►
From YouTube: Kubernetes Community Meeting 20160128
Description
We have PUBLIC and RECORDED weekly video meetings every Thursday at 10am US Pacific Time.
https://docs.google.com/document/d/1VQDIAB0OqiSjIHI8AWMvSdceWhnz56jNpZrLs6o7NJY
1.2 Release Update and Release process [T.J. Goltermann]
Demo (20 min): Deis/Helm - [Jason Hansen]
SIG Report - SIG Testing and Flaky tests update [Ike McCreery]
SIG Report - SIG Scale [Bob Wise]
A
A
28Th
I've
now
started
watching
a
bunch
of
these
and
publishing
them,
and
so
it
turns
out
to
be
very
convenient
if
I
say
the
dates
at
the
beginning.
So
this
is
January
28.
We
are
going
to
talk
about
first
up
the
1.2
release
miles
so
so
the
TJ
can
jump
off
and
head
to
his
next
meeting.
Do
we
have
a
note
taker
for
today?
Can
somebody
volunteer
and
do
that
we
did
get
the
notes
published.
We
did
get
the
videos
published.
A
The
notes
got
published
this
morning
on
the
Cooper
Nettie's
vlog
so
feel
free
to
share
them.
If
you
want
to
get
more
people
joining
us
and
the
videos.
The
four
videos
that
I
had
from
the
last
month
and
a
half
two
months
have
been
published
on
the
Cooper
Nettie's
youtube
channel
and
if
any
of
you
have
recording
meetings,
please
feel
free
to
send
me
those
dates
as
well
double
check
and
see.
If
we
can
publish
roles
as
well,
but
on
to
TJ,
hey.
B
So
my
name
is
TJ
cauldron,
I'm,
a
technical
program
manager
here
at
google,
and
so
that
is
a
title
that
is
focused
a
lot
around
planning
scheduling,
execution,
all
that
kind
of
stuff
to
keep
the
trains
running.
And
so
my
discussion
point
today
is
sort
of
about
the
1.2
release.
Cadence
to
it
see
me
and
so
I'm
sort
of
like
over
a
little
bit
of
history,
real
quick
to
put
it
in
context
of
10
and
11.
And
then
you
didn't
just
sort
of
thoughts
on
1.2.
B
B
Where
we
don't
let
any
more
major
features
in
we
kind
of
try
to
slow
down
the
churn
on
the
branch
fix
lots
of
above,
do
our
testing
and
then
do
the
release,
and
so
basically
that
was
those
are
trial,
run
for
1.1
for
1.2.
We
took
sort
of
sort
of
feedback
from
the
community
in
other
places
and
what
we've
done
differently
so
far
is
we
got
feedback
that
the
three-month
minor
release
version
people
like
that?
They
don't
want
us
to
go
terribly
fast,
officially
for
people
deploying
on-premise,
and
so
that
seems
to
be
positive.
B
But
then,
when
looking
at
1.2
on
some
of
the
area,
that
would
be
a
coating
over
lack
of
holidays
in
November
and
December,
where
a
lot
of
people
are
out
of
the
office
and
not
working,
and
so
we
lengthen
the
sort
of
coding
time
for
features
to
account
for
that.
And
then
we
also
found
it
for
1.1
the
sort
of
four
week
period
between
calling
it
more
or
less
feature
complete
and
actually
releasing
the
blessed.
B
For
that
and
then
one
other
piece
of
feedback
was
that
we
want
to
minimize
the
time
that
the
github
master
branch
is
in
sort
of
a
code
freeze,
code
slush,
and
so
we
we
want
to
iterate
on
our
wrenching
strategy
to
make
sure
that
you
know
features
for
for
v-necks
can
keep
flowing,
and
then
we
cut
the
branch
at
the
right
time
and
so
shows
up.
B
If
you,
if
you
do,
the
math
that
would
put
for
1.2
or
current
milestone,
feature
complete
goal
was
sort
of
end
of
January,
with
a
release
in
early
March
steps
or
the
six-week
period
there.
The
end
of
January
sort
of
I
did
a
straw
poll
about
a
week
ago
on
the
major
features
I
knew
about
that
were
coming
in.
Many
of
them
were
in
and
finished
on
prey
two-thirds
to
three-quarters,
but
the
remaining
twenty
twenty-five
percent
we're
not
yet
done,
and,
as
we
talked
last
week
in
this
meeting,
we
had
a
lot
of
tested.
B
B
B
Cool
yeah,
so
then,
essentially,
what's
happened.
Is
we
more
or
less
slipped
feature
complete?
We
were
aiming
for
last
week
but
decided
that
tests
were
more
important
and
so
been
spending
much
of
the
last
week,
both
at
Google
and
in
the
community
fixing
an
awful
lot
of
the
test
too
deep,
like
women,
make
sure
that's
more
solid.
Could
you
pause.
C
At
that
point
and
I
think
one
of
the
things
that
came
up
but
I'll,
we
may
ask
again
at
the
end
about
release
process,
but
since
you
said,
we
decided
blah
blah
blah.
Could
you
discuss
the
process
like
who
was
involved
in
the
discussion,
the
process
by
which
that
decision
was
made
and
the
communication
mechanism
that
was
utilized
to
communicate
that
outside
of
this
meeting
yeah.
B
So
it
was
basically
output
of
last
week's
community
meeting
a
week
ago
today
and
and
sort
of
the
need
to
refocus
on
test,
and
then
what
happened
was
at
Google.
There
was
a
meeting
that
came
exactly
thereafter
to
talk
about
precisely
that
issue.
We
also
had
seen
the
tested
been
an
issue
and,
and
mostly
it
was,
it
was
an
acute
issue,
because
the
submit
Q
was
was
blocked
up.
C
A
C
You'll
certainly
know
from
my
comments
last
week
that
I'm
entirely
supported
and
grateful
of
anything
that
puts
this
kind
of
testing
to
the
top
of
the
list.
So
it's
really
great
I
think
what
I'm
really
saying
is
is
this
kind
of
meeting
really
I
think
should
have
been
more
open
to
the
community
generally,
especially
if
decisions
being
made
about
the
process-
and
you
know,
one
of
the
other
comments
that
came
up
at
the
meeting
last
week
was
about
sort
of
be
a
papist
of
some
of
the
testing.
C
Some
of
the
testing
and
decision
process
here.
So
I
think
this
is
just
a
I
know.
I
know
this
is
google's.
Put
an
awful
lot
of
effort,
has
a
huge
number
of
people
and
got
this
project
off
the
ground,
but
it
seems
like
a
good
point
to
point
out
some
places
where
kind
of
as
a
open
source
community.
We
need
to
make
some
progress
so.
A
I
will,
I
will
wholeheartedly
agree
with
you
on
that
bob
and
point
out
that
we
had
our
first
cig
testing
group
last
meetup
group
last
week,
the
first
sig
meeting
for
that
and
two
people
from
Samsung
delightfully
attended
and
then
both
said
they
couldn't
help.
They
didn't
have
time.
So
if
you
also
want
to
prioritize
their
time
to
help
us,
that's
great,
it
will
push
us
that
much
harder.
A
But
we
also
had
a
meeting
just
yesterday
and
I
believe
that
Ike
is
going
to
talk
a
little
bit
about
both
the
sig
testing
and
the
internal
Google
testing
infrastructure
meeting
yesterday
about
how
we
are
trying
to
find
a
way
to
build
and
prioritize
something
that
makes
it
way
easier
for
people
externally
to
contribute.
But
so
far
we
have
a
lot
of
people
saying
we
should
be
contributing
and
not
a
lot
of
people
saying
I
have
time
for
it.
So.
C
Let's
not,
let's
not
conflate,
transparency
of
the
process.
One.
C
And
I
think
that
these
two
things
are
tied
and
I
think
that
we're
going
to
have
to
make
some
steps,
and
I
guess
I
say
we-
I
really
need
google,
in
this
case
google,
it's
going
to
have
to
take
some
steps
to
be
more
open
and
inclusive
a
while
in
order
to
get
were
involvement
from
the
community
over
time.
Yep.
A
D
Big
long
email-
I
did
see
that-
and
that
was
a
great
first
step,
but
to
sort
of
echo
Bob's
comments,
what
I'm
sort
of
interested
in
I
was
one
of
the
Samsung
folks
who
attended
the
same
testing
meeting
and
what
I
got
back
was
a
lack
of
vision
or
clear
commitment
on
how
much
of
the
testing
work
that
we
discussed
was
going
to
be
done
prior
to
the
1.2
release,
and
this
brings
me
back
to
I,
don't
actually
know
what
is
officially
slated
for
the
1.2
release.
D
I,
don't
know
what
officially
is
the
code
free
state
I,
don't
know
officially
when
it's
supposed
to
go
out,
and
so,
if
there
were
missions
made
around
this
I
didn't
get
to
see
what
those
decisions
were
aside
from
your
wonderful
email,
let
alone
participate
in
that
discussion.
That's
that
that
I
keep
waiting
to
hear
about
during
this
community
meeting.
Okay,.
A
A
B
So
then,
let's
see,
where
are
we
so
1.2?
It
sounds
like
again
a
pulse
of
looking
at
the
you
github
issues
for
the
features
that
are
not
yet
completed.
Hopefully,
we
can
wrap
up
sort
of
fixing.
The
most
acute
test
needs
this
week,
and
then
it
looks
like
another
week
will
be
needed
to
finish
up
the
majority
of
those
features.
So
if
I
had
to
sort
of
read
the
tea
leaves,
it
sounds
like
end
of
next
week
is
if
it's
a
viable
feature
complete
to
get
most
things
in.
B
B
If
there's
major
features
people
know
of
that
are
absolutely
critical
to
get
in
1.2,
please
let
me
and
others
know
and
speak
up,
and
we
will
see
what
we
can
to
get
those
in
I'm
and
obviously,
if
you
know
come
mid
to
late
februari
build
stability
is
still
not
there
for
the
1.2
candidate,
we
won't
push
anything
out
on
a
hard
deadline,
but
it's
the
intent
to
still
try
to
get
four
of
March.
So.
D
Where
can
we
go
to
actually
track
the
status
of
what
is
slated
for
1.2
and
how
much
that
has
been
completed?
For
example,
I
notice,
there's
a
1.2
mile
stone
and
the
cubed
Eddie's
get
her
project
issues
section,
there's
also
a
1.2
candidate,
milestone
they're.
Also
a
couple
issues
that
talk
about
features
proposed
for
1.2,
there's,
also
a
1.2
road
map
document
floating
around
somewhere.
D
B
Think
that's
completely
valid
feedback
here,
there's
a
bunch
of
stuff
in
assorted
places
and
most
of
it
is
a
graveyard
and
not
kept
up
to
date.
The
two
things
that
I
have
noted
that
we
need
to
get
better
at
in
1.3
is
seems
to
be
publishing
and
an
updating.
You
know
a
schedule
where
people
can
give
feedback
at
the
beginning
of
the
milestone
and
then
tracking
towards
that
and
also
be
very
firm
about
which
features
arc
in
or
out
or
rather
I
guess
interrupts,
not
brick,
not
really
the
right
phrase.
B
Obviously
the
project
take
any
feature
that
is
is
that
the
quality
barn
comes
in
but
which
are
critical
for
the
next
release
and
things
that
we'd
consider
delaying
that
release
for
and
so
for.
1.2
I
admit
it's
a
bit
messy
and
my
goal
is
for
1.3
to
be
absolutely
absolutely
nail
this
down
in
the
meantime,
if
anyone
knows
of
something
for
1.2,
that's
critical
I'd
urge
you
to
definitely
either
either
speak
up
now
or
send
me
an
email.
My
email
will
be
in
the
meeting
notes
or
send
to
be
on
on
slacker
github.
B
My
handles
just
my
last
name.
Vultureman
and
I'll
make
sure
that
that
is
taking
into
account
included
for
sure
for
1.2
and
then
for
1.3.
We
just
have
to
be
more
in
the
open
about
this.
It.
A
D
A
C
I,
don't
my
suggestion
or
request
would
be,
and
maybe
this
is
a
little
ambiguous,
I
think
having
I'm
testing
folks
TJ
show
up
is
awesome
and
I
hope
we
hear
from
him
every
single
meeting,
but
there's
some
difference
between
a
release,
testing
process
and
overall,
like
release
planning
and
I'm
not
sure
if
you're
the
person
for
that
sarah
or
TJ
is
really
that,
with
the
test
hat
on
but
use
the
person
on
point
for
that.
Okay,.
A
So
hold
on.
Let
me
separate
tease
this
out
a
little
bit
right
now.
The
person
that
has
been
leading
the
testing
sig,
which
to
be
clear,
has
been
around
for
a
week,
is
Ike
McCreary
and
he
is
I
think
going
to
chat
more
or
give
a
short
update
later,
although
I
haven't
seen
him
yet
in
the
meeting
so
like
if
you're
here,
you
want
to
just
drop
a
note
into
the
chat
and
all
know
that
you're
here
for
your
update.
A
I'll
try
and
have
I
cover
that
if
he
is,
if
he
has
joined
us
this
morning,
TJ
is
the
technical
program
manager
for
Cooper,
Nettie's
and
Google
computer
google
container
engine
here,
and
so
he
does
a
lot
of
the
wrangling
of
things
like
timelines.
We
also
have,
within
the
last
two
weeks,
hired
someone
to
manage
releases
and
I
haven't
brought
him
in
yet
to
this
meeting
to
talk
with
you
all,
but
it
sounds
like
we
should
add
him
for
next
week.
A
A
So
we
have
three
different
people
that
are
actually
all
trying
to
push
these
topics
forward,
plus
myself,
plus
the
rest
of
the
team,
who
is
obviously
in
support
of
bringing
the
community
in
it's
just
a
matter
of
which
thing
first
on
a
list
that's
way
too
long,
so
keep
pressing
us
but
keep
pressing
us
with
and
I
want
to
do.
This
and
it'll
happen
faster.
I
bet
that.
D
Piece
of
feedback
regarding
the
lack
of
a
hard
date
for
code
freeze
part
of
my
concern
is
that
the
phrase
code
freeze
has
been
tossed
around
through
a
couple
cigs
throughout
the
week
and
I.
Think
that's
symptomatic
of
the
fact
that
this
is
the
only
time
where
we
get
a
weekly
cadence
broadcast
of
the
community,
but
there's
information,
that's
changing
that
might
be
relevant
to
the
whole
community,
but
if
it
only
pops
up
briefly
during
the
cigs
or
some
of
the
work
is
getting
done.
D
Me
but
my.
A
Preference-
and
this
is
just
one
voice-
would
be
that
we
send
things
that
are
more
like
sig,
updates
or
sig
summaries
from
the
conversations
to
the
dev
mailing
list,
because
it
is
less
likely
to
get
lost
in
the
chatter,
as
you
might
find
happens
in
in
the
slack
channels.
Does
that
fit
with
others?
Experience
yeah,
okay,
I.
D
A
So
why
don't
we
take
that
as
a
how
we
want
to
bring
sig
updates
going
forward?
And
you
know
even
if
it's
a
two-paragraph
summary
after
a
sig
meeting
at
least
it
can
start
a
discussion
if
someone
else
on
the
sig
group
thinks
that
this
was.
You
know
that
there
was
another
piece
that
was
more
important
they
can
chime
in.
So
then
you
have
just
little
summaries
and
we
can
start
more
transparent
discussion
there
or
transparent
summaries,
at
least
of
it
awesome
all
right.
So
now
we
have
a
plan
for
more
communications.
A
Of
course,
this
means
that
the
sig
groups
have
to
appoint.
Someone
is
a
note
taker
which
usually
happens
and
that
that
person
sends
a
short
summary,
so
we
will
get
that
moving
forward,
all
right
anything
more
on
releases
before
we
release
TJ
to
us
next
meeting.
Oh
come
on.
That
was
funny
you're
all
on
mute
right.
A
You
thank
you
that
the
track
loss
all
right,
fantastic,
then
Thank,
You,
TJ
and
clearly
you
will
be
coming
back
absolutely
awesome.
So
we
had
next
on
our
agenda.
A
demo
from
Jason
Hanson
at
dias.
F
Everybody
good
morning,
good
afternoon,
good
evening,
whatever
the
case
may
be,
I've
been
with
Angie
our
neediest
459
fears,
I
wanted
to
kind
of
show
what
the
status
of
our
alpha
is.
This
implements
a
Heroku
inspired,
workflow
for
taking
source
to
image,
2
and
deploying
and
managing
an
application
across
series
of
nodes.
The
one
was
based
on
core
lesson
fleet
and
as
we're
moving
toward
be
two
we've
decided
to
replat
form
on
top
of
Cooper
Nettie's,
so
I'll
demo,
the
alpha
version,
which
is
currently
out
there
I'm
sure
my
screen
here.
F
So
the
current
work
is
going
up
at
github.com,
/
daya
/
would
flow,
which
is
where
all
the
v2
stuff
is
landing.
As
you
can
see,
we
on
top
of
communities
cluster,
we
implement
a
simple
REST.
Api
developers
interact
with
that
portion
of
the
platform
through
a
through
the
de
sel
out
your
child
show,
and
then
they
ship
code
into
the
platform
via
git
push
as
we
move
into
the
community
is
sort
of
universe.
One
of
the
things
do
we
want
to
be
is
a
good
community
citizen.
F
To
that
end,
all
the
applications
created
by
of
the
day's
platform
exist
in
their
own
namespace.
There's
a
service
that's
bound
to
the
light
cycle
of
that
application,
and
then
replication
controllers
are
created
for
each
of
the
versions
of
the
application
has
code
is
pushed
into
the
platform
or
configuration
changes
are
made
so
right
here.
F
I've
got
a
three
node
criminy's
cluster,
that's
running
on
AWS,
and
here
you
can
see
on
the
various
pieces
of
the
day's
platforms.
We've
got
builder,
which
is
responsible
for
source
to
image,
a
database
which
is
responsible
for
persistence,
a
small
NCD
cluster
Mineo,
which
is
argue,
element
solution
for
on
cluster
object,
storage,
a
dedicated
docker
image
registry,
an
edge
routing
component
and
in
the
workflow,
which
is
the
API
surface
area
that
handles
authentication
sort
of
coordination
between
all
the
various
bits
and
pieces.
F
So
on
my
alt
machine
here,
I've
just
got
a
simple
go
based
application
that
is
relatively
straightforward,
just
runs
a
simple
web
server.
So
what
we'll
do
is
I'll
use
the
day,
CLI
and
just
do
a
create.
So
this
create
created
an
automatic
name
for
us
web
cut
purse
and
I.
Do
actually
is
that's
bad,
so
you
can
see
here
we'll
take
a
look
at
this
namespace
right
now.
There's
nothing
there,
because
I
haven't
pushed
any
code
into
the
platform
yet
and
I
get
remote
has
been
added
to
my
well,
it's
my
repo
here.
F
So
if
I
get
push
yes
master
key,
this
pushes
the
graph
up
and
and
currently
we
launched
a
new
pod,
which
is
a
taking
the
source
code
and
building
a
slug
and
then
pushing
that
todrick
storage.
Once
that
is
done,
the
platform
then
creates
the
service,
the
replication
controller,
with
the
replicas
count
set
to
1.
So
that
is
all
up
and
running
which
point
you.
The
developer
could
scale
so
base
to
skin.
Well
be
kills
5,
but
this
will
do
is.
If
you
talk
to
the
controller
cool.
Oh,
is
that
a
question?
F
Maybe
just
some
feedback
modifies
the
replica
count
for
the
replication
controller
and
grenades.
Does
what
committees
does
best?
One
of
the
other
things
that
we
do
is
we
take
build,
so
some
version
of
the
source,
plus
the
configuration
that
set
for
the
application,
represents
a
release.
So,
if
I
fig
so.
F
This
will
create
a
new
release,
so
a
new
configuration
is
coming
from
the
developer.
We
join
that
with
set
of
environment
variables
for
a
replication
controller,
and
then
we
will
orchestrate
a
rolling
deploy.
Now
this
should
look
familiar
for
anyone,
who's
used,
coupe,
CTL,
rolling,
update
and
and
as
soon
as
the
diplomas
object
is
released
12.
We
will
retool
this
to
use
those
creators
primitives.
F
F
Now
you
can
see
that
this
is
powered
currently
by
food,
so
we
also
support
relax.
So
what
this
will
do
is
create
a
new
release,
but
go
back
to
the
previous
combination
of
the
code
cluster
configuration
for
that
for
that
version
of
the
hat,
as
you
can
see,
we're
going
it
rolling
to
be
for
replacing
v3
and
then
that
should
go
back
to
whatever
the
default
powered
by
was
so.
This
is
this
is
currently
alpha.
F
F
So
for
folks
who
don't
know,
we
also
created
a
package
manager
called
helmets
I'm
inspired
by
homebrew,
it's
a
simple
way
to
store
set
of
repositories
or
set
of
manifests
in
a
repository
and
then
interact
with
them
using
helm,
CLI
tool
they
can
download
fetch
edit.
We
just
added
a
template
generation
support
in
03.
This
is
how
we're
currently
installing
the
days
platform.
F
F
So
how
I'm
target
uses
cube
CTL
under
the
under
the
covers,
so
you
know
what
you're
pointing
at
helmets
all
day
is
this
just
walked
through
and
basically
creates
all
of
the
various
communities
resources
see
them
absolutely
come
back.
The
platform
takes
a
minute
to
converge.
It
said
CD,
cluster
sorts
itself
out
and
all
the
components
come
up.
We're
intending
for
all
of
these
components
to
be
reusable
on
their
own.
So
if
you
do
need
that,
grouting
take
a
look
at
a
router
and
use
it
in
your
own
project.
F
We
expect
separate
release
life
cycles
for
each
of
these
various
platform.
Components
as
we
move
forward
thoughts
would
be
great
there.
We
are
currently
working
toward
beta
we're
still
moving
some
some
of
the
load-bearing
posters
around
the
house
and
once
that's
done,
we'll
cut
a
beta
release
and
outside
of
the
base
users
mailing
list.
So
if
you're
interested
in
learning.
F
F
Now,
instrumented
to
see
line,
we
are
adding
a
an
administrative
component
to
it.
Just
so,
the
operators
of
the
cluster
can
see
what
the
status
of
all
of
the
various
components
are.
What
versions
they
are
where
they're
running
with
their
health
is,
but
not
at
this
point,
there's
not
a
developer
facing
sort
of
web
create
an
create
an
application,
deploy
it
manage
it
through
through
the
website.
So.
A
F
So
one
of
the
things
that
we
are
doing
is
with
it
with
the
service,
all
the
applications
get
one
their
own
namespace
and
then
over
the
late,
which
I
should
have
actually
dumped
the
replication
controller.
Usually
folks,
don't
know
a
lot
about
crew.
Burnetii,
sorry
just
gloss
over
all
of
the
important
details,
it's
so
so.
Each
of
the
applications
it
gets
label
of
the
version
is
labeled
and
then
the
reputation
controllers
salt,
like
their
pods
based
on.
F
So
we
have
a
set
of
conventions
internally
that
we've
used
for
that
and
expect
to
be
using
additional
communities
resources
so
way.
The
deployments
it
go.
Resource
works
will
be
the
way
that
we
deploy
applications.
So
it
should
feel
very
natural
for
someone
who's
done
to
benetti's
native
stuff
to
interact
with
applications
that
are
deployed
and
managed
by
it.
Idaeus
that
that's
your
question
there
yeah.
F
Helm
helm
currently
calls
out
to
a
coop
CTL
binary
it.
So,
whatever
the
often
context
set
for
your
good
CTO
environment
is
what
helm
uses
dinner
included.
Cluster.
A
A
F
A
Coupon
in
london
is
happening
march,
tenth
and
eleventh,
and
the
call
for
proposals
is
open
until
the
fifth
of
februari
and
we
would
love
many
more
proposals
about
awesome
things
that
people
are
doing
with
Cooper
Nettie's
on
Cooper
Nettie's
next
to
Cooper
Nettie's
around
Cooper
Nettie's.
You
know
all
all
the
different
ways
you
can
interact
with
it.
Oh.
A
D
So
yeah
couple
things
there:
the
testing
surge
internally
at
Google,
it's
going
pretty
well
we're
closing
bugs
quickly
and
submit
q
seems
to
be
picking
back
up
again,
which
everyone
here
is
excited
about
and
I
expect.
Folks
on
the
outside
are
too.
There
is
still
a
lot
of
work
to
be
done
there.
We
have
been
putting
a
few
things
in
place
to
try
to
document
better
what
kinds
of
things
the
community
outside
can
do.
D
D
E
D
Information
on
how
to
join
their,
but
the
the
first
meeting
happened
two
days
ago
and
I'd
say
it
went
pretty
well,
we
had
some
folks
from
Samsung
and
one
other
person
who
not
sure
where
exactly
they
were
from,
but
the
sort
of
overarching
thing
seems
to
be
everyone.
It's
better
test
infrastructure
and
no
one
is
really
willing
to
prioritize
that
and
cut
other
work.
At
least
that
was
sort
of
the
overarching
thing
that
we
heard
in
the
meeting
on
on
Tuesday.
So
you
know
at
Google
we
are
putting
head
count
on
this
myself.
D
Joe,
Finney
and
Jeff.
Grafton
are
all
full
time
on
that
and
we're
about
to
get
another
person
coming
in
full
time
and
actually
may
be
yet
another
and
that's
going
to
make
things
better,
but
we
would
really
love
to
have
some
some
help
from
the
outside.
So
you
know,
if
you
can,
you
know
chunk.
D
Of
the
week
and
help
out
there
are
concrete
things
that
can
be
done.
In
particular,
we've
heard
a
lot
that
conformance
testing
is
something
that
people
would
like
more
of
if
they
have
clusters
upon
different
providers
and
want
to
know
whether
or
not
they're
conformant,
there's
some
tooling,
that
is,
that
exists.
D
Erin
Criken
burg
from
Samsung
has
written
up
an
issue
about
that
and
Carl
from
May.
So
sphere
has
done
some
really
great
work
on
that
and
we
would
love
to
have
more.
So
if
you
are
interested
in
putting
in
some
time
definitely
hook
around
that
and
if
you
feel
like
you
want
to
be
putting
in
time,
but
don't
know
where
to
start
paying
us
on
slack,
and
we
can
definitely
direct
you
in
the
right
direction
and,
of
course,
join
us
for
the
suggesting
meeting.
So.
A
A
A
But,
as
like
mentioned,
we
have
dedicated
headcount
now
that
is
actually
trying
to
look
at
this
and
make
this
more
accessible
and
more
more
community
driven
or
engaged
at
the
very
least
because
we
know
right
now
it
sits
behind
a
you
know,
the
Google
firewall,
and
so
much
of
it
is
not
even
possible
for
somebody
to
unravel
from
the
outside.
But
there
are
things
and
the
more
we
talk
about.
A
D
D
Know
the
conformance
testing
has
has
been
something
that
folks
have
said
loud
and
clear.
That
is
something
that
we
want
more
infrastructure
around
and
we're
saying
great
and
we're
going
to
prioritize
it,
and
also
we
have
a
lot
of
other
things
that
we're
prioritizing,
and
so
it
folks
on
the
outside
want
to
want
to
put
in
hours
there.
You
know
it'll
it'll
be
better.
D
A
As
I
mentioned
a
couple
of
times
today,
I
think
the
best
way
to
do
this
is
to
push
us
with
specific.
I
want
to
help
in
this
way,
and
I
can't
because
that
gives
us
more
priority
than
please
tell
us
how
you're
doing
this
thing,
because
if
you
say
I
want
to
help,
we
want
you
to
help,
so
we
will
figure
out
a
way
to
make
that
go
as
quickly
as
we
can.
A
Awesome,
oh
and
Eric
also
says,
submit
Q
now
links
to
the
reason
friend
and
failures,
so
they're
not
invisible
anymore,
that's
right
there,
but
there
was
work
done
to
do
that.
There's
been
a
bit
of
work
done
this
week
also
to
make
make
it
easier
for
people
to
see
the
logs
with
when
a
Googler
drops
a
URL
in
for
what
the
log
is
actually
are,
so
we're
trying
to
make
this
more
transparent.
It
is
all
incremental
steps,
though,
so
knowing
where
the
friction
points
are
the
pain
points
and
where
you
can
help.
D
Questions
I
just
wanted
to
give
a
huge
shout
out:
I
can
the
rest
of
his
team.
They
have
done
phenomenal
work
in
improving
the
conformance
testing.
They've
done
a
lot
of
work
around
testing
the
upgrade
process,
so
there's
now
a
really
in-depth
seven-step
job
that
tests
out
migrating
from
a
previous
version
of
humidity's
to
a
new
version
of
humanities.
I
believe
this
was
one
of
the
larger
pain
points
from
going
from
one
point
out
of
one
dot,
one
so
I'm
pretty
happy
to
see
that
a
lot
of
effort
has
been
put
into
that.
D
B
D
D
A
E
A
D
D
To
do
with
you
know
what
community
thinks
is
highest:
priority,
p,
0
being
the
highest
priority:
p,
3,
being
the
lowest
p
to
being
lows:
3,
2
or
3
another
label
for
those
who
are
new
to
the
project.
There.
We
label
issues
with
help
wanted.
Those
are
issues
that
we've
sort
of
flagged
as
probably
good
beginners
projects
for
people
who
aren't
as
familiar
with
the
project
but
want
to
get
started.
D
In
their
hands
dirty,
so
you
can
search
github
for
that.
We
also
classified
labels
by
team,
an
area,
so
you
can
look
through
the
through
the
labels
there
and
see
we
have
in
pre
user
experience
and
test
infrastructure
and
clustering
control
plane
and
then
also
different
areas
for
things
like
performin,
so
yeah
yeah.
So
those
are
the
important
ones
that
come
to
mind.
There
are
other
ones
floating
around,
but
we
can
definitely
is
that.
E
Yeah,
the
github
issues
would
the
ones
relevant
to
what
you
were
talking
about,
I'm
still
new,
so
I.
I
am
trying
to
figure
out
how
the
selector
labels
and
the
communities
product
itself
has
been
used,
and
so,
while
you're
asking
about
that
again,
this
documentation
or
best
practices
around
that
will
be
helpful.
E
H
I
Sarah
somebody
else
is
talking
to.
I
A
A
I
C
C
If
these
tests
are
being
changed,
and
my
goal
is
in
the
next
week-
let's
say
before
the
next
meeting
and
I
are
going
to
work
on
writing
a
doc
PR
to
try
to
propose
something
so
just
wanted
to
give
everybody
a
fake,
and
we
were
thinking
about
that.
Working
on
that
and
please
touch
base
touch
base
with
me
on
at
all.
So
red
red
hat
stem
st.
Clair
has
been
working
with
us
on
this
as
well
so
yeah.
A
I've
had
a
few
conversations
with
him
and
the
big
long
email
that
I
wrote
on
Monday
that
you
didn't
read.
Yeah
was
going
to
reference
that
so
you
couldn't
possibly
have
known
was
going
to
reference
that,
but
we
ended
up
deciding
that
specific
references
and
that
were
not
necessary
because
people
had
been
paying
attention
and-
and
that
was
pretty
clearly
discussed,
even
in
the
pull
request
of
a
we
shouldn't
have
done
that.
So
so
I
didn't
end
up
putting
that
in,
but
there
have
been
internal
conversations
about
we
can't.
A
We
can't
do
that
just
to
allow
the
testing
to
go
through,
and
you
know
if
it
looks
like
it
needs
to
be
done.
Then
there
needs
to
be
a
broader
discussion
about
it,
so
I'm
completely
open
to
a
let's
try
and
make
this
a
little
more
formal,
I'm,
also
a
big
fan
of
not
having
too
many
pain
point
policies,
but
but
having
unknown
best
practices
on
this
is
a
really
good
idea.
Yeah.
C
Like
I
think,
our
intention
is
to
make
just
like
a
really
lightweight
kind
of
notice.
Thing
will
be
the
proposal,
there's
probably
something
else
we
should
do,
which
is
to
get
that
set
of
constants
kind
of
centralized
into
a
well-documented
place.
If
it's
easy
enough
to
make
a
policy
like
this,
but
it
would
be
probably
pretty
easy
for
someone
to
accidentally
change
something
and
it
realized
that
that
they
had
just
tripped
across
this
this
problem.
So
we,
you
should
probably
do
some
work
to
improve
that
as
well
over
time.
Yep.
A
Awesome
well
I,
amusingly
or
sadly,
because
so
much
of
testing
is
still
opaque
to
the
community.
I
think
this
is
another
Google
to
do
so.
Yeah,
so
I
will
take
it
as
such
and
we
will
see
what
we
can
do
from
there
until
labels
and
what
they
do.
That
strikes
me
is
something
that
anybody
who's
been
working
on
the
project
for
a
while
could
do
so.
If
someone
wants
to
pop
up
and
write
a
draft
of
that,
that
would
be
awesome.
D
Sounds
like
a
great
topic
for
the
next
big
testing
meeting
honestly
is
just
may
be
splitting
through
what
all
the
labels
mean
and
why
we're
trying
to
partition
them
up
that
way.
Yeah
I
think
we're
talking
about
two
different
kinds
of
labels.
There
and
I
guess.
C
D
C
D
So
the
labels
to
denote
you
know,
testing,
environment,
stuff
and
then
there's
also
the
labels
for
like
when
we
put
a
label
on
a
github
issue.
What
does
that
mean?
And
actually
you
know,
I
nodded
my
head
at
the
testing
contents.
Documentation
being
able
to
do
I
actually
think
that
you
know
maybe
Bob
you
and
the
rest
of
your
folks
might
be
great
people
to
contribute
that
to
since
it
sounds
like
you
had,
you
know
you
experience
that
pain
immediately.
E
D
D
Okay,
I
think
that
makes
a
ton
of
sense
just
for
what
it's
worth.
We
were
thinking
about,
looking
at
moving
the
documentation
for
the
responsibilities
of
the
build
cop
and
the
Google
staff
who
are
watching
the
infrastructure
out
of
the
wiki
and
into
document,
and
then
probably
opening
up
all
request
to
modify
that
document.
To
talk
about
the
policy
is
that
the
correct
approach
I
had
three,
I
would
give
you.
A
G
Got
a
less
than
four
minute
item,
which
was
a
couple
of
times
trying
to
find
the
time
for
meetings
for
cigs
has
been
difficult
because
you
have
to
just
kind
of
read
through
all
of
the
emails
in
the
archive
yeah
mom
I
know
that
networking
has.
If
calendar
I
know
that
storage
has
one
but
getting
access
to
them.
It's
not
fun.
So
if
we
could
get
an
organized
location,
I'm
going
to
put
at
least
the
networking
one
on
the
wiki
page
but
I
don't
know
all
the
other
ones
right.
A
A
We
will
move
to
having
an
Events
Calendar
there
that
people
can
invite
and
then
there's
at
least
sort
of
a
pull
request
to
accept
the
event
and
put
it
onto
the
calendar
for
lack
of
a
better
description
on
that.
But
anyway,
the
invitation
allows
a
little
bit
of
curation,
but
if
we
want
it
to
be
fairly
light,
so
both
of
those
are
on
my
to
do
is:
if
someone
wants
to
start
one
of
those
and
move
move
them
forward
more
aptly
again.
A
But
if
one
of
you
wants
to
get
started
on
it,
you
know
hit
me
in
email
or
on
slack,
and
we
can
find
space
to
work
on
it
together,
all
right
that
takes
us
to
one
minute.
So
thank
you
all
for
joining
and
we
will
again
do
this
next
week.
So
if
you
have
items
that
you
want
on
the
agenda
for
next
week,
let's
build
that
out
we're
looking
for
a
demo,
we're
looking
for
use
cases
we're
looking
for
sig
updates.