►
From YouTube: Kubernetes SIG Contributor Experience 20170517
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
Good
morning,
everybody
you
share
my
screen
here.
A
B
So
what
I'm
going
to
tell
about
is
a
local
cluster
tool.
I
developed
this
proposal
in
kubernetes
community
repository
for
a
new
version
of
search
of
this
tool,
I'm
going
to
rewrite
it
go
actually
as
of
now
it's
mostly
comprised
of
shell
scripts.
So,
basically,
what's
the
problem,
it
is
salts
so,
for
instance,
this
two
most
commonly
used
local
cluster
tools,
of
course,
is
much
more,
but
I'll
just
name
this
two,
for
example,
namely
local
plaster
and
mini
cube,
but
there
are
numerous
problems
with
it.
B
For
example,
if
you
want
to
build
the
local
binaries
for
mini
cube,
it's
actually
another
complicated
process,
I
mean
if
you
want
to
make
a
change
in
copernicus
itself
and
see
it
work.
It's
rather
hard
to
do.
Another
problem
is
that
both
local
cluster
and
mini
cube
are
single
node
processors,
which
limits
possibilities
of
both
applications
and
say
end-to-end
tests
that
you
can
run
here
so
I'll
just
start
demonstrating
the
tool
and
well
takes
some
time
to
run
so
I'll
have
time
to
explain
other
possibilities.
B
B
B
So
well,
it
starts
with
our
other
products
that
soaps,
so,
for
instance,
irritable
netis
is
a
rather
big,
so
it
takes
quite
some
time
to
build
and
if
you
have
not
so
powerful
laptop
or
it
may
take
age
to
build
and
start
the
cluster,
and
if
you
use
remote
dr.
engine,
you
can,
for
instance,
use
it
to
start
to
build
your
kubernetes
source,
but
not,
but
not
run
your
test
busters.
B
Another
problem
so
suggestion
local
pastor
implementations
is
support
for
in
different
c9
limitations.
So
you
may
want
to
have
some
kind
of
realistic
cluster
with
a
flannel
or
calico,
and
this
can
be
actually
a
problem
with
a
nd.
You
can
choose
Sienna
implementation.
I
will
not
do
this
during
this
demo,
because
the
time
limits,
but
in
general
it's
possible.
B
So,
oh
sorry,
here,
environment,
finding
repeating
myself
sources,
I
mean,
if
you
say,
run
intestine
TARDIS
and
your
project
depends
on
people
Metis
and
if
you
use
local
cluster,
you
will
have
to
build
kinetics
from
source,
and
here
you
can
use
pre-built
images
and
mini
cuba.
Well,
it's
hard
to
use
GML
side
vm.
So
here
we
have
the
cluster
up
and
running.
B
You
can
see
it
well,
it
took
some
time
to
start,
but
if
we
will
restart
it
now
it
will
be
much
faster.
So
basically,
during
your
workflow,
you
start
faster
runs
and
then
it's
can
be
restarted,
much
quicker
because
it
takes
snapshots
of
the
cluster.
So
in
about
cycle
model
half
a
minute,
you
will
have
cluster
dragon
again
and
here
I
will
demonstrate
another
interesting
trait
I
mean
a
possibility
of
running
adventists.
B
Here,
I
have
slightly
broken
kubernetes
kettle
cone
from
burn
a
Toastmaster.
You
see,
I
have
introduced
a
break
and
change,
and
now
I
will
try
to
run
end
to
end
test
here.
For
the
sake
of
brevity,
I
already
started,
the
cluster
and
I
will
just
show
how
to
respect
it.
It's
done
with
the
same
comment
as
initial
setup.
B
B
B
B
Unfortunately,
there
are
rough
edges
and
some
tests
flip
some
times,
but
basically
here
you
see,
this
exercise
is
different.
Keep
with
next
versions
and
building
stable
and
master
key
Burnett
is
from
source
or
master.
Of
course,
I
have
to
fix
it
from
time
to
time.
Another
thing
where
you
can
use
this
project
easier.
For
instance,
you
are
running
writing
some
Cuban
artists
extension.
B
B
B
B
So
you
can
see,
for
instance,
start
a
GC
instances,
helper
script
for
it
and
use
it
for
building
a
cubed,
Metis
and
so
on,
in
this
case,
I'm
running
at
the
base,
local
Booker,
because
it's
slightly
faster
this
way
but
well,
it's
also
works
just
fine
on
med
docker,
but
my
Mac
is
not
very
powerful,
so
it
would
take
it
and
longer
if
I
was
running
it
here
so
concerned
in
this
community
proposal
Oh.
What
else?
B
What
I'm
going
to
do
is
besides,
go
rewrite
I'm,
going
to
make
some
clean
configuration
for
the
plaster-
and
this
is
not
mentioned
in
proposal
yet,
but
also
there's
possibility
of
introducing
multiple
backends
later
so
docker
and
ochre
may
not
be
the
other
possibilities.
For
instance,
it's
possible
to
add
the
local
class
local,
a
plaster
like
single
mode
operation,
operation
mode.
B
B
Now
or
the
well
folks,
now
it's
just
taking
some
time
to
build
a
hypercube,
and
so
but
after
it's
built,
the
cluster
restart
is
also
much
faster
because
the
scripts
are
made.
This
way
that
just
the
binary
sorry
placed
and
the
snapshot
and
snapshot
is
reused,
so
most
of
the
time
during
a
pasta,
restart
in
this
case,
the
most
time
has
taken
actually
by
kubernetes,
builded,
separate
contributor
experience,
issue,
I
think
can
be
done
or
now
it's
placing
binary,
and
now
it's
using
this
binaries
to
start
the
cluster.
B
Actually,
so
here
starts
it's
actually
three
node
cluster,
in
this
case
one
master
and
to
work
with
nodes
and
actually,
if
I
I
could
specify,
for
instance,
six
nodes,
and
it
will
not
take
much
more
time,
but
with
six
nodes,
for
instance,
you
can
run
all
the
conformance
tests,
both
serial
and
serial
and
under
ten
minutes,
and
they
all
pass
so
now
I
will
be
stacked,
the
deck
master.
Sorry
let
it's
taken
some
time,
but
here
let's
restart
this
implant
test
and
see
if
it
passes
now.
B
B
B
B
Here
it
is
so
it
describes
a
bit
different.
So
now,
let's
see
see
here
you
we
have
success.
So
basically
we
updated
kubernetes
and
rebuilt
it,
and
now
all
the
tests
pass
and
it's
possible
its
lookers
wide,
very
much
wider
use
than
say
local
up
cluster
and
also
it
can
do
some
cases
not
handled
by
mini
cube.
D
F
A
D
So
it
took
him
more
time
than
it
should
have
to
figure
out
what
the
problem
was
and
because
he
was
checking
his
build.
Configuration
and
I
think
he
had
some
ideas
for
improving
the
visibility
of
this.
Maybe
an
email
to
kubernetes
dev
isn't
sufficient.
Maybe
we
need
to
do
slack
or
have
the
sig
leads
kind
of
propagated
at
community
meetings
whatever
it
is,
is
if
you,
if
you
make
a
change
that
is
no
longer
backwards
compatible
to
head
meaning
all
developers
will
need
to
change
the
way
that
they
build.
D
G
Good
morning
everybody
I'm
gonna
paste
the
document
into
chat.
Can
you
guys
hear
me?
Okay,
yeah,
okay,
so
I
work
with
these
two
Googlers
Lee
and
Rupert.
They
are
sres
or
not
exactly
like
familiar
with
how
to
work
with
kubernetes
and
I
sat
down
with
them
for
about
an
hour
hour
and
a
half
and
they
basically
brain
dumped
and
I
wrote
it
all
down.
So
I
posted
this
little
list.
G
I
think
last
week
at
some
some
point
and
for
a
lot
of
these
I
think
they're
known
issues
like
as
far
as
things
that
keep
being
brought
up
by
people
but
I
figured
it
wouldn't
hurt.
Did
anybody
see
anything
here
that
looked
out
of
place
or
is
a
surprise
compared
to
what
you've
heard
and
then
I
was
thinking?
If
someone
wants
to
make
actual
little
items,
maybe
we
can
do
some
things
about
things
like
contreras
been
one
of
those
we
should
get
rid
of
it
for
a
long
time
kind
of
things,
yeah.
H
I
think
the
choices
we
need
to
make
here
are:
what
are?
How
are
we
prioritizing
the
pain
point?
To
else,
are
we
asking
for
on
feedback
because
they
know
there's
also
some
IBM
folks
that
are
trying
to
help
with
on
you
new
contributor
onboarding
and
then
what
are
we?
So
what
are
we
prioritizing?
What
are
we
going
to
do
and
then
start
making
proposals
out
of
it,
so
feedback
on
all
of
this
would
be
awesome,
sir.
G
Do
yeah
they
mail
me
today
about
like
how
do
we
triage
a
bug
report?
Is
there
like
a
page?
It
tells
us
how
to
triage
bug
reports
and
the
best
I
was
able
to
do
is
find
some
SIG's
have
like
specific
steps
on
how
to
do
it
and
some
don't
so
one
of
them
was
like.
Is
this
something
we
should
all
work
on
together
and
I
was
like
that
one
good
question
I'm
going
to
bring
up
today?
G
D
Hey
Chris,
that's
awesome,
I,
don't
can
we
start
with
the
proposal
event
on
how
to
triage
your
bug
report
and
what
we
expect
from
like
individual
contributors
and
what
we
expect
from
SIG's,
because
I
think
a
number
of
people
senior
people
very
siblings,
that
I've
talked
to
all
have
different
ideas
on
what
it
means
to
triage
a
bug,
and
so
maybe
we
we
need
to
discuss
what
this
means.
I
know
there's
been
a
group
of
us
at
Google
that
have
looked
at
the
five
dozen
issue.
They're
like
we
need
to
do
something
about
I've.
D
You
know
5,000
plus
open
issues.
Let's
try
doing
it
manually
and
see
what
if
we
can
pick
out
patterns
or
find
some
because
a
strategy
that
we
can
turn
into
a
longer
process
but
I
think
step.
One
is
determining
what
it
means
to
be
triage
and
in
our
reading
the
options
where
every
issue
should
have
a
sig
label,
meaning
if
some
sig
should
own
it
and
from
there.
We
need
to
have
expectations
that
say:
go
through
two,
because
just
adding
a
sig
label
to
5,000
issues
isn't
the
right
answer
either
from
there
I
think.
D
There's
a
github
team
called
for
each
sig
sig
bug
right,
like
sig
network
bugs
sig
note
bugs,
and
we
can
leave
it
up
to
the
six
to
come
up
with
some
kind
of
rotation
or
we
could
help
them
draft
a
process
for
that
as
well.
But
that
and
I
think
the
kind
of
issue
is
also
important.
So
there's
also
kind
labels
in
the
repo
and
we
were
trying
to
add
like
kind
bug
and
kind.
You
know,
feature
requests
or
kind
enhancement.
D
There
are
various
one
of
those
right
now
there
is
automation,
a
very
poorly
publicized
automation
that
allows
people
to
actually
annotated
label.
Can
we've
talked
about
it
in
the
contrive
X,
but
if,
if
you
go
to
kubernetes,
kubernetes
and
you're
on
an
issue,
you
type
slash
sig
node,
it
should
add
the
signal
label.
D
So
we
need
to.
We
need
to
work
on
publicizing
both
the
mechanism
and
the
policy
and
one
defining
the
policy.
Is
anyone
interested
in
doing
this
defining
the
policy
for
triaging
what
it
means
for
bugs
to
be
triaged,
at
least
taking
a
first
pass,
and
we
can
review
it
at
the
decumanus
at
the
next
big
meeting
and
we
can
get
some
feedback
from
chris
and
the
other
folks
at
IBM
that
are
working
on
it.
Yeah.
G
F
Network
go
through
mark
things
that
have
clearly
been
fixed
as
fixed
and
closing
them,
or
else
figuring
out
if
they
are
going
to
be
available
and
or
if
they
need
to
be
fixed
or
if
it's
still
relevant
and
then
commenting
on
their
after
giving
them
out
of
time,
and
they
will
feedback
just
close
the
issue
I
think
maybe
in
general,
a
greater
knowledge
here
needs
to
happen
between
what
should
be
contained
in
the
issues.
You
know,
if
there's
a
feature
for
your
cig,
that
you
hope
eventually
will
come
to
fruition.
F
Does
it
still
make
sense
to
have
an
actual
physical
issue
open
for
it,
or
is
it
something
you
want
to
put
in
like
a
like
an
epic
or
roadmaps
like
this
thing?
Is
it
nice
to
have,
as
I
can
slash
these
issues
down
to
a
more
feasible
number
for
the
sake,
but
they're
just
going
to
stack
back
up
with
the
number
of
issues
that
get
opened
within
the
automated
fashion
and
duplicates
for
people
that
are
still
looking
for
it
be
sure
that
it's
like
I
think.
H
This
is
one
of
the
reasons
we
wanted
to
find
what
it
means
to
triage,
try
to
make
it
in
some
way
helpful
and
useful
for
people
who
might
be
coming
in
as
new
contributors
to
be
able
to
contribute.
Now
they
won't
always
have
contacts,
but
maybe
there's
a
first
level
triage
second
level,
triage
kind
of
thing
so
that
if
the
Nutri
contributor
wants
to
come
in
and
try
and
help
us
bugs
they
can,
they
can
get.
D
D
And
yeah,
so
defining
a
process
is
going
to
be
key
and
then
a
lot
of
them
not
being
useful,
would
kind
of
go
towards
I.
Don't
know
if
the
templates
going
to
be
the
answer,
but
right
now
it
certainly
isn't
the
answer.
There
is
an
issue
template
like
if
you
go
to
kubernetes
communities
and
say
create
a
new
issue.
There's
an
issue
template
that
everybody
just
deletes
because
it's
long
and
it's
a
pain
to
Bri
through,
but
maybe
if
they
were
just
like
three
questions
or
you
know
the
minimum
viable
MPT
right.
G
E
G
D
G
D
We
should
worry
about
the
big
pile
I
think.
Maybe
we
can
have
a
suggested
process
for
the
cigs.
It's
going
to
be
hard
to
standardize,
I
mean
we
can
have
requirements
for
all
the
SIG's
like.
We
need
them
to
be
done,
but
I
think
if
we
tell
them
how
to
be
done,
it'll
be
challenging
I.
Don't
what
do
you
think
Jay
go
I.
E
Think
what
we're
getting
at
is
that
we
need
some
sort
of
higher
level
issue
lifecycle
and
like
triage
stage.
One
is
one
part
and
then
think
triage
does
another
and
I
think
what
we
were.
What
we're
getting
close
to
is
some
rules
where
the
automation
can
take
care
of
bail
issues
that
don't
see
action
for
some
amount
of
time
and
get
kicked
back
to
the
filer
for
more
information.
And
if
no
information
comes
in
three
days,
then
it
gets
closed.
E
That
kind
of
thing,
but
you
know
the
higher
issue
that
I
see,
is
that
there's
a
lot
of
communication
on
many
of
these
issues,
but
not
a
lot
of
yes.
I
have
ownership
of
this
and
I
will
close
by
next
Tuesday
kind
of
language
and
so
I
think
if
we
can
help
out
with
a
general
process
and
issue
lifecycle
that
will
go
a
long
way.
H
And
then,
to
my
request
for
ways
to
get
people
on
boarding.
If
that's
rough
level,
first
level
triage
could
in
any
way
have
easy
starting
points
for
some
of
the
rinse
and
repeat,
rinse
and
repeat
grunt
work
that
needs
to
happen.
If
it's
like
go
search
for
something
related
and
say,
I
think
this
is
related
to
this
bug,
without
necessarily
clearing.
H
H
F
A
D
Yeah
I'm
trying
to
look
at
see
where,
where
that's
at
right
now,
Lucas
is
actually
working
on
it
and
I
believe
he's
in
the
European
time
zone.
So
I,
don't
think
he's
here,
oh,
but
the
idea,
just
to
reiterate,
is
that
all
repos
within
the
kubernetes
org
and
I
need
to
change
my
definition
of
all
now.
I
posted
a
proposal
to
a
career
instead
of
mailing.
D
Lucas
has
done
a
demo
on
his
own
github
account,
and
so
every
I
think
every
repo
in
his
public,
like
org,
has
all
the
kubernetes
labels,
which
is
kind
of
cool,
but
it's
defining
which
labels
we're
going
to
use
and
I
think
we're
going
to
start
with
a
small
group
like
the
cig
labels
and
maybe
the
kind
labels
and
possibly
LG
GM,
so
we're
going
to
skip
out
on
I
think
they're,
like
60
area
labels.
That
may
not
be
useful,
for
you
know
things
like
Helmer,
some
of
the
other
other
repos.
A
G
Sorry
I
got
one
last
second,
the
few
guy
Jamie
I
was
kind
of
like
walking
through
with
him
Joe
and
I
decided
that
we
would
grab
a
person
and
walking
through
something
he
did
this
cool
thing
where
he
was
trying
to
update
a
SIG's
information
like
in
the
metadata
for
the
docs
for
the
like
a
list
of
six
page,
and
then
we
had
this
crazy
idea.
We
should
just
make
it
all
Auto
generate.
G
So
all
cigs
are
consistent
in
things
like
that,
and
we
could
just
have
one
yamo
file
with
all
the
sig
metadata,
and
then
we
could
generate
all
the
sig
individual
pages
and
the
big
list,
oh
SIG's,
in
a
way
that
people
can
only
update
in
one
spot
and
then
it
would
just
Auto
generate
Auto,
generate
everything,
so
I
got
merged
and
he
announced
it
a
few
times
like
in
dev
and
everything.
But
I.
Don't
think
people
were
paying
attention
or
responded.
So
I,
just
kind
of
wanted
to
like
remind
everybody
that
we
have
this.
Oh.
A
G
Alright,
so
you
know
when
you
go
to
us
in
the
community
page
to
go
to
a
cig
and
it
has
like
a
readme
and
some
like
they're,
not
the
same
format,
so
this
guy
initially
started
with
like
hey.
We
should
just
make
a
template,
that's
the
same
for
all
the
SIG's,
so
they
all
have
the
same
information
and
that
can
I
transformed
into
let's
make
a
yamaha
file.
G
G
G
G
D
H
H
Is
a
sharing
point
for
cig
leads
and
I
think
putting
it
in,
as
you
noticed
the
community
meeting
and
even
sharing
the
go
team,
we
had
a
new
contributor
who
is
super
excited
about
it
kind
of
thing
to
cover
duties,
dev,
it's
a
good
idea,
because
I
think
we
need
to
start
more
of
those
rallying
points
and
more
about
excitement
points
even
if
it's
not
super
chattery.
Ok,
so.
G
H
C
C
G
Can
you
put
that
in
the
feedback
in
the
pull
requests
so
I
can
relay
it,
get
it
true
yeah
that
would
be
very
all
right,
cool
all
right.
That's
all
I
had
sorry,
no.
F
Had
a
note
on
feature
stability-
and
this
is
probably
like
some
that
will
cross
across
I'm
too
sick
p.m.
but
do
we
have
like
to
find
guidance
on
what
it
means
for
feature
to
be
alpha
and
or
beta
and
or
GA,
as
in
like
the
strict
requirements
and
how
one
progresses?
And
what
can
you
do
and
can't
you
do
one
matriculating
between.
C
Well,
it's
also
the
question
for
6p
and
and
if
you're
going
to
like
when
we
discuss
the
prob'ly,
the
next
6:00
p.m.
meeting
or
you
may
type
your
question
into
this
sick
p.m.
island
estimate
when
I
discuss
occur.
If
someone
has
any
ideas
ready
already,
I
would
share
them,
but
I
would
propose
to
switch.
It
started
to
Signia
sorry.
D
Chris
I
just
searched
I,
think
I,
remember,
reading
it
in
the
kubernetes
Doc's
a
while
back
at
least
for
the
API
it
that's
probably
what
I've
seen
before
on
the
alpha
level.
It
is
something
that
sits
behind
a
flag,
so
it's
disabled
by
default
and
then
something
you
have
to
actively
enable
for
your
cluster
I.
Think
that's
one
of
the
things
and
it's
not
supported
at
all
support
for
it
be
dropped.
It
may
like
go
away
as
a
feature.
D
It's
not
something!
That's
like
production-ready
I
think
these
are
all
like
general
guidelines.
I,
don't
know
if
we
have
a
stricter
definition
on
that
and
then
beta
implies
it
a
little
bit
better
tested,
I
think
there's
some
backward
compatibility.
The
overall
features
proper,
like
not
going
to
be
dropped,
I'm
just
reading
from
the
page
at
this
point
in
going
from
memory,
but
beta
is
still
intended
for
like
giving
feedback,
whereas
it's
a
stable
level,
it's
just
like
production
and
ready
I,
don't
know.
Do
you
see
the
link
that
I
said
did
well.
F
Yeah
this
thing:
okay,
yeah
I'll,
take
a
look
at
this
I
bring
it
up
here
only
because
I
think
it
is
a
another
potential
trigger
problem
whereby
maybe
you
get
so
far
with
something
and
beta
always
a
note
that
it
things
that
have
to
be
reversed
right
before
it's
about
two
kids,
GA
I've
seen
this
in
a
couple
of
six
now,
where
maybe
there
wasn't
popular-
and
this
probably
lends
to
the
discussion
about
how
to
how
do
we
announce
backwards,
incompatible
changes,
but
sometimes
something
will
make
it
into
the
release
and
be
right.
There.
F
G
A
Anybody
doesn't
have
anything
else,
go
wrap
up
to
the
day,
okay,
well!
Well,
we
can
injure
and
week
after
next.
Okay,
thanks.