something
that
you
should
never
see:
mini
cubes
mini
mini
cube,
it's
not
meaning
so
capital.
A
It
has
two
new
drivers,
we
have
multi-node,
we
have
a
new
command
called
pause
and
a
pause
very
exciting,
and
we
auto
select
flags
for
you
and
we
have
five
tools
that
graduated
out
of
mini
cube.
I'm
gonna
give
you
a
deeper
look
at
all
of
these
things
that
I
just
announced
before
that.
Let's
take
a
mental
picture
of
this
slide
and
not
forget
it.
Let's
take
a
mental
picture.
A
Minitube
supports
all
of
these
users.
Also,
those
green
ones
are
the
ones
that
we
have
integration
tests
for
it
and
we
cannot
break
it
on
every
pr
mini
cube.
You
have
to
support
them.
You
have
to
support
a
guy
with
linux
and
amd64
kvm
driver
and
any
of
those
container
run
times
and
any
of
those
cni's
and
any
of
those
combinations
that
is
possible.
We
support
it.
For
example,
you
can
have
linux
and
docker
and
arm
processor
and
cryo
and
weave.
B
A
Cni-
and
this
is
very
important-
minicube
has
more
than
3.3
million
downloads
on
github
we
have
tens
of
thousands
of
users
and
they're
very
diverse
in
technology.
This
was
my
community
very
unique.
We
support
so
many
container
runtimes
cni's
and
drivers
and
we're
not
just
one
technology.
We
are
very
diverse.
A
B
A
B
A
A
A
So
in
k3d
you
can
see
roughly
is
around
25,
where
mini
q
is
roughly
around
45
and
four
kind
is
roughly
around
65
and
k3d
by
the
way
does
not
use
kubernetes
source
code
completely.
They
fork
some
part
of
it.
They
replace
the
lcd.
That's
why
they
are
faster,
so
minicube
seems
like
in
this
script
that
was
run
113
times.
We,
you
can
check
the
time
to
cates
on
github
and
thomas
ripple
to
reproduce
this
result.
A
How
did
we
get
from
2
minutes
and
35
seconds
to
21
seconds
step?
One
was
collecting
metrics.
We
wanted
to
know
where
we
can
invest
to
make
pr.
There
are
so
many
ideas
we
had
how
to
make
miniature
faster.
So
many-
and
we
said
how
about
you-
write
a
pr
bot
github,
but
that
will
comment
and
says
this
pr
makes
mini
cube
this
much
faster
or
slower
and
priya
did
this,
and
it
helped
us
massively
to
avoid
accepting
prs.
A
That
was
adding
tons
of
complexity
to
minicube
and
was
not
doing
actually
any
significant,
meaningful
difference.
Another
tool
was
slow
jam
which
thomas
wrote,
which
is
a
ui
visualizing
the
stack
traces.
Let's
show
that
exactly
which
function
exactly
which
function
was
the
one
that
we
should
attack
to
break
it
down
or
to
make
it
faster
or
make
it
run
in
parallel,
so
big
targets
was
creating
the
vm
and
also
loading
images
into
the
container
on
time.
These
are
the
two
big
ones.
The
first
one
when
I
first
joined,
mini
cube
a
team.
A
B
A
A
Joined
I
was
like
why
nobody
did
this,
why
nobody
added
a
docker
driver
for
mini
cube.
What
would
happen
if
we
just
create
a
container
just
the
same
way
as
we
create
a
vm
and
then
bootstrap
the
kubernetes
inside
of
it,
and
and
just
do
exactly
what
we
do
for
moviems?
Why
we
are
not
doing
that
and
the
answer
was
when
you
do
that
it
crashes
and
I
and
there
was
a
project
called
kind,
and
I
looked
at
it.
A
It's
like
what
kind
is
doing
that
we
are
not
doing
why
they
are
not
crashing
and
we
are
crashing
when
we
do
that,
and
the
answer
was
in
this
red
hat
blog.
It's
like
the
real
problem
was
system
d
cannot
run
in
a
container.
If
you
don't
do
some
tricks,
so
our
problem
become
much
much
smaller.
Now
now
our
problem
was
like
how
to
make
systemd
work
inside
container
inside
a
container,
and
once
we
solve
this
problem
by
basically
using
the
same
entry
point
as
a
kind.
The
entry
point
that
fixes
this
problem.
A
We
basically
added
docker
driver
by
by
basically
creating
the
container
the
same
way.
We
created
the
vm
and
do
the
same
thing
we
do
for
the
vm
and
but
there
was
still
other
things
that
we
had
to
take
care
of,
like
refactoring,
a
lot
of
the
mini
cube
code
to
be
more
abstract
and
more
reusable,
and
the
answer
was
like.
A
A
A
A
A
B
A
The
next
target
was
preloading,
the
images.
So
can
we
preload
images
for
vms
as
well,
for
example,
the
project
like
kind?
They
would
build
an
image
that
has
all
the
things
they
want
to
pre-load
it
per
kubernetes
version.
So
if
you
want
different
kubernetes
version,
you
want
a
different,
build,
a
different
image,
and
that
has
all
the
things
in
there
that
would
not
be
applicable
for
mini
q,
for
pulse
vm
and
docker
driver.
We
just
follow
a
different
workflow.
A
A
So
how
do
we
do
that?
We
came
up
with
this
solution
that
we
have
the
base
image
or
the
iso
I
so
for
the
vm
based
image
for
the
container
and
they
are
they
don't
have
the
kubernetes
image
inside
of
them.
We
pull
that
from
a
from
internet
from
a
tar
file
that
has
a
frozen
file
system
of
a
loaded
overlay
file
system
and
we
just
untie
it
into
the
vm.
A
That
makes
a
little
bit
of
a
problem
that
that
takes
time
to
untar
something
and
also
download
something
the
download
takes
only
one
time
this
next
time.
You
will
not
download
it
again,
but
the
untouring
could
could
be
adding
overhead.
So
we
came
up
with
a
solution
that
we
spin
up
a
side
container.
In
parallel,
we
do
the
untiring
there
make
a
volume
out
of
it
and
then
attach
it
to
the
container,
while
the
other
container
is
spinning
up.
A
So
we
save
time
by
parallelization
here
and
that's
how
we
actually
got
to
the
most
of
the
86
percent
goal
of
minikit,
there's
actually
tons
of
other
small
optimization.
Each
one
would
cut
two
three
four
five
seconds
per
minute
cube,
but
these
two
were
the
top
biggest
one
and
priya
implemented
the
preload
as
well.
A
A
A
Nothing
really.
It
means
means
like
only
docker
and
nothing
on
it.
That's
how
much
an
empty
and
nothing
on
its
docker
would
would
use
from
the
cpu
0.47
and
a
mini
q
on
hyperkit
uses,
3.89
and
kind
uses
5.5
by
the
way
the
scripts
to
generate
this
are
all
available.
The
links
will
be
at
the
end
of
the
slide.
A
A
A
Another
exciting
feature,
sorry
about
that
is
mini
q
paws.
So
we
did
a
lot
of
analyzation
using
slow
jam
using
perf
tools
that
we
developed
and
we
found
out
the
kubernetes
api
server
is
actually
the
one
which
is
the
source
of
the
problem
like
if
the
kubernetes
api
server
was
not
there.
Minikit
was
like
almost
nothing
in
terms
of
cpu
overhead.
If
you
look
at
that,
yeah
see
what
if
we
pause
the
api
server
and
we
we
did
that
and.
A
A
Another
exciting
thing
was
shaital
gomal.
He
implemented
the
most
reacted
feature
in
the
history
of
mini
cube,
also
a
2016
issue.
This
was
actually
a
sooner
issue
on
the
vm
free
issue.
Issue
number
94,
it's
incredible!
Looking
at
this,
because
now
we
have
tens
of
thousands
of
issues
I
mean
in
number,
so
we
have
multi-nodes
and
thanks
to
sheriff
gomer
for
very
implementing
this.
A
Another
exciting
thing
here
is:
we
have
auto
select
options.
Mini
cube,
has
so
much
options.
If
you
do
mini
cube,
start
dash
dash
help
you
see
so
much
and
we
used
to
make
people
pass
the
vm
driver
to
them
to
the
start.
Now
you
no
longer
have
to
pass
anything.
We
check
your
system
find
the
best
driver
based
on
your
system.
What
is
healthy,
what
is
the
best
health
score
for
each
driver
and
we
pick
the
best
driver
for
you
and
also
based
on
your
system.
You
pick
the
best
cni
the
best
memory.
A
Everything
is
auto
selected
for
you
in
a
smart
way,
based
on
your
needs,
not
a
generic
default
value
and
it's
another
exciting
part
of
miniqueue
another
new
thing
in
minicube
we
have
a
add-on
that
helps
you
develop
apps
on
mini
cube,
easier
when
you
want
to
when
you
want
your
apps
once
you
do,
credential
calls
to
gcp,
so
you
no
longer
have
to
mount
your
secret
to
every
pod.
If
you
enable
this
add-on,
it
will
do
it
for
you
automatically.
A
B
Now
you
need
to
create
linux
foundation.
Identity.
For
that
you
need
to
log
in
using
your
github
account
once
you're
logged
in
into
your
github
account,
your
github
username
will
be
automatically
pre-populated.
You
need
to
fill
out
your
first
name
last
name
and
submit
the
form
it
will
create
your
identity
on
the
linux
foundation,
identity
portal.
B
If
you
are
signing
the
cra
using
your
organization
name,
then,
managers
of
the
organization
name,
email
addresses
and
all
the
individual
contributors
name
and
email
addresses
are
required
to
sign
the
celi.
Once
this
form
is
submitted,
a
cla
will
be
sent
to
the
individual
email
and
it
must
be
signed
electronically.
B
B
B
B
B
B
B
Once
make
command
is
completed,
it
will
create
a
fresh
mini
cube
binary
in
the
r
directory.
So
if
you
have
cm
system
local,
then
many
cubes
start
from
the
out
directory
gonna
work
as
a
test.
If
you
have
different
system
local,
you
can
override
the
printed
version
using
lc
all
environment
variable.
So
I
am
using
here.
Japanese.
B
B
B
B
A
B
A
Remember
the
mental
picture,
I
guess,
as
you
guys
to
take,
which
which
was
this
one
remember
this
one.
We
support
all
of
this
combination
of
things.
We
support
any
possible
combination
of
this.
How
do
we
not
break
them
and
the
thing
is
we
do
break
them,
but
each
time
we
break
them,
we
add
an
integration
test
for
it.
So
it's
the
last
time
we
ever
broke
it.
So
I.
A
A
So
if
you
see
in
each
minicube
pr,
we
have
21
integration
tests
of
different
kinds,
for
example
docker
and
windows,
docker
and
mac,
os
hyper-v
and
windows,
hyperkit
and
macos,
and
each
of
those
integration
tests
is
about
200
tests.
Some
of
them
take
an
hour
or
sorry
40
minutes.
So
we
invested
in
testing,
and
this
is
one
tool
that
came
out
of
mini
cube
that
I
built
over
a
hackathon
and
when
we
were
looking
at
integration
test
results,
here's
how
it
looks
like
from
200
integration
tests.
Let's
say
something
failed!
A
A
So
and
you
can
see
this
is
and
then
you
can
search
with
it
and
open
a
new
window
copy
the
logs
folder
permanently.
So
this
is
done
by
gopro
and
gopro.
I
initially
named
it,
go
pretty
and
then
it
just
auto
completed
myself,
go
pretty
or
go
home.
It
was
a
hackathon
ideas.
I
just
came
up
with
a
quick
name
and
just
became
go
poke
and
we
use
gopro
in
our
integration
test
to
convert
a
text.
A
B
A
Recently
added
a
stress
test,
so
we
had
a
bug
in
mini
cube
that
we
noticed
if
we
had
many
docker
drivers.
At
the
same
time,
we
had
a
specific
issue
and
we
said
how
about
you
make
that
issue
never
happen
again.
We
write
a
test
for
it
and
forget
about
it.
So,
each
time
we
make
a
pr,
the
the
testing
infrastructure
helps
us
to
catch
them
and
we
develop
like,
I
would
say,
an
empire
of
testing.
Really.
A
We
have
so
many
vms
in
google
cloud
and
we
reach
out
to
sig
windows
and
thank
you
so
much
for
sig
windows
to
hooking
up
us
with
windows
vms
in
azure.
So
now
we
test
windows.
We
used
to
not
have
test
windows,
and
now
we
do
tests
in
parallel.
One
year
ago
we
were
not
doing
tests
in
parallel
now
the
test,
as
opposed
to
one
hour
and
a
half.
It
takes
30
minutes
or
40
minutes,
but
we
do
way
more
testing
we
more
testing
because
we
do
the
tests.
B
A
Work
development:
you
can
actually
have
mini
cube,
run
in
github
actions
by
adding
three
lines
just
say:
uses
media,
gh,
slash,
setup,
dash,
mini
cube,
and
then
you
can
test
your
apps
against
a
running
mini
cube
in
every
github
action
job
so
for
full
example.
Click
on
that,
but
before
I
get
distracted
about
this
slide,
we
also
have
customers
from
ide
teams
that
use
mini
cube,
underneath,
as
you
would
not
notice
that
they
are
using.
A
A
Of
running
mini
cuban
ci,
by
the
way
that's
how
this
set
up
mini
cubit
of
action
in
when
we
implemented
it,
we
use
those
for
you.
Minitube
speaks
your
language.
We
you
saw
that
rohit
showed
us
how
to
add
japanese
language.
This
is
meaningful.
Speaking
french
or
issue
templates.
We
have
issue
terms
in
multiple
languages.
We
accept
contribution
from
all
languages.
A
A
Anybody.
We
try
to
make
sure
everybody
can
access
us.
We
we
have
detections
how
to
help
them
to
run
mini
cubing
corporate
environments
with
proxy
environments
and
also
we
we
have
a
tool
called
triage
party
that
also
graduated
from
mini
cube
project,
and
I
encourage
you
to
come
with
us.
We
have
a
trash
party
every
wednesday.
You
can
come
play
with
us
and
I
will
let
you
read
more
about
it
because
we're
running
out
of
time.
This
is
a
tool
that
we
wrote
to
help
us
triage
the
makeup
issues.
A
A
Here's
the
link
to
all
of
the
tools
that
I
mentioned
in
this
slide
and
also
the
link
to
the
slack
take
a
look
at
them.
By
the
way
we
have
surveys
on
every
release.
Notes
if
you
click
on
it,
we
do
listen
to
the
surveys.
This
is
a
survey
from
2020
to
one
the
top
two
categories
that
people
asked
us
to
improve
on
was
multi-node
and
vm
free
and
then
the
next
choir
when
we
delivered
those
things
and
fixed
all
the
bugs
for
it,
we
saw
that
there
was
none
of
them
anymore.