►
From YouTube: Kubernetes Community Meeting 20170713
Description
We have PUBLIC and RECORDED weekly video meetings every Thursday at 10am US Pacific Time.
In this meeting we get a demo of Spinnaker + Azure ACS Engine for CI/CD, 1.8 Feature planning updates, SIG Cloud Provider proposal, SIG Architecure update, SIG Release release process improvements, and a SIG Testing update.
https://docs.google.com/document/d/1VQDIAB0OqiSjIHI8AWMvSdceWhnz56jNpZrLs6o7NJY
A
All
right
are
we
recording
all
right.
Hi
everybody
welcome
to
the
July
13th
kubernetes
community
meeting.
My
name
is
Ken
Simon
I'm
going
to
be
emceeing.
The
meeting
today
I
reminder
to
everybody.
This
is
a
publicly
recorded
meeting
and
the
streaming
and
available
on
YouTube.
So
it's
10
o'clock
exactly
so.
First
up
we
have
Arun
Chandra
Sekhar
from
Microsoft
to
talk
about
spinnaker
plus
as
your
ACS
engine
for
CI
CD
room.
If
you
want
to
share
your
screen,
yeah.
B
Let
me
know
if
you
can
see
the
screen.
I
can
see
it
just
fine
great
hi,
everyone,
I'm
John,
shaker
and
the
principal
program
manager
for
as
your
open
source
integrations
that
include
Netflix,
spinnaker,
Hoshi,
coughs,
sweet
as
well
as
suite
of
other
things
like
Jenkins
and
so
on.
Today,
I'm
gonna
primarily
be
focusing
on
how
to
do
sort
of
continuous
deployment
to
Azure.
B
So
the
idea
is
that,
irrespective
of
whether
you
have
a
containerized
app
or
like
a
regular
stateful
legacy
app
on
a
virtual
machine
with
with
a
lot
of
data,
you
can
still
do
immutable
by
baking.
Your
images
either
using
Packer
for
your
machine
images
or
using
some
kind
of
private
registry,
such
as
as
your
container,
is
you
for
your
docker
container
images
and
spinnaker
basically
provides
a
nice
cluster
management
and
pipeline
view
for
you
to
go
ahead
and
push
these
into
auto-scaling
sets
of
nodes,
and
these
nodes
could
be
either
auto
scaling.
B
B
Secondly,
it's
a
pretty
good
OSS
community,
which
sort
of
comes
within
the
confluence
of
netflix
jenkins
and
communities,
which
is
very
interesting
to
us
and
as
I
mentioned
immutable,
is
what
the
skew
is
and
which
is
very
containers.
Microsomes
and
cloud
native
friendly.
Very
few.
Fully
open
source
pipelines
can
claim
that
that
exclusively
focus
on
the
sweet
spot,
and
then
it
is
also
building
out
a
whole
bunch
of
new
sophisticated
tooling
such
as
things
which,
which
enable
resilience
over
robustness,
which
includes
things
like
chaos,
monkey
2.0,
which
is
completely
built
on
spinnaker
as
a
platform.
B
And
then
you
also
have
new
things
that
what
Google
and
we
are
investigating
around
things
like
testing
production,
using
cannery
type
deployments,
as
well
as
strengthening
the
sort
of
expressing
of
pipelines
completely
as
coal.
So
this
makes
it
really
interesting
and
plus
it's
pretty
hard
and
because
Netflix
is
using
it
quite
heavily
and
being
able
to
achieve
large
number
of
deployments
every
day.
So
this
gives
us
some
amount
of
confidence,
because
I
just
go
and
go
quickly
into
the
demo.
Now
to
show
you
a
little
bit
of
what
this
will
feel
like.
B
B
And
as
that's
flowing
through,
what
does
basically,
what
we
are
provided
is
a
QuickStart
template
that
can
let
you
pretty
much
stitch
together
the
tool
chain
that
I
described
earlier,
which
lets
you
go
all
the
way
from
github
to
Jenkins
to
spinnaker,
so
the
disk.
It
would
basically
trigger
a
new
sort
of
pipeline
for
doing
a
basic
docker
build
in
Jenkins,
and
what
this
does
is.
B
So
what
does
this
go
at
this
pipeline
is
going
to
do
is
deploy
to
the
day
of
sort
of
a
cluster
that
I've
defined
here,
which
I'll
talk
about
in
a
second
and
once
it's
done
deploying
it
will
go
ahead
and
do
like
it's
just
stopping
at
a
manual
judgment,
so
I
can
verify
that
things
are
good,
so
this
is
just
for
demo.
You
normally
want
this
just
flow
through
in
a
continuous
deparment
flow.
So
here
it
basically
tells
you
to
run
coupe
CD
of
ceramics
for
the
client.
B
Everything
here
is
using
SSH
I'll
just
go
ahead
and
run
proxy
for
my
local
dev
and
once
I
have
that
I
should
start
seeing
the
change
that
I
made
show
up
here
so
basically
I'm
having
the
sort
of
the
hello
kubernetes
loaded
here
once
I
know
that
everything
is
good.
I
can
go
to
spinnaker
and
go
ahead
and
tell
it
to
continue
at
which
point
it'll
go
and
try
to
deploy,
to
start
and
start
deploying
to
production.
B
What's
very
interesting
about
spinnaker
is
because
it's
completely
immutable
based
you're,
not
really
doing
multiple
publishes
or
multiple
retrievals.
What
it's
gonna
do
is
it's
gonna,
take
the
image
that
it
found
in
the
dev
cluster
and
then
directly
deploy
it
with
fraud,
because
you
want
complete
consistency
across
your
or
your
environments,
which
is
what
immutable
enables,
and
this
gives
you
a
nice
way
to
ensure
that
the
exact
image
that
you
had
in
dev
is
not
moving
into
prod
and
what's
interesting.
Here
is
what's
very
interesting
about
the
spinnaker
view.
B
Is
this
notion
of
clusters,
which
kind
of
tries
to
abstract
away
sort
of
the
cloud
specific
logic?
So
this
is
where
we
work
together
with
all
the
other
cloud
providers.
So
then,
now
you
have
a
single
pane
where
you
can
do
both
deployments
with
kubernetes.
You
could
deploy
two
VM
virtual
machine
scale
search.
You
could
deploy
to
anything
that
you
want,
and
here
we
are
abstracted
away
sort
of
the
load
balancers.
If
you
look
at
the
sort
of
the
dev
load
balancer
here.
This
is
just
sort
of
a
simple
cluster
within
private
IP.
A
C
A
D
A
E
You
so
if
you--if
you
look
in
the
meeting
notes,
I've
dropped
in
a
proposal
about
a
working
group
to
kind
of
consolidate
the
efforts
of
the
the
crossing
efforts
of
the
mystic
cloud
providers.
Essentially-
and
let
me
see-
I'll
drop
a
link
to
this
in
the
in
the
chat
so
that
everyone
can
see
it
and
maybe
I'll
share
the
screen.
Here
too,
let
me
see
all.
E
But
right
now,
one
of
the
couple
of
the
challenges
that
we
face
are
there's
a
bit
of
inconsistency
across
how
the
cloud
providers
operate
and
how
they
offer
a
certain
behavior
and
and
there's
also
been
some
discussion
to
move
some
of
the
provider
code
out
of
the
out
of
upstream
kubernetes.
For
example,
the
the
OpenStack
community
has
has
has
a
little
bit
of
difficulty
in
maintaining
the
upstream
provider
and
is
looking
at
moving
that
towards
you
know
outside
of
the
community,
and
you
know
kind
of
getting
a
greater
volume
of
experts
maintaining
that
provider.
E
So
what
this
proposal
is
is
either
the
formation
of
a
new
working
group
or
a
new
sig
that
will
that
will
kind
of
evaluate
these
crusts.
These
these
cross
stick
efforts,
and
so
looking
so
looking
at.
How
do
we
define
particular
behaviors
like
if
they
like
I've
been
known,
is
no
longer
available?
Should
that
be
deleted,
or
should
that
be
parked
or
if
there
can't?
If
we
can't
reach
a
consensus
about
that
shared
behavior,
then
how
can
we.
E
E
E
I
mean
I
mean,
certainly
you
know
one
of
the
reasons
why
we're
you
know
we're
bringing
this
up
as
a
proposal
and,
in
my
mind,
there's
a
question
mark
as
to
whether
it
should
be
a
cig
or
a
working
group
is,
you
know,
is
in
some
ways
to
answer
these
questions.
You
know
you
know
personally,
I
I
don't
have
a
strong
preference
either
way
aside
from
I
want
to
you
know
if
this
is
established,
it's
something
that
I.
E
Would
you
know
that
I'm
there
that
I
would
like
there
to
be
consensus
on
on
how
its
run
and
buy-in
from
the
other
from
the
other
SIG's?
And
so,
if
it
turns
into
a
sig,
that's
fine.
You
know
if
it's
a
working
group
I'm
fine
with
that.
Also
you
know.
That's
that's
one
of
the
reasons
for
for
bringing
this
forward.
So
I
don't
have
strong
opinions
either
way
and.
G
G
We
need
to
make
sure
that
we
have
the
right
interfaces
and
all
those
other
parts
of
the
system
and
tooling,
to
ensure
that
cloud
providers
can
plug
in
so
in
a
working
group
to
bring
together
the
necessary
members
of
all
the
cities
and
all
the
different
stakeholders
for
the
inaudible
cloud
providers
makes
sense,
but
in
terms
of
long-term
ownership,
like
the
service
controller
should
be
sitting
network
or
you
know
that
whoever
we
decide
owns.
The
no
controller
should
own
that
part.
No.
G
H
Yeah
I
have
the
implications
of
this
on
the
build
release
and
distribute
is
the
part
of
it.
I
keep
coming
back
to
binary,
doesn't
work
across
providers
and
I
think
there
needs
to
be
a
place
to
discuss
the
determine
proposed
solutions
for
that
I
think
the
working
group
seems
like
a
good
starting
point.
It
definitely
needs
a
place
that
have
that
conversation,
James.
J
J
But
what
will
each
move
east
and
we
should
have
a
clear
goal
and
we
want
to
buy
some
time
and
have
the
proposal
for
certain
problem.
Then
we
connect
the
include
all
those
we
involve
of
the
sequel.
Once
those
have
to
go
is
needed.
We
should
dismiss
those
were
cool
if
we
idiot
and
then
we
can
start
news
what
problem
it
is
about.
Yes,
tonight
songs,
what
group
is
connected?
J
Still
they
didn't
make
the
decision,
they
didn't
really
participate
with
the
code
and
feature,
but
if
they
will
now
be
be
and
end
up
track
that
all
those
cannot
relate
to
the
feature
goes
to
the
work
pool
and
even
a
not
go
back
to
the
reality
of
the
sick.
Look
to
drive
those
finished
so
in
that
we
have
an
actor,
always
some
proposal,
but
no
action.
J
E
So
listening
to
these
comments,
I
think
I
I,
guess
them.
Maybe
what
what
am
I
proposed
that
we
do
is
I
can
go
back
and
I
can
revise
this
proposal.
That
include
a
little
bit
more
information
about
specific
goals.
We
have
a
couple
of
specific
goals
right
there
and
I.
Think
the
focusing
on
coming
up
with
the
definition
of
common
behaviors
and
a
you
know:
a
proposed
milestone
for
migrating
providers
out
of
tree
and
into
their
into
sig
sig
managed
projects.
E
That
would
be
my
that
would
be
my
proposal.
You
know.
I
also
just
want
to
make
sure
too
that
this
is
something
that
there
is.
There
is
buy-in
from
from
from
the
sig
providers,
because
I
wouldn't
want
to
run
ahead
with
the
proposal
and
then
have
half
of
the
providers
not
being
involved
with
it,
and
so
you
know
so
the
the
AWS
and
Azure-
and
you
know
the
OpenStack
and
other
you
know
providers.
You
know
actively
being
involved
in
this
and
working
towards
the
common
goal
of
these
of
these
procedure.
Artifacts.
A
F
K
K
I
like
when
it
comes
to
assigning
ownership
of
this
or
being
able
to
triage
an
issue
off
and
say:
oh
that's,
clearly
something
that
needs
to
me
to
decide
the
time
working
group.
What's
the
team,
I'm
gonna
think
what's
the
label
I'm
going
to
find
the
issue,
so
the
box
stops
nagging
me
and
how
do
I
know
something's
playing
attention
to
this
I.
F
Would
expect
that
one
of
the
outputs
of
this
working
group,
if
it
stays
a
working
group,
would
be
clear
lines
of
ownership
and
essentially
so
that
you
know
so
that
this
say
this
working
group
doesn't
own
the
issue.
It's
very
obvious
that
hey!
This
is
part
of
the
league.
You
know
cloud
provider,
storage
interface,
so
they.
K
Know
I
have
I,
have
a
I
file,
a
bug
that
says
a
storage
interface
that
names
a
particular
way
on,
let's
say
a
sure,
but
doesn't
behave
that
same
way
in
other
cloud
providers.
I
need
to
escalate
to
somebody
to
make
a
decision.
I
would
like
to
trio
actually
said
she
was
something
that
should
be
decided
as
such.
K
Just
trying
to
highlight
this
means
that
triage
and
these
sorts
of
issues
requires
a
lot
of
Sigma
wareness
and
it
makes
it
difficult
to
delegate
or
mechanize
this
role,
providing
a
label
that
says
this
clearly
falls
into
this
working
groups
bucket.
So
I,
don't
know
if
that
means
we're
not
going
to
have
working
directly
was
in
addition
to
sig
labels,
or
if
this
working
group
is
going
to
focus
on
a
particular
area
label
or
whatnot
I,
just
an.
G
Area
cloud
provider,
label
I:
the
premise
of
this
being
a
working
group
is
that
they
shouldn't
I,
don't
see
them
owning
code
right.
It's
a
cross-cutting
thing
across
all
the
different
cloud
provider.
Individual
cloud
providers,
things
as
well
as
state
stores.
Think
networking
is
across
the
life
cycle
without
or
other
touch
points
are
in
the
system.
It.
J
Clear
we
could
draw
more
more
reasonable
sake,
oops,
so,
but
right
now
it
is.
We
just
need
to
try
to
define
those
API.
They
find
those
integration
in
time,
that's
kind
of
things.
Otherwise
it's
the
only
issue
to
them
in
the
power
of
the
kubernetes
coupling
and
adhere
to
a
small
set
up
in
time,
an
incoming
it
still
not
super.
M
Important
point
about
this,
which
is
that,
as
the
cloud
private
provider
extraction
happens,
all
that
cloud
provider
code
is
going
to
be
open
source
repos
in
other
people's
organizations.
So
we're
gonna
have
a
really
interesting
challenge
and
neither
labels
nor
cigs
and
or
none
of
that
actually
covers
that
use
case.
So
we're
gonna
have
to
look
into
that.
I
think.
M
M
Yeah
so
I
I
I
just
was
putting
this
in
there
because
we
want
to
make
sure
that
we
have
really
good
attendance
for
the
next
meeting.
I
didn't
do
the
best
job
of
radiating.
The
meeting
we
had
last
time.
So,
if
you
are
so
inclined,
please
join
the
meeting
a
really
critical
area
of
concern
for
the
project
and
it's
on
July
24th
at
10
a.m.
Pacific,
Standard,
1
p.m.
Eastern,
Standard
and
1700
UTC.
M
M
M
A
J
J
J
J
It
is
there's
no
visibility
after
what
it
is
in
each
piece,
even
for
that
component
to
their
area
and
I
feel
it
is
I
feel
there
is
even
more
serious
things
we
are
trying
to
build
out
to
the
past,
like
that's
a
pretty
more
repository
and
for
kubernetes,
and
so
for
you
don't
know
like
the
for
the
web.
I'm
seemingly
means
it's
like
that.
We
already
call
the
face
and
there's
the
woman,
small
PR
and
the
people
definitely
approve
its
especially
1959
change.
J
That
is
a
great
example
up
great
hipster
English
and
to
the
new
version.
But
what's
in
that
image,
there's
no
your
second
or
there's
no
base
opinion
here
and
you
have
to
ask
the
minis
team
member
and
you
have
to
do
they
ask,
can
I
have
the
other
information
and
all
those
kind
of
things
attacks
make
hard
that
we
just
earned
for
talk
about
the
cloud
provider
separate
the
cloud
provider
separate
repository
and
what,
in
those
things
in
that
one,
we
have
no
clear
signals.
So
another
problem
is
from
the
major
funnels.
J
Is
it's
difficult
to
track
the
progress
of
the
feature
and
tactics
we
have
like
today?
We
have
a
weapon
and
clicks
connector
at
the
issue
of
sticks
are
those
kind
of
opiates.
How
did
you
know-
and
you
have
to
go
to
read
all
the
PR
implementation
detail
and
there's
new
enhancement
to
the
existing
feature
in
their
cities?
We
are
really
nice
in
the
previous
era
and
there's
new
new
enhancement,
a
new
new
issue.
J
Enhancement,
so
it
is
really
hard
to
tell
based
on
all
those
kind
of
an
strategy,
the
signals
and
another
one
is
all
the
puppies
fix
that
I
know
we
closed
the
issue
better
than
I
actually
described
fix.
It's
not
right.
It's
the
new
protection,
let's
reward
and
you
nobody
knows
how
things
and
it
is
going
to
be
a
way.
Example,
you
know
what
I'm
say:
meaning
is
one
picture
and
it's
been
being
fixed.
Anything
is
introduced,
a
new
feature
introduced
in
1.7
and
then
later
they
just
there's
the
new
release.
Notes
it's
like.
J
Oh
I
want
a
public
resource
memory
source
to
another
level
and
the
nature
of
surprise.
Community
I
want
to
thank
again
I
want
to
increase
that
again.
It's
almost
reuse,
those
that
up
together
and
then
this
last
one
it
is
at
least
a
modest
feature
exists.
This
feature
it
is
called
West,
huge
integration.
Finally,
new
stream
and
try
to
learn,
we
hone
them
disable
that
feature
so
our
it
is
have
the
release
notes,
but
in
the
well
released-
and
there
are
no
mystical
or
you
are
coordinators
together.
J
J
Our
document
leader
for
the
mean
is
in
standby,
if
you
want
punish
them,
these
notes
make
that
is
more
user-friendly,
instant,
an
epic
event
mejor
until
after
these
Lots
we
just
don't
have
and
I
know,
there's
the
content
is
just
so
hard
for
him
and
regenerated,
readable
explodes
so
at
end,
which
is
a
well.
We
only
have
a
high
level
of
the
feature
base
of
the
release,
notes
and
the
rest
of
stuff,
which
is
simply
rehab
with
categories
and
disorder,
but
why
we
need
a
community
to
do
that.
J
That's
kind
of
a
risk
for
our
our
factors
of
engineering,
the
amp
and
and
also
a
lot
of
things
we
have
close.
There's
no
minis
notes.
Do
we
eat
believe
or
not.
Those
kind
of
things
is
not
no
impact
for
the
user
and
only
a
glue
believer
if
there's
no
local
grass
and
there's
no
belong
to
any
city
because
Mexico
for
those
some
things
we
even
so
in
the
web
host,
even
even
behind
seemingly
lists
Jim
proposed
something
like
that.
J
Some
new
process
and
you
would
roll
so
one
of
these
things
is
if
every
issue
in
the
one
constable
milestone
we
are
because
if
there
as
Venice
broker
so
in
this
cases,
so
once
you
back,
this
is
one
considering
the
request.
So
we
are
kind
of
the
development
of
the
process.
Is
your
single
belong
Bekasi
rule
and
the
wheel
being
a
sub-directory
issue
for
the
factory
needs
so
some
time
process.
J
Another
one
is
that
every
issue
are
at
least
once
equal
labels,
so
it
means
that
they
are
sponsored
a
issue
that
feature
and
every
peer
should
be
associated
with
that
he's
the
one
issue.
So,
and
so
don't
you
you
make
sure
the
almost
kind
of
the
process
is
working,
so
we
can
adapt,
introduce
my
country,
they
want
development.
Our
issue
associate
active
state
machine
and
actually
an
octave
slide
about
director
initial
spirit
and
make
a
small
modification.
J
So
then
we
introduced
the
first
renew
our
labels,
two
photos
together
the
factors
in
machine,
so
it
is
a
crew,
Kemal,
stone
and
also
stickers
in
process
and
the
status
in
the
vehicle
soon.
So
we
have
to
try
to
impose
the
signal.
Each
group,
a
second
each,
to
understand
issues
for
each
tyrannies,
and
then
we
use
an
approval
for
mountain,
so
only
see
girl
might
unnerve
is.
J
Of
have
to
decide
it
is
approval
for
that
milestone
like
a
pro
master.
No,
we,
it
is
means
you
are
going
to
restaurants,
testicle
dressed
up
for
reading
hitter,
feature
implementation,
testing
and
also
the
documentation
and
release
notes.
So
we
kind
of
doing
some
of
those
kind
of
things
and
the
same
cigar
act
again
and
he
went
with
us
be
nice
to
you,
those
kind
of
things
he
worker
warehouse
weights,
that's
that
bill
and
I
try
to
build
it
is
this
aggregate
reveal
and
staff
do
those
kind
of
things
and
I?
J
Just
do
me
a
quick
house
reshot
for
the
one
page,
so
you
can
see
here's
the
so
many
opens
opening
issues
Mac
for
the
1.8
and
the
four
by
the
different
groups.
But
it's
because
we
in
dreaming
still
is
not
really
emphasized
by
the
for
the
PR,
and
so
you
can
see.
There's
the
red
bow
on
the
left.
There
is
for
Pia,
so
most
of
us
know
stables,
so
sopia
actually
should
be
automatic.
If
the
ones
appear
it's
a
security
issue.
This
would
be
automatically
on
track
to
be
done
without
rules
the
group's.
J
J
So
you
can
see,
for
example,
was
a
cool
book.
We
have
like
the
feature
read
for
issue
and
for
the
army
Mossad
without
that
one,
so
you
can
see
in
1.9
alpha
one
and
you
know
it
fixed.
We
have.
We
have
to
make
the
incremental
improvement
in
the
other
one.
Then
we
can
either
adding
just
whatever
process
process
and
then
and
I
have
not
supported
for
Durbin
for
the
number
of
the
open
files
and
in
the
next
release.
Actually,
we
are
continue
using
the
same
feature,
ambreena
issue.
Then
we
sit
up
here.
J
It
is
I
will
add
a
meet
with
each
new
enhancement
for
the
same
pictures,
so
we
also
connect
the
show
that
so
there
are
some
a
graphics
in
the
theater,
whether
the
caustic
group,
little
restaurants,
for
the
city
wagon
things,
it
would
take
something
add
some
new
features,
so
something
like
and
so
a
manage
in
is
once
we
apply
enforce
those
kind
of
process.
We
can
even
go
further
to
auto-generate
of
the
renews
notes,
so
I
did
a
song,
like
example,
recipient,
and
that
it
is
what
we
had
I
just
copy
paste
once
sleep.
J
Yet
I
mention
that
earlier
we
have
more
than
400
like
this
and
image
of
the
point.
There's
no
way
to
process.
Actually
we
spend
engineering
are
and
all
those
entries
and
I'm.
You
know.
That
is
what
many
work
human
being
I
mean.
What
what
we
do.
Actually
this
is
can
be
auto
generated
by
other
tools
once
we
enforce
those
rules,
so
so
the
kind
of
kind
of
status
in
one
hand
7
it
is
rewind.
We
were
influenced
that
every
PR
should
be
associated
this
issue.
It
is,
it
is
automated
and
in
fact
you
don't
see.
J
I,
don't
have
the
issue
associate
and
then
you
can
approve,
and
we
will
remind
you
and
but
this
one
this
is
sweetest,
you
need
the
enhancement
I've.
Always
perhaps
it
right
now
only
look
at
an
image
repository.
They
shouldn't
know
Campbell
repository
and
those
kind
of
things
and
organized.
So
we
have
two
coherent
real,
coherent,
coherent
rule
against
are
in
depository,
so
we
we
should
have
work
on.
Those
kind
of
things
includes
another
one
you
are
done
in
the
100
means
it
is.
J
We
require
every
issue
at
least
once
a
group
label,
so
I
think
this
is
a
huge
help.
People
solve
anything
that
settles
and
the
wheel
so
so
engineer
how
many
educators
release
team
many
people
own
me,
the
staff
field,
with
your
actually
the
most
important
most
useful
for
them.
It
is
like
appeal
based
on
the
Penguins
by
the
state
group,
so
they
can
see
the
data
what
it
is.
Let
me
new
things
open
all
those
kind
of
things.
So
so
there's
the
we
already
done.
J
Those
kind
of
things
at
this
so
I
will
also
put
the
ink
in
April
and
like,
for
example,
we
still
optional
for
the
kind
put
a
kind
and
area
label
and
also,
as
you
see,
that
people
may
be
mistake.
So
so
we
need,
but
it
is
kind
of
the
work
lead.
It
has
most
it
ends.
We
need
a
stick
group
or
really
responsible
for
the
charge,
their
religion
of
issues
and
make
sure
that
is
cracked
up.
In
fact,
things
here
and
another
one
is
immediate:
did
we
already
added
a
link,
posted
and
others?
J
You
know
to
do
all
those
kind
of
things
visually
enhance
and
once
you
start
the
peer
assignment
in
the
video
and
then
you
should
have
issued
the
system
to
label
and
once
you
start
connect
the
review
process,
the
review
on
you
speak
of
a
sign
will
appear,
then
that's
a
review
process
and
all
those
kind
of
things.
Of
course
that's
the
other
thing.
We
need
to
iterate
our
mistake
machine
for
the
issue
like
that.
The
issue
reopen
what
we
should
do
of
all
those
have
things
we
can
discuss.
J
J
In
this
reason
they
start
thinking
about,
and
this
is
way
and
we
can
lab
loop
or
after
archetype
right
her
and
earlier
the
lanthanum
cycle,
and
so
we
could
one
things:
what
I
propose,
how
many
forces
cannabis?
We
could
operate
much
of
not
merge
the
peoc
without
the
Steve
acceptable
each
issue,
and
so
so
we
also
need
the
enhance
of
that
those
think
of
who-ville
a
prolonged
PR
without
no
issue
things
like
no
secret
office,
those
things
that
happen
so
another
thing
don't.
A
J
J
It's
generator
script
I,
we
need
country
and
each
protein
to
help
to
to
have
the
script,
to
the
build.
Really
nice
report
for
each
sake,
and
another
thing
is:
we
need
up.
Your
own
ring
is
tool
and
to
auto-generate
base
notes
based
categories
by
mistake,
kind
and
area
Legos.
So
then,
then
epically
it
can
be
some
at
home
to
produce
more
readable.
J
F
K
K
I'll
see
if
I
can
share
my
school
quickly,
as
always
I'll
be
going
off
of
meeting
news
for
this.
I
could
always
put
some
thing.
I've
talked
because,
like
I
preload
these
tabs
and
then
it's
always
have
to
actually
click
there
under
funding.
Okay,
so
let's
see
things
I
want
to
talk
about.
Today
we
have
a
b18
milestone
and
interesting
for
replay.
The
testing
for
Rico
is
where
same
testing
has
pretty
much
understood.
The
cooperation
with
folks
experience,
I'm
Stevie,
improving
the
automation,
labeling
issue
at
all
across
the
page.
K
Our
hope,
going
from
there
is,
must
use
a
script
called
bootstrap
that
is
also
written
in
Python
that
you
can
eventually
turn
this
as
we
go
like
I'd
like
to
once
again
improve
the
ease
of
use
of
this.
You
have
a
significant
amount
of
work,
you'd
like
to
do
any
areas
of
crowd
and
tester,
so
proud,
even
get
other
than
doing
a
thing
with
it.
We
would
like
to
make
it
more
reusable
for
more
people,
I
think
ultimate
national
ears.
K
We
want
to
be
able
to
have
people
stand
up
the
kubernetes
cluster
and
have
them
run
all
the
same
components
that
me
ourselves
are
using
to
test
the
kubernetes
project
as
a
whole
problems.
But
one
component
of
that
there
are
some
hard
coded
things
that
assume
that
it's
operating
on
the
kubernetes
organization
or
follows
some
communities
conference,
and
we
have
folks
from
other
projects
not
to
necessarily
call
them
out
by
name.
K
But
if
you
want
to
come
check
out
views
and
also
working
on
the
crab
code,
ace
I'm
super
excited
to
see
them
some
things
that
get
crowd
into
their
own
workflows.
We
are
hacking,
we're
making
this
more
reusable,
the
other,
the
other
key
component.
A
lot
of
people
have
day-to-day
experience
with
is
munch
github,
which
is
also
known
as
the
submit
to
you.
That's
the
thing
that
sweeps
through
github
and
does
things
to
issues
or
pull
requests.
K
K
Github
is
you've
around
one
instance
per
repository,
whereas
we
have
one
instance
that
can
cover
multiple
organization
is
in
multiple
repositories,
so
for
things
like
ensuring
that
the
same
set
of
labels
is
consistently
applied
through
all
repositories
across
the
litany
of
kubernetes
related
organizations
and
posit
or
ease,
although
there
is
a
monitor
to
do
that,
it
makes
more
sense
to
extract
that
functionality
into
a
proud
look.
I,
along
with
another
col
Widener,
are
planning
on
working.
H
K
Extracting
as
much
done
as
we
can,
but
at
the
same
time
making
these
what
remains
a
bunch,
you
have
a
little
more
operations
friendly.
It's
a
piece
of
code,
that's
been
organically
maintained
and
passed
around
from
person
to
person,
so
some
examples
of
what
we
have
done
there.
Let's
see
now,
instead
of
having
a
ton
of
flags,
it
shouldn't
be
reading
from
a
config
map
that
can
be
a
little
more
piece
of
urgent.
K
This
also
means
it
should
be
able
to
swap
the
config
map
on-the-fly
instead
of
automatically
restarting
submit
queue
and
waiting
the
hour,
so
it
takes
one.
For
instance,
we
also
discovered
that
much
of
the
same
kind
of
and
there's
a
bunch
github
was
responsible
for
running
a
school
called
both
LG
Chem.
It
was
unfortunately
using
the
same
token,
that's
the
submit
queue
was
using.
K
You
can
probably
tell
from
this
something
you
subscribe
up
top
here,
that
the
tool
was
deployed
roughly
around
April
8th,
and
we
didn't
actually
attribute
it
to
the
fact
that
this
make
you
was
really
slowing
down,
and
so,
unfortunately,
after
the
one
side
of
the
race
was
cut.
But
for
those
of
you
who
are
merely
writing
the
release
burned
down
train
all
the
way
you
may
notice
at
times.
This
is
a
bit
queue.
Take
a
while
to
get
to
your
pull
request
and
it
turns
out.
This
is
because
it
was
being
ranked
limited
by
github.
K
We
have
since
disabled
that
tool
and
had
an
open
issue
to
track
their,
why
it
was
using
their
token
so
much.
We
can
also
stand
it
up
again
with
a
separate
something,
although
thinking
is,
and
probably
slamming,
to
get
that
API
usage
once
again,
our
goal
is
for
2017
and
beyond,
are
also
in
a
placeholder
milestone,
and
it
doesn't
have
a
date
attached
to
it
and
talked
about
right.
The
other
thing
I
wanted
to
say
a
test
grade.
Iii
will
get
there,
but
we
think
test
grid
is
fantastic.
K
It
is,
unfortunately,
the
one
component
of
our
test
infrastructure
that
is
not
open
source,
specifically
in
the
piece,
that's
responsible
for
straight
being
all
the
data
from
the
buckets
and
turning
it
into
something
that
can
be
consumed
by
the
App
Engine
that
further
we
per
meter
and
test
grid
are
both
apps
that
run
on
App
Engine.
They
don't
run
on
a
kubernetes
cluster
and
someday
in
the
future.
It'd
be
great
if
we
could
change
that,
so
that
everything
could
growing
on
a
cluster.
K
K
Dashboard
it'd
be
great
if
we
had
a
little
more
documentation
on
how
it
works,
so
speaking
of
class
spirit
and
all
SIGG's
having
dashboards
in
lieu
mechanically
consuming
this
from
the
job
stock
config
that
JSON
I
think
it
is
that
has
any
signal
owners
field
that
people
pull
update.
We've
just
gone
ahead
and
manually
created
tester
dashboards
or
most
of
the
six.
Some
of
these
are
filled
out
with
data.
Some
of
them
are
not
so,
for
example,
the
dashboard
or
for
EPP.
K
Our
machinery
is
an
example
of
a
dashboard
that
the
API
machineries
ASIC
on
themselves
to
clear
out
the
configuration
that
is
useful
to
them
right
now,
I
believe
it
is
populated
with
a
GCE
slow
job.
Meanwhile,
the
Federation
has
a
dashboard
that
has
tests
that
only
relate
to
the
Federation
related
results
of
giving
the
jobs.
So
in
these
dashboards
work
right
is
they
can
take
other
jobs
like
the
GC
Eastland
job
that
what
it's
like
all
the
tests
and
then
pre
apply
regex.
That
only
applies
to
you
that
I
suppose
takes
interest
in.
M
K
Okay
right,
so,
if
you
want
to
find
out
more
about
test
grid,
we
recently
had
a
presentation
about
two
weeks,
obviously
custom
meeting.
You
can
find
a
link
to
the
recording
with
India
medians
and
also
find
that
at
the
cig
testing
meeting
YouTube
channel,
which
is
on
the
gratis
YouTube
channel.
Sorry,
this
is
a
playlist
on
the
communities.
Youtube
channel
and
Michelle
was
kind
enough
to
provide
her
slides,
which
again,
at
present
the
working
honest,
are
probably
the
best
form
of
documentation.
K
Part
about
a
ton
of
neat
features
that
you
may
not
necessarily
be
using
to
our
best
advantage
right.
So
that's
available
for
your
consumption.
If
you
want
to
find
out
more
about
this
possible
tool,
let's
see
so
I
think
that's
about
it
for
what
we've
done.
What
would
we
like
to
do
going
forward?
Well,
anything
that
came
up
during
the
retrospective
meeting
earlier
this
morning
is
I
would
like
to
have
people
specifically
for
we
sing,
but
those
people
do
care
about
this
admit
she
was
down.
K
What
can
I
do
about
it
to
show
up,
and
let's
share
our
stories
about
this?
Let's,
let's
identify
the
pain
points
and,
let's
see
if
we
can't
identify
requirements
or
how
we
can
make
it
easy
to
figure
out
if
this
makes
us,
why
am
I
talking
about
is
reviewing
some
history
on
this
when
I
remember
that
if
you
use
that
aid
are
these
page,
this
page
used
to
have
a
big
red
bar
at
the
top
that
says
each
me
test
fail
and
submit
cute
blocked.
We
don't
have
that
anymore.
Why
is
that?
K
It's
because
we
no
longer
block
the
submit
keel
on
tests
that
run
after
encode
has
merged.
We
only
prevent
code
for
merging
if
the
pull
request
tests
there
past
is
also
known
as
crucibles.
That's
great
that
has
improved
the
velocity
of
this
project.
I
think
I
feel
like
it
has
I
believe
the
rest
of
the
week's
team
would
agree
with
me
in
this
regard.
You
don't
want
to
go
back
to
the
battle
days
of
having
tests
run,
walk
the
whole
project
that
he
works.
Aware
of
at
all
request
time.
K
However,
this
means
it
can
be
a
little
difficult
to
kick
her
out
when
the
submit
queue
isn't
doing.
What
it's
supposed
to
mean.
We
merge
encode
a
good
example
of
this
was
recently
from
around
Hawaii
to
July,
11th
or
so.
The
chaos
AWS
jock
was
timing
out
pretty
consistently
ninety
minutes
across
all
pull
requests
and
it
took
a
while
for
this
to
be
really
obvious
to
notice.
K
The
problem
is
that
if
you
look
at
crowd
the
thing
that
loans
are
pre
submit
jobs,
it
shows
all
the
jobs
or
any
for
all
the
pull
requests
and
sometimes
be
difficult.
You
rock
a
pattern
out
of
this
real
quickly.
There's
also
crowd
running
attachment,
which
is
basically,
if
giving
a
bunch
of
test
results
on
behalf
of
the
civic
use
and
once
the
test
is
passed
all
in
its
pre
submits
and
then
price
action
to
merge
the
thing
we
were
also
rating
into
that
same
job.
K
Failing
right
before
a
merge
time
and
again
it
can
it's
cool
that
it
has
multiple
requests
at
once,
but
it
can
be
difficult
to
scan
through
this
and
identify
a
pattern
quickly.
So
it
just
sort
of
took
humans
organically,
starting
to
notice
that
hey
a
lot
of
pull
requests
are
failing
on
this
one
company,
a
job
for
us
to
realize
that
something
was
up
with
to
submit
to
you,
we'd
like
to
be
find
a
way
to
this
process
and
I'm
very
open
to
suggestions
on
any
other
topics.
People
would
like
to
discuss
at
suggested
meetings.
H
A
L
So,
in
the
meeting
notes,
there
is
a
link
to
the
discussion
thread
in
the
state
cluster
life
cycle
group.
When
I
asked
that
each
of
the
installer
tools,
please
reply
to
say
that
you
will
be
participating
and,
of
course
anyone
is
also
welcome
to
participate
whatever
your
interest,
but
ideally
a
representation
from
all
these
four
tools.
F
I
think
Justin
said
it,
but
I
just
want
to
make
sure
it
comes
across.
This
is
not
like
folks
can
use
cube
admin
as
a
sort
of
a
way
to
actually
get
an
installer,
but
we
also
really
want
a
tool
box.
So,
even
if
your
install
methods
are
very
different
from
the
wait
cube
and
it
works,
let's
try
and
find
the
pieces
of
what
cube
Edmund
does
that
we
can
extract
and
actually
sort
of
you
know,
share
across
as
many
installers
as
many
situations
as
possible.