►
From YouTube: Kubernetes Sig Docs 20190205
Description
Meeting notes: https://docs.google.com/document/d/1zg6By77SGg90EVUrhDIhopjZlSDg2jCebU-Ks9cYx0w/edit#
The Kubernetes special interest group for documentation (SIG Docs) meets weekly to discuss improving Kubernetes documentation. This video is the meeting for 05 February 2019.
https://github.com/kubernetes/website
A
A
B
B
So
I
am
switching
my
roles
here.
At
Google
from
being
a
tech
writer
to
a
program,
manager
and
I
will
be
working
with
Sarah
Maddox
on
her
season
of
Docs
project,
which
is
to
learning
technical
writers
pair
them
with
open
source
projects
to
help
improve
the
documentation
and
also
give
the
technical
writers
some
experience
in
the
open
source
space.
So
that
means
I'm
going
to
need
to
step
down
as
co-chair
for
the
sig.
Docs
I'll
still
be
involved
here
and
there,
but
I
won't
have
as
much
bandwidth
so
I
think.
B
A
A
Some
of
this
has
to
do
with
what
I
guess,
one
might
call
succession
planning,
but
some
of
it
has
to
do
with
working
on
opening
up
leadership
opportunities
to
a
broader
group
of
people.
Also,
so
is
that
wanted
to
follow
up
on
this
topic,
but
he
couldn't
be
here
today,
so
I'm,
just
putting
this
out
as
another
ask
to
a
somewhat
expanded
group
from
what
we
had
last
week,
inviting
people
to
step
up
to
shadow
or
learn
more
about
leadership
roles
and
sig
Docs
and
contact
any
of
the
chairs
for
more
information.
B
You
know
if
one
of
the
co
chair
positions
opens
up,
you
know
they
could
step
in
to
do
that,
or
we
can
like
add
a
co-chair
position
where
that's
you
know,
based
on
terms,
as
you
know,
have
not
exactly
term
limit
but
likes.
Maybe
a
suggested
term.
Lengths
like
I,
don't
know
could
be
like
I
think
it
should
be
based
on
units
of
quarters.
Since
that's
kind
of
our
you
know,
our
release
cycles
are
based
on
that.
B
So
maybe
it's
like
three
quarters
or
four
quarters
or
something
like
that,
but
just
something
so
that
people
know
what
they're
committing
to
and
I
think
that
might
make
it
more
really
because
I
know
people
have
been
you
know,
are
they
have
regular
day
jobs
and
may
not
have
time
for
it?
But
if
there's
a
more
defined
I
think
term,
and
that
might
make
it
a
little
easier
to
know
what
their
commitment
is.
A
A
You
know
sort
of
meeting
with
any
of
the
regular
contributors
to
the
docs
to
find
out
more
about
how
we
do
things
to
help
document
processes
any
and
all
of
that
work
is
very
much
related
to
leadership
and
also
work
that
we're
happy
to
help
you
sort
out
if
you're
interested
and
aren't
sure
about
what
to
do
where,
where
to
start
and
spread.
The
word.
A
So
it's
not
yet
11
a.m.
on
the
west
coast
and
I
think
it
would
be
cool
if
we
could
hear
from
Steve
about
his
updates.
I
can
report
on
the
net
Liffe
eye
situation,
although
really
there's
no
news
except
what's
in
the
agenda,
which
is
that
the
outstanding
issue
from
last
week
appears
to
have
been
cleared
up
and
that
their
branch
is
good
to
go.
Have
you
seen
anything
to
the
contrary,
Andrew.
D
D
E
E
A
A
F
Yeah
I
can
do
that,
so
the
controller
factsheets.
The
idea
was
taken:
okay,
later
rised
from
sites
you
see,
for
example,
Google
or
or
sold
Wikipedia.
When
you
have
a
topic
like,
for
example,
you,
google
for
a
person
or
you
look
up
a
person
on
Wikipedia,
you
usually
get
a
fact
sheet
based
on
the
type
like,
for
example,
the
type
person
and
the
fact
sheet
gives
you
a
few
glances
like
birth
dates,
status,
ex-spouses
and
spouse
occupation,
a
picture.
What
have
you
so
for
our
controllers?
F
There
is.
There
is
a
possibility
to
include
a
fact
sheet
like
that,
to
get
a
quick
glimpse
of
what
the
controller
is,
what
it
does
which
objects.
It
reads
which
object
writes.
So,
for
example,
we
could
use
scheduler
as
an
example
or
the
replica
set
is
an
example.
What
type
of
controller
is
a
replica
set?
The
replica
set
is
a
core
controller.
It
only
reads
and
writes
from
the
kubernetes
object
store
what
objects
trigger
the
replica
set.
It
is
the
replica
set
object.
What
objects
does
a
replica
set
controller?
F
Modify
the
replica
set
controller.
Modifies
parts
can
create
parts,
it
can
update
parts.
It
can
delete
parts
and,
as
any
controller,
it's
the
trigger
object
there,
the
status.
So
you
see
that
in
a
fact
sheet
for
a
controller,
you
can
actually
communicate
a
lot
of
information
fair
fairly
efficiently
and
it
is
consistent
across
all
the
controllers,
because
the
controller
pattern
is
applied
over
and
over
again
in
the
come
in
the
controllers.
So
the
fact
she
did
self
is
fairly
stable.
B
Yeah
because
then
users
don't
have
to
think
of
them
as
totally
different
beasts
and
try
to
you
know
keep
all
that
in
the
head.
If
they
know
they
all
follow,
you
know
a
similar
pattern
and
then
they
can
whenever
they
need
to
reference
a
particular
control.
They
can
just
look
at
the
fact
sheet
and
just
they
should
get
a
good
idea
of
what
it
does
pretty
quickly.
F
And
also
what
it
doesn't
do
right
so
for
the
replicas
set
again,
you
want
to
understand
the
replica
said:
all
you
have
to
look
at
is
the
interaction
of
the
replica
set
controller
between
the
replica
set
objects
and
the
pod
objects.
Nothing
else.
Well,
nothing
else,
matters,
no
node
object
matters.
No
persistent
volume
claim
errors,
nothing
matters
to
it,
but
that
and
the
fact
sheet
would
actually
detail
that
out
pretty
quickly.
F
F
A
A
Chris
has
indicated
that
he's
interested
in
helping
out
with
getting
started
guide
if
anybody
else
is
also
interested
in
that
area
of
the
content.
Just
speak
up
on
slack
I
think,
and
we
can
take
a
small
working
group
sort
of
separately
from
this
meeting
to
discuss
what
would
be
useful
to
work
on
there
and
Steve
has
joined
us.
Oh.
G
A
G
G
I'm
glad
to
see
that
these
these
things
are
moving
forward.
The
things
we've
chosen
to
work
on,
as
you
know,
writing
or
updating
content
are
are
happening.
These
are
the
ones
that
that
I'm
involved
in
so
I
listed.
These
the
replica
set,
Cody's
work
on
the
ingress
and
and
Dominic
and
I
will
take
the
deployment.
Does
anybody
else
want
to
report
on
any
any
work?
They've
been
doing
on
the
on
the
content
updates
or
the
writings
new
content.
A
About
at
cube
con,
but
in
which
I
said
I
would
sort
of
lead
the
way
on,
but
have
not
in
fact
yet
kicked
off,
but
Chris
showed
up
and
said:
hey
I
was
there
and
I
offered
to
help,
but
I'm
here
and
I'm
offering
to
help.
So,
yes,
we
will
try
to
get
that
going
at
least
get
a
plan
in
place
and
I
will
the
one
piece
of
content
that
I
said
I
would
work
on?
Is
a
better
and
better
placed
introduction
to
Cuban
am
right.
G
A
Right
now
you
just
sort
of
dive
into
it
in
the
setup
section
and
there's
some
internal
content
in
the
reference
section
and
things
aren't
sort
of
where
you
expect
to
find
them.
Nor
do
they
really
I
think
provides
the
right
kind
of
top-level
picture
of
what
you
can
do
it
with
cube
ATM
what
you
need
to
be
aware
of,
etc.
B
Now,
at
the
revamped
Docs
homepage
you
could
add,
you
know
a
card
that
was
for
the
getting
started
and
stuff
and
put
it
over
there.
All
right,
I
wasn't
sure
if
you
wanted
to
create
like
a
new
section
again,
because
we
at
some
point
we
did
have
a
getting
started
like
section
and
then
we
deprecated
it
so
I
didn't
write.
A
B
A
Because
we've
been,
you
know
this
is
this
has
been
a
point
of
discussion
and
some
churn
for
a
while
now
and
I
mean
I
think
that
that
was
pretty
clear
in
figuring
out
what
the
links
ought
to
be
for
that
section
of
the
new
card
homepage
right
and
it's
not
it's.
That's
still
not
the
greatest
clearest
sort
of
entry
point
for
set
up
that
card
isn't
and
that's
not
because
of
any
problems
with
the
car,
its
problem
with
the
organization
content.
G
A
Right,
possibly
I,
just
I,
think
you're,
absolutely
right.
Steve
I
was
also
listening
to
you,
I'm
wondering
whether
getting
started
even
for
the
sort
of
introductory
get
to
know
kubernetes
for
the
first
time
content,
whether
getting
started
isn't
also
a
little
bit
of
a
misnomer,
simply
because,
historically
not
in
the
world
of
kubernetes
but
historically
getting
started.
A
If
which
are
important,
I
think
they're,
really
these
things
might
also
help
people
understand
better
sort
of
a
trajectory
of
learning
communities,
because
it's
one
thing
altogether
to
explore.
You
know,
deploying
and
and
and
deploying
engine
X
in
some
kind
of
a
sample,
sandbox
or
playground
and
another
saying
altogether
to
worry
about
spinning
up
your
own
clusters,
so,
whether
it's
with
a
cloud
service
or
on
your
own
servers
with
an
eye
to
eventually
getting
to
production.
It's
almost
those
are
almost
two
really
different
tracks:
right,
yeah,
I'm,.
G
A
G
Yeah,
it's
so
varied.
What
then
good
to
have
a
cluster?
The
people's
experiences
will
be
into
you
know,
really
varied.
Some
will
just
push
the
button
for
some
cloud
providers
cluster
and
that's
that
you
know
others.
It
will
not
be
that
easy
they'll
want
to
use,
create
their
own
cluster
real.
They
want
to
create
a
hybrid.
You
know,
hybrid
environment
is.
A
F
F
So
working
title
kubernetes
in
15
minutes
and
let
me
try
to
bounce
off
various
concepts
of
you
so
once
again,
this
is
not
about
the
usage
man,
it's
not
about
the
usage
of
kubernetes
in
15
minutes,
but
the
high-level
concepts
of
kubernetes
in
15
minutes.
And
the
first
point
is
the
introduction,
the
definition.
What
is
kubernetes
and
commonly
you
find
something
along
the
lines.
The
kubernetes
is
a
container
orchestration
engine
designed
to
manage
containerized
applications.
F
What
does
manage
actually
mean
well
managed
translates
into
a
set
of
commands
that
you
can
send
to
the
machine.
So
in
that
case,
two
commands
would
be
deploy.
You
need
to
deploy
a
binary
and
run.
You
need
to
start
the
binary
so
that
you
have
a
running
process.
Also
in
this
case,
it
is
up
to
you
to
monitor
the
processes.
So
if
you
have
an
application,
come
so
three
different
processes
with
one
of
these
processes
terminates
in
this
simple
model.
It
is
your
responsibility
to
bring
this
process
back
up
now.
F
Let's
look
at
the
first
dimension
that
is
containerization.
What
is
the
difference
between
an
application
and
a
containerized
application?
That's
actually
fairly
straightforward
and
instead
of
binaries
we
have
images
and
instead
of
processes
on
this
abstraction
level,
we
have
containers.
Yet
you
still
have
about
the
same
commands.
They
come
in
somewhat
different
flavors,
but
the
same
commands
you
need
to
deploy
an
image,
but
just
usually
done
via
a
pool
from
a
registry.
So
there
is
an
indirection
in
there.
F
You
need
to
run
an
image
to
get
a
run
in
container
and
once
again,
in
this
simple
model,
it
is
up
to
you
once
the
container
fails.
You
have
to
bring
it
back
up
right
and
the
question
now
when
we
look
at
the
dimension
of
infrastructure,
automation
is
where
does
kubernetes
come
in
and
the
interesting
part.
The
interesting
realization
here
is
that
kubernetes
takes
your
place
very
much
like
in
the
early
factories
in
an
assembly
line
that
were
actually
humans
at
the
assembly
line.
F
Assembling
the
car,
whereas
in
later
on,
a
robot
arm
would
take
the
place
of
the
human
kubernetes
does
very
much
the
same
thing.
It
takes
your
place.
You
tell
kubernetes
either
quote
unquote
what
you
want
or
quote
unquote
what
you,
what
to
do
and
kubernetes
issues
these
commands.
The
interesting
thing
here
to
realize
is
that
kubernetes
issues
say
exact
same
commands
that
you
would
issue
it's
like
deploy
the
container
I'm
sorry
deploy
the
image.
Yeah
run
the
container
if
the
container
dies.
F
So
here
the
realization
is
that
in
the
data
plane,
kubernetes
does
not
offer
anything
when
the
container
engines
already
there,
the
network
subsystem
already
there
storage
subsystem
already
there,
the
load
balancers
already
there
kubernetes
just
controls
or
manages
the
data
plane,
just
as
you
would
have
done
before
right.
So,
while
you
speak,
the
kubernetes
api
to
kubernetes
create
an
object
like
create
a
create
replicas
set
up.
F
F
The
kubernetes
can
issue
the
same
set
of
commands
that
you
could
issue
if
you
were
to
manage
it
manually.
So
then,
one
little
drill
down
into
the
kubernetes
control
plane
also
reveals
that
kubernetes
has
two
different
types
of
controllers:
that
I
just
named
core
controller
and
edge
controller
in
the
interim
core
controllers.
Are
the
controllers
that
read
or
query
and
modify
kubernetes
objects?
Tour
exclusively
like
the
replica
set
controller
will
never
touch
anything
else,
but
the
kubernetes
api
server.
F
The
scheduler
will
never
touch
anything
else,
but
the
kubernetes
api
server
and
then
you
have
the
edge
controllers
and
the
edge
controllers
are
the
ones
that
actually
make
the
the
change
in
the
data
plane
happen.
They
read
from
the
kubernetes
api
server
and
then
issue
the
appropriate
commands
towards
the
kubernetes
plane.
Best
example
for
an
edge
controller.
Is
the
cube
lit
right.
It
reads
the
information
about
the
pods
bound
to
this
node
and
then
issues
commands
to
create
the
containers
we
create
containers
or
bring
the
containers
down
in
the
container
engine
so
over.
F
Here
is
an
example
for
the
for
the
edge
controller
right,
and
we
talked
about
the
reconciliation
loop,
a
lot,
but
usually
we
also.
We
only
talk
about
50%.
That's
we
always
talk
about
this
pick
nine.
But
if
you
look
into
like
the
infinity
loop
of
kubernetes
right,
so
you
have
the
kubernetes
object
store
in
its
initial
state.
You
have
the
data
plane
in
its
initial
State
and
then
a
user
action
happens
like
I,
create
a
replica
set
or
I
create
a
pod
right.
So
the
user
changes.
F
The
kubernetes
object
store
state
here,
like
visualized
by
a
tick
but,
as
I
said,
it's
very
small
yeah,
so
I
will
put
the
tick
into
words,
but
does
not
change
the
data
plane
state.
Then
the
controller
does
the
first
50%
of
the
reconciliation.
That
is,
it
swings
to
specification.
It
does
not
change
the
kubernetes
object
store,
yet
it
changes.
The
data
plane
state,
for
example,
bring
up
this
comparing
up
this
con
container.
F
So
here
governor's
object,
store,
tick
and
the
first
change
to
the
data
plane,
state
tick
and
the
second
50%
of
its
job
is
to
swim
the
status.
Now
it
reads:
the
state
from
the
data
plane
does
not
change
the
state
from
the
data
plane,
but
changes
the
state
in
the
kubernetes
object,
store,
so
kubernetes
objects
or
ticktick
to
reflect
the
current
state
of
kubernetes.
B
F
F
Yes,
so
the
same
is
true
for
the
for
the
storage
system
right,
so
we
have
the
volumes
and
kubernetes
talks
to
the
storage
system,
whether
it's
the
Google
cloud
platform
or
AWS
or
Asia,
or
what
have
you
or
your
own
data
center
to
get
the
volume
mounted
on
the
node?
An
interesting
interesting
now
that
we're
at
that
topic
interesting
point
is
that
these
channels
are
usually
standardized
right.
You
can
actually
find
the
standard
for
it,
for
if
it
is
the
container
engine,
it's
a
container
runtime
interface
for
the
for
configuring.
The
network
subsystem.
F
It's
the
cni,
the
container
network
interface
for
the
storage
stuff.
It's
a
CSI,
the
container
storage
interface
right,
so
the
set
of
commands
towards
the
data
plane
is
actually
standardized
in
the
in
the
context
of
kubernetes
and
Marx.
Also,
very
nice
clean
cut
between
the
kubernetes
control
plane
and
the
structure
also
note,
in
that
case
right,
the
the
infrastructure
never
ever
ever
or
the
kubernetes
data
plane.
Never
ever
ever
comes
in
contact
with
any
kubernetes
object,
doesn't
know
about
any
semantics
of
kubernetes
objects
doesn't
know
anything
about
it.
F
C
So
I
took
some
notes.
So
when
you
talk
about
the
container
versus
processes,
you
got
to
be
a
little
careful
one
of
the
things
that
got
lost
in
and
how
it
was
coming
across.
Was
you
know
a
container
really
is
just
a
process
that
has
some
really
cool,
Linux
isolation,
security
features
with
it.
You
know,
you've
got
your
your
your
your
file
space,
you
know
your
layered
file
space
and
you
got
other
things
so
the
way
it
was
presented.
C
You
talked
about
containers
as
images
which
they
are
I,
get
that
but
I
always
try
and
emphasize
that
that
it
really
is
still
just
a
process
that
has
some
nice
features
that
isolate
it
with
its
own
little
file
system.
What
have
you
namespaces
and
and
that's
why
that
there's
the
power
of
containers?
They
they
like
a
process.
You
can
start
them,
you
can
stop
them,
really
kill
them.
A
snapshot
I'm
really
really
quickly.
C
So
so,
if
you
could
somehow
capture
that
I
think
it
would
help
help
folks
to
understand
how
close
those
are
and
then
I
did
like.
Obviously,
how
you
were
talking
about
how
kubernetes
becomes
you
that
was
great,
but
the
one
thing
that
got
lost
was
the
power
of
kubernetes
and
it's
declarative
model.
So
it's
it's
not
an
imperative
procedural.
In
most
cases,
it
really
is
a
declarative
model,
and-
and
that
gives
you
all
the
auto
scaling
and
the
failure
resilience
the
robustness.
C
So
if
you
could
somehow
make
sure
you
emphasize
that
it's
it's
got
a
declarative
model
and
you
know
if
the
state
changes
from
the
desired
state
to
what
it
needs
to
be.
It's
going
to
something
fails,
it's
gonna
start
up,
and
you
know
a
new
container
new
pod.
Rather
what
have
you
that
I
think
would
help?
C
So
if
you
get,
you
good
make
sure
those
key
features
that
that
have
made
kubernetes
so
successful
the
declarative
model,
the
built
in
auto
scaling
the
built
or
failure,
resilience
I,
think
that
would
help
as
you
go
through
the
presentation.
Obviously,
I
really
do
like
what
you
covered
and
and
don't
get
me
wrong.
I
thought
that
was
really
cool
and
it'll
be
very
helpful,
but
I
just
want
to
make
sure
those
key
pieces,
weren't
weren't,
forgotten.
F
Actually,
if
you
don't
mind,
I
jump
in
on
that
to
solve
for
perfect,
but
probably
years
from
now
then
find
the
time
for
that.
I
would
also
very
much
enjoy
like,
for
example,
a
section
of
inside
containers.
I
understand
that
kubernetes
is
an
infrastructure,
automation,
platform
yet
or
container
orchestration
platform.
Yet
I
think
it
is
it's
an
important
part
to
understand
the
essence
of
containers
and
yes,
exactly
on
an
operating
system
level
right
the
operating
system
doesn't
know
that
containers
exist
right.
The
the
reasoning
but
I
got
that
feedback
before
so.
F
Yes,
absolutely
now
the
difference
why
I
make
the
or
the
reason
why
I
make
the
difference
is
that
on
the
level
of
container
engine
the
container
knows
what
a
process
is
right.
It
has
an
I'm,
sorry
what
a
container
is.
The
container
has
an
ID
I
need
to
use
that
ID
in
order
to
talk
to
the
container
right.
So
from
a
user
experience,
point
of
view
process
doesn't
equal
container
from.
F
Reading
system
point
of
view
totally
agree
same
thing
right,
so
that's
why
I
also
this
here
right,
it
looks
like.
Oh,
the
operating
system
manages
a
set
of
processes
right
and
here
it
looks
like
now.
The
Dayna
engine
manages
a
set
of
containers.
If
you
wanted
to
be
actually
very
exact,
you
would
need
to
put
that
into
another
thing
here
right
and
convey
the
fact
that
this
is
still
actually
managed
by
the
operating
system
managing
the
the
processes.
It's
so
far.
F
I
have
shied
away
from
it
a
little
bit
because
it
complicates
the
picture
but
yeah
there
is
I
have
one
sentence
here
so
far.
However,
that
sentence
is
terrible,
cannot
be
used
yeah,
but,
as
you
see
here
in
the
user
right,
there
is
what
does
the
user
care
about?
I
was
also
thinking.
The
declarative
state
right
replica
said
somewhat
of
an
interesting
example
right.
It's
like
okay,
replica
set
declares
I
want
three
parts.
That's
like
again
from
the
point
of
user
I,
don't
care!
F
Yes,
like
replica
said
pod
I
won't
contain
us,
so
the
user
has
a
desired
data.
Plane
state
never
desire,
kubernetes
object,
store
state,
couldn't
care
less
right
and
there
is
a
data
plane
state.
So
now
that
sentence
as
I
said,
this
is
like
don't
use
it
anymore.
It's
like
the
user
needs
to
translate
the
desired
data
plane
state
into
a
set
of
kubernetes
objects
such
that
the
state
of
the
data
plane
equals
the
desired
state
of
the
data
plane.
F
When
kubernetes
is
in
a
steady
state,
it's
a
terrible
sentence
but
fairly
accurate,
but
yes,
that
that
it's
like
you're,
not
you're,
not
giving
kubernetes
commands
right,
you're,
you're,
basically
creating
a
representation
of
the
desired
data
plane
state
within
kubernetes
and
kubernetes
just
works
tirelessly
to
make
sure
which
is
still
cool,
because
this
is
kind
of
how
a
user
works
right.
If
I
have
two
notes
with
a
wordpress
application,
my
desired
state
WordPress
up
Postgres
up
or
my
scale
or
whatever
you
using
right
and
then
I
worked
tirelessly
to
do
that.
E
F
And
the
interesting
thing
here
also
to
notice,
usually
I,
use
this
sentence.
Kubernetes
a
container
orchestration
engine
designed
to
host
container
eyes
application,
actually
doesn't
host
nothing
right.
Kubernetes
itself
doesn't
run
anything
for
you,
kubernetes
just
manages
stuff
for
you
and
on
the
data
plane,
which
is
all
the
stuff
you
already
had
in
your
data
center,
but
kubernetes
automates.
That.