►
From YouTube: Kubernetes SIG Apps 20171106
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
A
So
when
you
see
sig
apps
and
the
thing
that's
who
is
talking,
and
so
with
that
we'll
jump
into
the
agenda
here,
there
are
a
whole
bunch
of
announcements,
so
I'm
gonna
walk
through
some
of
them
and
there
might
be
others
who
added
some
who
want
to
dig
in
a
little
bit
to
it.
So
I'll
see
if
you're
here
the
first
one
is
the
meeting
time
and
time
zones.
A
So
some
of
you
may
have
noticed
that
what's
on
the
community
page
refers
to
an
hour
earlier
because
it's
set
in
UTC
and
yet
the
meeting
started
at
a
different
time.
We
kind
of
around
these
daylight
savings
time
time.
Zone
changes
have
a
little
bit
of
shuffle
instead
of
going,
and
but
yet,
if
you
go
to
the
kubernetes
calendar
at
kubernetes,
dot,
io
/
community,
it
had
a
different
time.
It
actually
says
9:00
a.m.
A
Pacific,
because
that
calendar
set
to
Pacific
time
and
that's
when
the
meeting
invite
is
set
for
and
so
I
kept
it
today.
As
the
meeting
time
that
the
meeting
was
set
for
figuring,
people
would
show
up
for
that,
rather
than
the
shift
on
the
community
repo
and
because
we
have
such
a
presence
in
the
Pacific
time
zone.
We'll
probably
keep
that,
and
so
it's
going
to
be
9:00
a.m.
Pacific
time,
we'll
update
the
community
repo
today
I
know
it
gets
a
little
weird
setting
around
and
positioning
around
that
kind
of
time
zone.
A
A
The
next
one
is
you'll
see
a
link
to
there
is
another
I
wanted
to
share
this
real
briefly
I
know
people
are
working
on,
go
apps
at
interface
with
kubernetes,
and
there
is
another
go
client.
That's
been
talked
about
the
community
a
bit
lately
and
so
I
wanted
to
just
share
that
with
folks
to
go.
Take
a
look
at
eventually.
We
will
probably
be
curious
to
talk
more
about
it,
but
I
just
want
to
throw
that
out
there.
A
The
next
one
is
the
app
definition
working
group
update
I,
wanted
to
just
give
a
quick
update
on
this,
because
we
had
our
first
official
meeting
as
in
the
app
definition
working
group
and
the
things
to
give
is
we
spent
a
little
bit
of
time
focusing
on
what
are
the
goals
of
this
working
group
just
to
clarify
it?
So
that
way
we
don't
go
off
the
rails
or
go
into
all
kinds
of
other
places.
In
addition
to
that,
Antoine
had
a
demo
on
a
way
to
kind
of
organize
applications.
A
He
implemented
it
through
a
CRD
I
think
was
called
packages,
and
so
there
was
a
demo
there.
It
was
not
recorded.
So,
unfortunately,
I
don't
have
a
link
to
it.
The
goals
that
we
came
to
in
that
that
we're
still
iterating
on
are
to
help
drive
the
actual
completion
of
issues
with
regard
to
like
the
sig
CLI
interfaces,
so
things
like
apply
and
dry
run
and
some
of
these
things
that
are
currently
in
coop
control.
How
do
we
help
facilitate
them
with
some
of
their
goals?
A
The
second
one
is
to
do
a
survey
and
pull
up
examples
to
provide
guidance
for
folks
who
want
to
do
definition,
things
to
try
to
understand
the
personas.
What
are
the
needs
and
where
are
we
going
with
it?
And
then
the
third
one
which
we
haven't
douve
into
very
much,
is
to
look
at
interoperability
between
tools
and
some
of
the
conversation
path.
Sense
and
is
even
said.
A
Then
the
next
two
and
I
don't
know.
If
somebody
else
is
here
to
talk
about
it,
but
I'll
talk
about
it
with
coop
gun.
We
have
to
see
gaps
meetings
at
coop
con
one
is
Thursday
December
7th,
which
will
be
an
update,
that'll,
be
a
half
an
hour
and
it's
similar
to
the
updates
we
give
in
community,
so
we'll
be
kind
of
highlighting
everything
that's
going
on
there
and
then
the
next
one
is
on
Friday.
The
8th.
B
A
So
the
telepresence
yeah
there's
one
on
here,
I'll
just
read
it.
It
says
the
telepresence
team
is
testing
a
more
robust
method
that
proxies
a
local
container
to
your
remote
container
for
fast
development,
they'd,
be
interested
in
feedback
and
there's
a
Doc's
link
here.
So
folks
want
to
take
a
look
into
that
I'm
not
familiar
with
it.
It
was
just
something
they
obviously
added
here
to
talk
about
so.
C
A
All
right,
so
the
first
one
we
have
up
here
before
the
demo
and
our
discussion
topics
is
I,
got
a
short
discussion
on
here
for
the
clients
and
SDKs.
These
are
formerly
owned
by
API
machinery
when
it
comes
to
generating
things
and
stuff
like
that,
but
they
haven't
had
much
time
to
actually
dig
into
the
UX
or
the
implementation
or
the
other
things,
and
so
now
ideas
have
been
thrown
out
about
a
cig,
SDK
or
a
working
group,
or
something
like
that
to
try
to
improve
them.
A
There's
also
work
being
done
like
I
shared
earlier
in
the
meeting
with
another
go
client
and
there's
actually
a
number
of
clients
out
there,
there's
one
for
Python
that
are
in
the
community.
Things
like
that,
and
so
there's
talk
now
about.
How
do
we
kind
of
attack
this
problem
and
where
do
we
go
with
it
to
have
better
shared
tooling
and
so
without
getting
into
the
going
and
actually
building
it?
Is
there
anybody
here
or
around
or
do
we
know,
folks
who
are
actually
interested
in?
A
How
do
we
organize
around
that
and
work
together
on
it
for
that
shared
tooling,
because
we
already
have
like
a
kubernetes,
client,
org
and
there's
some
open
API
work
being
done
there?
Is
there
anybody
here,
who's
actually
work?
You
know
interested
in
jumping
over
there
and
helping
I
can
I
can
plug
you
into
the
right
place.
You
can
reach
out
offline,
but
that's
kind
of
what
I
wanted
to
talk
about
was.
How
did
we
say?
Gaps
want
to
engage
in
that?
C
I've
been
following
identity,
girl,
sick
absence
again,
machineries
I've
been
following
a
lot
of
the
client
work
Polly
as
well.
I'd
been
with
the
cuvette
committees,
json
schema
stuff,
finding
bugs
in
the
swagger
specifications
which
a
lot
of
the
clients
are
using
but
haven't
had
enough
time
to
do
anything
more
active,
I.
C
Think
there's
like
one
of
the
things
that's
cropped
up
there,
as
well
as
the
sort
of
the
weird
self
first-class
nature
of
the
girl
client
and
which
I
think
is
sort
of
like
one
of
the
reasons
why
I
think
we
probably
see
a
lot
of
tools
written
go
in
cig
apps
is
because
of
like
that.
It's
just
easier
from
a
client
perspective
because
they
actually
they're
not
generated
by
a
via
external,
so
I
guess
tough
at
the
moment
that
sort
of
a
weird
vendor,
Native
communities,
yeah.
A
And
it
changes
with
every
release
of
kubernetes.
You
get
a
major
API
breaking
a
release
which
the
best
I've
heard
is
people
taking
hours
to
upgrade
some
people
take
days
to
weeks,
yeah,
it's
a
because
it's
it's
staged
and
it's
supposed
to
be
synced
out
and
designed
to
be
separate,
even
though
actually
pulling
it
out.
I
believe
is
currently
on
hold,
while
all
their
work
is
being
done.
E
Really
go
ahead.
I
think
one
of
the
best
things
we
can
do
here
is
try
to
provide,
like
churning.
We've
always
had
this
notion
of
what
our
promise
is
in
terms
of
supporting
backward
compatibility
and
supporting
an
API
version
once
it's
released,
but
we
also
have
a
promise
with
respect
to
how
long
we're
going
to
leave
something
in
beta
before
we
pretty
stable
and
we've
been
a
little
bit
lacks
on
that.
A
A
A
Okay,
so,
but
this
this
is
really
talking
about
go
and
I.
Think
an
important
point
was
brought
up
earlier,
that
a
lot
of
folks
like
to
write
applications
that
interact
with
stuff
in
other
languages,
Python
Ruby,
a
number
of
different
languages,
and
so
just
having
CL
eyes,
and
you
know
an
SDK
or
client
in
go.
That's
a
first-class,
probably
isn't
all
that
we
want
or
need,
and
so
how
goes
in
teresting
in
helping
get
some
of
this
other
stuff
going.
E
D
Okay
to
your
point,
Matt
like
what
I
guess
what's
the
next
step
here,
is
there
already
a
cig,
that's
talking
about
this,
or
is
this
looking
to
kind
of
organize
that
effort.
A
C
C
He
was
he's
actually
come
up
with
a
whole
scheme
of
badging
all
the
different
clients
in
terms
of
where
they
are
from
so
yeah.
If
you
actually
go
into
the
like,
if
you
go,
go
and
look
at
all
the
different
clients,
they
have
these
sort
of
committees,
client
touches
and
they're,
either
in
like
two
alpha
beta
or
their
I,
think.
C
However,
the
like
having
them
all
so,
they
can
all
be
automatically
generated
from
all
of
the
the
swagga
definitions
and
having
all
of
that
then
have
an
interface
that
feels
native
to
people
is
sort
of
quite
a
bit
a
challenge.
But
maybe
it
is
this
odd
person
who's
had
time
across
their
variants.
There.
D
A
A
Something
at
all
so
I,
maybe
a
different
way
to
say
this,
though
it's
not
to
say
what
do
we
have
to
do
like
a
company
providing
them,
but
we've
got
a
community
of
people
who
are
developing
apps.
How
do
we
enable
people
who
are
in
that
community
who
are
not
current
contributors
today?
Who
are
interested
in
these
problems
to
know,
there's
a
place
to
get
involved
and
to
get
involved
because
I
don't
want
to
say
okay,
this
is
an
extra
burden
on
people
who
are
already
busy
today.
A
I
want
to
look
at
and
say,
there's
a
whole
lot
of
app
developers
we're
struggling
with
these
problems
if
they're
interested
in
it.
Maybe
they
who
don't
contribute
today
are
willing
to
work
together
to
solve
this
common
problem
for
them,
because
there's
a
lot
more
app
developers
out
there
who
write
stuff
that
talk
to
api's,
then
and
even
show
up
here
regularly
I.
D
A
Yeah
and
I'm
thinking
is
there
a
way
to
facilitate
that
rather
than
try
and
do
something
else,
the
client,
Python,
okay
and.
E
There's
a
quiet
job
as
well
may
he's
been
kind
of
working
on
both
of
those
and
there
may
be
a
plenty
more
out
there.
These
are
three
that
I
know
that
are
actually
induced
by
other
projects
and
that
people
had
internally
started
to
contribute
to
already
so
those
are
like
if
you're,
a
Java
developer
or
a
Python
developer
and
you're
looking
to
pitch
in
those
are
some
things
that
you
can
start
with.
But
right
now,
okay,.
A
How
do
we
start
enabling
this
stuff,
or
how
do
we
come
up
with
a
plan
to
do
it
because
I,
don't
I,
really
don't
want
to
ask
more
people
who
are
already
really
busy
to
take
on
more
work,
so
anybody
who's,
interested
and
I
grabbed
a
couple
of
names
here
will
try
and
start
looping
a
few
people
together
over
the
coming
weeks,
not
like
this
week
or
next
week,
but
the
coming
weeks
to
figure
out
how
do
we
attack
this
problem?
Should.
A
A
A
F
F
The
basics
of
this
were
outlined
in
the
tutorial
I
made
last
week,
but
what
I
wanted
to
do
is
integrate
Jenkins,
Griffin
and
Prometheus
to
automate
the
process
for
not
only
deploying
but
issuing
row
backs
when
things
don't
go
well.
So
our
current
pipeline
is
to
push
to
a
canary
branch
on
github,
have
Jenkins,
listen
for
those
events,
build
a
new
docker
image
with
the
updated
application
deploy
a
canary
version
of
our
deployment.
F
Have
that
get
deployed
to
kubernetes
and
have
Prometheus
and
gravano
then
scrape
metrics
from
that
new
deployment
and
maybe
potentially
fire
alerts
depending
upon
whether
things
don't
go
well.
So
that's
the
overall
pipeline
when
things
get
merged
into
master,
will
update
that
docker
image
and
we
will
delete
the
canary
application
so
right
now,
I
have
the
current
baseline
up
and
running
from
that
canary
app.
So.
F
So
if
people
have
a
terminal,
they
can
start
watching
the
responses,
or
you
can
see
it
here.
It's
a
very
simple
application
that
scrapes
the
environment
inside
of
kubernetes
to
look
for
particular
labels
and
that
are
that
are
present
on
the
pod.
So
we
get
return
to
JSON
object
that
has
the
pod
name.
What
app
the
application
is
running
in
ie?
F
What
deployment
a
label
for
the
release
version
and
a
master
that
release
will
be
tagged
as
stable
and
on
a
canary
branch
should
be
tagged
as
canary
and
then
any
additional
labels
that
are
in
the
that
are
on
the
pod,
and
here
we
see
I
appended
the
kid
commit
for
that
particular
one.
Here
we
can
see
the
the
pods
running
in
the
namespace.
F
F
F
We
make
sure
that
we
replace
the
app
name
with
app
info
canary,
and
we
only
have
one
replica
instead
of
three
and
we
set
this
release
to
be
canary
instead
of
stable
and
then
for
the
master,
we're
just
updating
the
docker
image.
So
it's
a
pretty
straight
set
up
a
lot
of
it
was
taken
from
another
tutorial
that
GCP
provided
on
doing
this
and
it's
a
very
similar
process.
F
So
if
we
in
our
app
info
service,
we
see
that
there's
the
problem
lines
here
that
we're
not
pulling
the
namespace
value
correctly
from
the
environment,
variable
that
we're
passing
in.
We
also
have
this
timeout
that
we've
commented
out,
and
this
is
going
to
simulate
when
we
uncomment
this
a
poorly
performing
canary
deployment,
where
responses
are
coming
back
correctly,
but
maybe
your
SLO
times
are
two
seconds
respond,
so
this
is
going
to
guarantee
a
three-second
response,
so
we're
going
to
simulate
what
happens
here
when
we.
F
So
we've
done
is:
we've
checked
out
a
version
new
branch
on
our
app
repository
we've
pushed
into
the
canary
branch
github
has
notified
Jenkins
that
there
is
now
a
new
version
of
the
canary.
It
is
going
through
and
building
as
we
see
this
conclude
its
build.
We
should
see
a
change
in
the
responses
here
to
show
that
there
is
a
now
a
canary.
Oh
we
see
over
here
on
the
the
pod
side
that
we
have
a
canary
version
of
our
pod,
and
we
saw
that
this
changed
back
from
from
the
canary
there.
F
This
is
the
graph
on
a
page.
It's
going
to
be
in
the
repo
the
kind
of
documents
this
entire
tutorial.
It's
showing
the
response
time
for
the
different
app
info,
backends,
the
stable
version
of
the
canary
version
and
right
now,
there's
only
a
stable
version
hasn't
been
scraped
by
Prometheus.
Yes,
but
as
we
get
updates
great,
we
see
that
there
is
now
a
version
that
has
a
canary
release
tag
that
is
taking
over
three
seconds,
which
is
very
significant.
F
We
see
this
alert
this
graph
down
here
has
an
alert
built
into
it
that
if
it
goes
over
two
seconds
which
was
our
fake
SLO,
it's
going
to
fire
an
event.
So
we
look
back.
The
event
that
gets
fired
is
actually
just
the
Jenkins
job
called
canary
rollback,
which
is
very
simple,
simple
and
it
just
fired
Brand
12
seconds
ago,
and
if
we
look,
we
now
no
longer
have
our
app
info
canary
deployment
that
was
deleted
and
all
these
are
coming
back
from
the
stable.
F
Now
the
rollback
feature
I
put
in
as
a
Jenkins
build.
It
doesn't
have
to
be
that
it
could
be
some
sort
of
other
script
that
could
be
run.
I
just
wanted
to
kind
of
consolidate
the
the
stack
that
I
was
using
to
kind
of
leverage
the
same
pieces
and
that
Jenkins
around
can
also
be
manually
run
if
things
are
looking
right,
but
they
don't
need
some
alert
mechanism.
That's
built
into
your
monitoring
system.
F
Now
this
middle
graph
here
will
document
any
error,
responses
that
come
back
from
the
service
and
we
don't
show
any
of
them
happening
here,
but
that
might
also
be
another
SLO.
That's
that's
important.
You
guys
making
sure
that
you
have
200
response
codes,
some
percentage
of
the
time.
Now
we
can
look
at
the
responses
coming
in.
We
see
our
canary
deployment
come
in
we're
seeing
the
proper
name
space
value,
get
populated
when
Canaries
there
we'll
start
seeing
our
response
times
go
down
once
it
gets
enough
data
to
get
a
new
version.
F
F
F
And
we're
not
going
to
squash
and
merge
here,
because
that
would
rewrite
the
canary
branch
and
then
you'd
have
a
race
conditions
between
the
updated
canary,
build
getting
run
and
the
master
build,
getting
run
and
master
would
try
to
delete
the
canary
app
and
at
the
canary
build,
would
try
to
deploy
a
new
version
of
the
canary.
So
you
can't
really
do
that
as
and
then
successfully
delete.
Our
canary
branch
will
see.
Jenkins
start
our
master
builds.
F
H
F
A
F
A
F
Yeah,
so
I
wanted
to
get
a
lot
of
people's
feedback.
We
don't
use
kubernetes
yet
in
production,
we're
on
a
marathon
stack
right
now
and
we're
trying
to
move
away
from
that
for
obvious
reasons.
So
from
trying
to
figure
out
how
to
do
the
things
you
do
there
on
kubernetes,
it
seems
like
a
lot
of
people
have
been
using
third-party
tools.
Iste
o
is
a
common
one
for
controlling
your
canary
deployments
and
there's
a
good
or
I
thought.
H
H
One
is
to
be
very
interesting
to
if
you're
gonna
use
Jenkins
to
use
pipeline
just
so
that
you
can
use
blue
ocean
and
some
of
the
more
perfect
he's
you
throw
them
up-to-date,
Jenkins
like
loveliness,
if
you're
gonna
use
it,
and
secondly,
I
was
wondering
if
you
adapted
this,
to
use
a
a
church.
You
know
element
charts,
not
kind
of
just
over
complicate
your
demo
or
do
you
think
that
would
be
a
good
way,
maybe
a
good
step
for
moving
this.
You
know
I,
guess
canary
demo,
into
something
more.
What
to
do.
F
Agreed
the
infrastructure
can
be
complicated
in
setting
ups
and
I
did
use
helm
for
the
first
time,
trying
to
setup
and
deploy
a
Prometheus,
earphone
and
Jenkins.
There
were
some
things:
I
had
to
modify
slightly
to
get
working
and
I
think
it
worked
well,
because
I
took
mini,
cube
first
and
stood
it
up
locally
and
then
I
really
wanted
the
github
notifications
working
so
that
I
didn't
have
to
go
manually,
run
things
because
I
know
that
always
looks
poor
in
the
demo.
F
F
So
I'd
love
someone
else
to
walk
through
that
and
see
if
they
get
the
same
end
result
that
I
got
just
to
make
sure
that
it's
you
know
repeatable
and
for
someone
who's
used
Jenkins
in
the
past,
word
of
bamboo
shop,
in-house
and
I
didn't
want
to
buy
a
bamboo
license
for
some
public
demo.
I'd
love
to
get
your
feedback
on
how
that
could
have
been
improved
and
best
practices
there
I'm,
not
a
Jenkins
person,
cool.
E
One
other
alternative
type
of
canary
and
staged
rollout
that
I've
seen
with
deployment,
that's
also
prevalent
so
you're,
creating
a
new
deployment
for
your
canary
and
updating
the
main
deployment
after
you're
satisfied
with
the
help
of
that
canary.
The
other
way
of
seeing
this
done
is
to
have
statically
sized
deployments
so,
like
you
might
have
another
point
in
that,
I
was
one
in
an
actual
canary.
Then
you
have
a
deployment.
That's
got
like
10%
of
the
fleet,
which
is
like
a
staged.
E
Rollout
and
you've
got
another
deployment,
which
is
the
other
89
percent
of
your
fleet,
and
rather
Li
create
a
new
deployment.
They
just
apply
an
update
to
the
canary
first
and
they're,
satisfied
with
that.
They
update
the
deployment
that
represents
10%
and
they
update
the
larger
one
when
they're
satisfied
after
they've
taken
10%
of
it
any
thoughts
to
doing
it
that
way
or
this
way
or
why
you
choose
just
to
do
it
by
create
new
deployment,
as
opposed
to
the
more
statically
articulate.
F
No,
no,
no
real
reason
from
my
end,
I
thought
it
made
it
clear
when
the
deployment
was
active
and
not
based
on
the
presence
of
that
deployment
object
and
in
kubernetes.
So
from
my
end,
seeing
the
the
change
in
state
of
the
cluster
was
easier
using
a
brand
new
deployment
rather
than
updating
versions.
Along
that
end,
one
thing
that
would
be
nice
about
either
of
those
is
kind
of
having
this
automated
rollout
or
upgrading
of
the
canary
deployment
could
happen
regularly
as
like
a
nightly
build,
and
you
know,
you've
got
in
your
requirements.
F
Documents
got
to
work
for
a
day
and
then
it
gets
staged
up
to
the
next
one
and
it's
a
three
day
rollout
or
whatever
you're.
You
know
you're
required
to
do
it's
nice
when
it's
it's
just
a
present
deployment
there
that
you
can
update,
rather
than
having
to
do
checks
to
make
sure
it's
whether
it's
present
or
not,
or
treat
things
differently
depending
on
what
state
it's
in.
H
It
and
just
a
very
quick
cyber
on
that,
her
piggyback
on
that,
as
it
kind
of
depends
on
what
what
app,
what
type
of
app
you're
deploying
right.
If
you're
talking
about
a
stateless
app,
that's
you
know,
staged
builds
have
a
totally
different
meaning
than
if
you
have
a
stateful
app
that
has
shared
live
data.
E
E
The
community
documentation
for
staple
set
when
we
added
rolling,
update
documents,
how
to
do
staged,
rollouts
and
canary
updates
on
sample,
said
using
the
partition,
and
then
there
was
a
presentation
in
big
apps
when
the
same
infrastructure
was
used
to
implement
lifecycle,
hooks
for
staples
us
and
there's
over
proposal
for
lifecycle
hooks
based
on
that
as
well.
Doom
links
Anthony
that
sounds
awesome.
H
E
It's
it's
still
under
discussion,
but
there's
been
multiple
attempts
to
do
that
type
of
auto
pause
with
deployment
and
demons
that
it
staples.
That
offers
a
much
more
clear
semantics
in
terms
of
length
of
Oz,
given
the
way
that
expectations
work
inside
of
the
claimant.
It
is
very
hard
to
implement
something
like
Auto
pause
in
a
reliable
and
predictable
way.
E
E
A
I
know
the
docs
folks
are
looking
at
restructuring
the
docs
to
try
to
make
it
a
little
bit
better,
but
even
as
they're
they're
doing
that,
and
that's
something
if
you're
curious,
you
weren't
here
last
week
we
just
last
week.
Last
week
we
talked
about
it.
This
is
something
we
should
probably
get
into
the
documentation
somewhere.
So
if
it's
not
already
there
and
you're
interested
in
contributing
something,
Doc's
are
pretty
good
way
to
go.
A
All
right
there
are
please
go
ahead
and
bring
it
up
on
the
mailing
list
and
we
can
keep
this
going
or
bring
it
up
again.
So
with
that
I'd
like
to
move
to
the
next
item
on
the
agenda,
which
we
have
Dan
here
from
the
CN
CF
to
talk
about
the
CN
CF
is
organizing
a
certified
kubernetes
application,
developer,
curriculum
training
and
certification
process,
I'm
kicking
off
around
goop
con
and
Dan
I
I
see.
Are
you
still
on
I
sure,
AM
Thanks?
All
right
the
floor
is
yours.
You
got
about
15
minutes.
Oh
that's,.
B
B
They'd
be
great,
if
you
don't
mind
all
right
just
a
moment
or
folks
can
just
pop
it
up
here
so
I
think.
Hopefully
most
people
are
familiar
with
CN
CF,
we're
the
nonprofit
Software
Foundation,
that
hosts
kubernetes
and
a
number
of
other
projects.
Our
parent,
the
Linux
Foundation,
has
an
internal
training
group
that
we've
built
out
over
the
last
several
years.
That
has
a
like
a
very
successful
training
for
Linux
system,
mins
for
OpenStack
for
Cloud
Foundry,
and
then
we
just
launched
the
kubernetes
certified
kubernetes
administrator
a
year
month
ago.
B
So
if
you
go
to
slide
two,
you
can
see
that
the
the
model
here
is
that
we
pick
a
topic
and
then
we
pull
together
a
group
of
experts
and
jointly
design
a
curriculum
that
curriculum
is
open
source.
We
then
do
a
free
intro
course
through
our
partnership
with
EDX,
which
is
the
MIT
online
training
group
and
and
we've
gotten
tons
of
excitement
and
energy
through
that,
like
folks
for
141
different
countries
have
signed
up
for
the
kubernetes
one.
B
We
then
do
paid
training
course,
299
bucks
that
is
tied
to
the
curriculum
and
then
an
exam,
and
the
key
idea
here
is
that
the
exam
is
not
a
multiple-choice,
but
it
actually
you
a
command-line
and
requires
you
to
go
in
in
the
CK
a
case
to
configure
a
kubernetes
cluster.
So
that's
that's
also
300
bucks,
although
when
you
get
the
two
together,
it's
five
hundred
so
slide
three.
The
key
thought
is
that
we're
assembling
the
industry
experts,
but
we're
not
actually
trying
to
define
the
exam
or
I
mean
what
what
the
topic
should
be.
B
We're
really
looking
for
the
group
of
people
who
come
together
on
it
to
do
that.
We
also
fund
the
development,
then
on
the
internal
LF
training
group
designs
it
using
some
outside
experts
and
develops
and
runs
the
training
exam,
and
then
we
do
encourage
third
party
training
partners.
So
the
idea
is
that
we
kind
of
the
curriculums
open
source
the
exam
is
kind
of
monopoly,
but
the
training
we
encourage
lots
of
folks
to
compete
with
us
and
then
offer
the
coupons
for
the
exam
at
the
end
of
it.
So
in
slide
4.
B
The
exam
has
gone
quite
well,
so
far.
There's
definitely
been
some
teething
issues
on
it.
There's
built
around
1.6
we
had
are
now
funding
it
to
do,
for
updates
a
year
to
match,
cover
news
releases
and
in
particular
it
had
some
canonical
isms
that
people
were
not
happy
with
so
those
are
being
removed
and
it's
about
two
weeks
away
from
having
the
1.8
version
of
it,
and
then
we
hope
the
fisa
versions
are
even
closer
in
on
the
release
of
kubernetes.
We're
gonna
do
a
Chucky's
translation
of
all
the
text.
B
B
Slide
5
is
a
little
bit
on
our
aspirations
where
we
would
like
to
offer
a
whole
set
of
add-on
exams,
particularly
once
you've
become
a
certified,
kubernetes
administrator
and
there's
an
analogy
here
to
scuba
diving,
where,
just
like
you
can
do
deep
dive,
a
wreck,
dive
or
underwater
photography
that
you
could
imagine
having
things
that
cover
configuring,
envoy
or
ISTE,
o
or
Prometheus,
or
foodie,
or
security
or
networking
or
all
sorts
of
other
aspects
of
it.
But
the
first
part
that
we'd,
like
to
start
on
is
slide.
B
6
super
certified
in
kubernetes
application,
developer
and
I.
Think.
The
headline
point
that
I'd
like
to
make
to
this
group
is
that
we
feel
like
there
is
a
natural
compliment
to
the
administration
exam
that
in
any
given
company
you're,
probably
only
gonna
have
two
or
three
people
who
really
should
become
CK
is,
but
you
definitely,
but
you
should
then
correspondingly
have
dozens
or
even
hundreds
of
different
developers
who
are
going
to
be
creating
applications
that
deploy
on
to
that
kubernetes
application.
B
And
so
this
has
been
the
case
with
OpenStack,
with
cloud
foundry
with
others
that
we
think
that
there's
a
natural
kind
of
application
developer
exams.
That
necessarily
should
be
easier
shorter,
not
as
involved
as
the
kubernetes
administrator
point
and
then
the
the
other
kind
of
headline
point
here
is
that
yes,
this
this
area
is
changing
dramatically
right
now
and
so
I
can
easily
imagine
that
every
three
months
there
will
be
much
much
more
churn
in
this
content
in
the
curriculum,
the
exam
etc.
Then
there
isn't
a
certified
kubernetes
administrator
exam,
but
we're
okay.
B
With
that
we're
willing
to
fund
the
effort,
we
think
the
experts
with
Quinn
are
willing
to
keep
this
up
to
date,
and
so
you
know,
hypothetically,
if
nine
months
or
twelve
months
from
now
building
a
helm,
shard
is
no
longer
best
practice
and
instead
there's
a
standard
tool.
Not
that's
universal,
but
70
or
80
percent
of
people
are
interested
using
that
tool.
That,
then
you
know
produces
the
home
chart
then
great.
B
Let's
change
this
there's
absolutely
no
lock-in
on
any
of
this
content,
but
the
goal
would
be
that
for
each
three
months
that
we
issue
this.
It
would
represent
something
of
the
best
practices
today,
and
so
you
can
see
in
some
things
in
here
like
how
to
design
twelve
factor
apps
and
when
you
should
sometimes
want
to
break
out
from
the
guardrails
that
should
not
actually
age
the
Prometheus,
metrics
and
alerts.
It
is
probably
pretty
steep
stateful,
so
it's
they're
stable,
as
are
I,
think
the
cron
jobs
and
staple
sets,
but
I
do
an
emphasize.
B
These
bullet
points
are
six
of
them
here,
we're
just
my
guess
at
what
it
should
be
to
sort
of
calibrate
things.
It
really
is
up
to
the
group
of
experts
to
decide
what
they
think
should
be
in
it,
and
so
then
that
slide,
seven
that
we're
looking
for
curriculum,
design,
volunteers
and
slide
age
shows
the
schedule
that
you
arrive
Trudie's
early
for
cube
con
and
then
Monday
and
Tuesday.
We
go
through
and
come
up
with
what
we
call
a
job
task
analysis
where
we
talk
about.
B
But
the
goal
would
be
that
whatever
we
ask,
people
to
do
wouldn't
actually
require
you
to
be
familiar
with
JavaScript
applications
or
PHP
or
whatever.
It
is
that
it
would
be
more
Universal.
So
that's
the
background.
It's
completely
open
to
anyone
to
participate,
I'd
love
to
get
engagement
from
this
group
and
I.
Believe
man
is
already
volunteered.
There's
no
cost
involved
other
than
obviously
all
of
your
time
is
extremely
valuable.
So
we,
you
know,
respect
the
effort
berth
for
people
willing
to
contribute
here
and
I'll
stop
and
take
any
questions
or
accusations.
B
Gotta
say
the
outside
of
its
just
a
tiny
bit
problematic
and
I.
Do
really
appreciate
that
that
and
limits
things,
but
we've
just
found
with
these
groups
and
then
again
just
being
really
bluntly
here,
the
OpenStack
effort
extended
to
like
12
months
because
they
were
trying
to
do
it
on
weekly
conference
calls
and
for
kubernetes.
We
got
all
the
people
in
the
same
room
over
two
weekends
and
accomplished
more
faster
yeah.
B
I
Hi
from
the
kubernetes
certified
administrator
syllabus
and
the
one
that
we
saw
for
the
application
developer,
there
seems
to
be
some
sort
of
overlap.
So
what
are
the
plans
I
mean?
Are
we
going
to
remove
any
traces
of
application,
related
maintenance
in
DCAA
and
keep
everything
in
this
certification
or
house
how's
the
plan
going
forward
the.
B
Aspiration,
of
course,
I
mean
we're
looking
to
these
expert
groups
to
recommend
it,
but
is
that
the
cka
is
what
an
administrator
needs
to
go.
No,
not
an
application
developer.
The
person
whose
job
it
is
to
take
care
of
the
cluster
and
the
CK
ad
is
what
an
application
developer
needs
to
know.
Who
doesn't
actually
isn't
responsible
for
the
cluster
now
and
a
lot
of
companies.
It
would
be
the
same
person
or
startups.
It
would
be.
I
B
I
think
an
administrator
needs
to
know
how
to
do
deployments
just
to
be
able
to
answer
questions
for
mother.
So
just
but
I
don't
know
that
they're
there
necessarily
keep
in
mind
sort
of
best
practices
on
application,
design,
I
guess
my
I'm.
Let
me
choose
my
answer:
I'm
really
looking
to
the
expert
groups
for
both
projects,
which
are
designed
to
run
in
parallel
to
make
their
recommendations
on
what
the
curriculum
should
be.
It
seems
to
me,
like
there's,
naturally
going
to
be
some
overlap.
B
H
B
Mean
it's:
it's
literally
an
in-person
process
over
two
days
that
it
gets
done
on
Excel,
spreadsheets
and
Google
sheets
and
stuff
I.
Guess:
there's
an
ongoing
working
group
that
then
refines
it
and
such,
but
we're
I
mean
again
somewhat
bluntly,
were:
were
not
opening
it
up
for
everyone
to
evaluate
it,
partly
because
we
don't
want
to
have
the
backlogs
of
it.
Be
you
know
the
list
of
answers.
Everything
needs
to
go
do
so
we
actually
require
all
the
participants
to
sign
something
saying:
no,
they
won't
share
with
their
colleagues.
B
B
We
literally
have
13
different
people
from
13
different
companies
that
are
engaged
in
this,
that
are
all
active
in
the
kubernetes
community,
and
my
hope
is
that,
together
that
they
would
be
able
to
agree
on
essentially
the
what
the
common
ground
is.
Not
necessarily
the
idea
or
anything,
and
then
I
mean
my
other
pitch,
for
it
is
that
we
will
iterate
it
over
time
and
it
hopefully
improves
rapid
worse
for.
A
Those
13
folks
who
signed
up
or
for
anybody
else
who
gets
involved
is
there
a
way
to
look
at
who
actually
does
the
operation
of
applications
in
kubernetes
and
I
say
that?
Because,
if
I'm
being
very
honest
here
there,
a
number
of
folks
who
work
on
building
kubernetes
and
are
interested
in
operating
kubernetes
itself,
whose
tend
to
see
viewing
how
you
operate
applications
a
little
bit
different
than
a
lot
of
the
things
that
we
see
in
practice
and
so
I'm
curious.
How
much
of
it
is
actually
people
who,
if
there's
a
way
to
say.
B
B
I
will
say
that
I'm
familiar
with
most
of
the
people
on
the
list
and
the
vast
majority
of
other
companies
that
are
engaged
with
real
enterprises
today
and
so
I
think
it's
totally
valid
to
say
yeah,
but
they're
teaching
them
the
wrong
thing
or
you
know
their
their
proposed
solutions
are
not
ideal
or
something
but
again,
the
goal
is
that
the
the
wisdom
of
crowds
here
by
bringing
a
group
of
them
together
that
they
could
agree
on
the
common
subset?
Okay,
I.
B
Mean
all
of
our
time
is
incredibly
valuable,
so
it's
a
little
hard
for
me
to
imagine.
People
signing
up
just
to
propose
something
really
impractical,
but
I
do
want
to
emphasize
that
I
see
this
area
as
even
more
up
in
the
air
than
to
seek
a
a
curriculum,
so
I
think
I
understand
some
of
the
sensitivities.
A
Part
of
it
is
the
rapidly
changing
space.
Part
of
it
is
just
opinions
on
how
you
do
things,
and
we
are
a
community
with
a
lot
of
varying
strong
opinions,
and
how
do
we
make
sure
that
one
strong
opinions
does
not
dominate
something
over
somebody
else's?
That
is
that's
a
definite
concern
for
me
right
now,
given
the
wide
variety
of
opinions
in
the
application
space,
nobody.
B
A
See
what
I
can
show
up
for
I
don't
know
yet?
Okay,
great,
but
don't
worry,
I
won't.