►
From YouTube: Kubernetes Community Meeting 20160407
Description
We have PUBLIC and RECORDED weekly video meetings every Thursday at 10am US Pacific Time.
https://docs.google.com/document/d/1VQDIAB0OqiSjIHI8AWMvSdceWhnz56jNpZrLs6o7NJY
Elastickube demo, Kubernetes 1.2 launch postmortem, Community Development Process (good and bad examples).
A
Bertha
says:
I
always
start
this,
which
is
saying
that
I
have
turned
on
the
recordings
public
and
recorded
conversation
with
me.
Cooper
Nettie's.
Can
you
get
April
seventh
and
we
are
going
through
a
demo
today
from
elastic
cube,
then
we
are
going
to
go
through
a
short
post
mortem
success
with
David
Iran
chick
who
I
haven't
seen
David.
Are
you
on
the
call
yet
I'm
going
to
go
with
not
yet
we
shall
see
him
shortly,
so
David
Iran
chick
is
going
to
lead
a
modified
post-mortem
around
the
1.2
launch.
A
A
A
A
To
write
blog
posts
where
the
Cooper
Nettie's
community
blog
as
well
all
right
one
last
announcement,
then
we'll
get
this
kicked
off.
The
last
announcement
is
we
are
going
to
do
our
first
asian
friend
time
zone,
community
meeting
so
expect
updates
to
your
calendar
items,
the
mist
monday
or
sorry,
the
first
thursday
in
May,
when
we
have
when
we
would
normally
have
our
meeting
at
ten
a.m.
we
have
it
5pm,
so
I
p.m.
pacific
10
a.m.
Pacific
will
make
one
meeting
a
month
happy
and
friendly
for
the
Asia
community.
A
C
Give
me
know,
guess
that's
think
otherwise,
and
can
you
see
my
screen,
though
too
fantastic
world?
Thank
you
everybody.
For
your
time.
We
are
very
excited
to
show
you
what
we
have
done
here,
that
elastic
box.
My
name
is
Alberto
areas
maestro
and
funded
incipio
of
elastic
parts,
and
a
few
weeks
ago
we
release
a
project
called
last
week
cube,
and
so
what
is
this
elastic
cube?
So
far,
are
we
having
found
a
category
for
these
projects,
but
we
call
it
a
convenience
management
platform.
C
Our
goal
is
to
accelerate
the
adoption
of
coordinators
by
complex
large
organization.
We
believe
that
we
have
to
provide
us
with
some
extra
capabilities
on
the
Batphone
and
those
capabilities.
We
have
this
code
of
the
first
relief
of
elastic
fault
as
the
following
authentication
and
authorization
of
trust
of
resources.
C
We
also
are
very
diverse
of
collaboration
and
the
Fox,
and
we
believe
that
for
individuals
to
be
able
to
participate
on
the
operations,
you
have
to
have
a
system
of
notifications
I
like
to
know
what
is
happening
with
the
resources
that
you
have
provided,
have
traded
and
deployed
on
the
communities
cluster
and
finally,
for
the
overall
organization
to
open
up
access
to
critical
resources.
You
have
to
have
anything
on
recording
of
the
coast.
Air
operations.
C
Casa
q
is
available.
Today
we
try
to
make
it
simple
as
possible
to
deploy
on
any
communities
Custer.
So
for
that
we
have
created
a
pastor
script.
You
can
easily
download
execute.
We
also
have
open
source
elastic
you.
We
use
the
Apache
to
parole
license
with
accepting
contributions.
We
use
in
exactly
the
same
CLA
than
Google
areas.
So
in
terms
of
looking
at
how
to
contribute
is
true.
Beauty
is
three
forwards,
so
please.
C
Co-Executors
comp
bypass
in
reality
takes
a
few
minutes
to
deploy
elastic
cube.
It
runs
a
set
of
containers,
pots
of
reciprocation
controllers
inside
your
permanent
cluster
in
similar
fashion
that
the
dashboard
app
already
it
install
it.
So
it
took
just
a
few
seconds
to
verify
the
old
parts
of
the
system
of
elastic
q
bar.
I
will
run
once
you
have
installed
elastic
you,
you're
presented
with
the
UI
that
helps
you
through
the
creation
of
the
first
user.
That
is
going
to
be
better
for
the
cluster
administrator
of
elastic
you.
So
I'm
going
to
sign
up.
C
And
after
bad,
what
I'm
presented
is
with
any
console
I'm
going
to
switch
now
to
another
instance
and
a
half
of
elastic
cube,
because
I
have
already
configured
or
anything,
you
don't
have
to
do
a
lot
of
configuration
and
of
things
that
you
can
do
decide
who's
going
to
be
a
cluster
administrator,
enable
single
sign-on.
So
far
we
support
the
Google
Earth
open
ID
at
last
Monday.
We
have
an
ad
the
support
for
some,
oh,
so
you
will
be
rather
12
indicate
and
authorize
after
resources.
C
We
also
support
is
non
password
and
I,
don't
know
about
expression,
and
you
can
we
integrate
with
get
to
full
application
definition.
Space
come
for
much
my
default.
We
point
you
to
the
hub
/
charts,
github
repo,
but
you
can
actually
provide
your
own
repo
and
we
already
have
a
feature
request
to
add
authentication,
so
you
can
have
a
private
repo
with
your
charts
and
that
is
also
common
queries.
C
Finally,
you
can
also
configure
an
SMTP
server
and
this
is
critical
to
be
able
to
send
invitations
to
other
users
of
elastic
cube
and
to
be
able
to
get
notifications,
email,
notifications
of
the
life
cycle
of
resources
that
you
have
deployed
our
managing
to
invite
other
users
is
relatively
simple:
you
switched
the
users
menu
and
you
can
get
an
invitation
and
to
send
an
invitation.
Just
like
the
email
of
the
person
you
want
to
invite
and
select.
The
name
is
spaces
that
are
going
to
have
access
to.
You
can
also
put
in
it
small
note.
C
It's
a
welcome
to
Last
give
he
will
receive
an
email
right
now
and
he
would
like
that
email.
There
would
be
a
link
once
key
clicks
on
that
link.
He
will
go
through
a
similar
experience
that
we
had
when
we
create
the
first
user,
but
once
he
loves
saying
he
will
only
have
access
to
that
name
is
base
and
he
will
not
have
access
to
them
in
console
will
only
have
access
to
the
instacharge
menu
that
I'll
show
you
in
a
few
minutes.
C
Finally,
all
the
things
that
you
can
do
is
you
can
create
an
image
basis
and
you
can
invite
who
launched
to
those
living
spaces
and
if
you
should
be
also
going
to
add
the
ability
to
manage
quarters
such
as
you
can
restrict
among
how
many
resources,
those
members
of
those
mini
spaces
can
actually
use
of
the
cluster.
You
can
also
see
the
list
of
charges
that
we
have
synchronized
from
the
git
repo.
This
is
maintained,
synced
constantly
on
the
back
up
on
the
background.
C
So
that
means
that
if
somebody
does
made
an
ads
in
your
chart,
it
will
appear
immediately
and
they
always
we've
been
pointing
to
the
latest
version
across
chart.
So,
finally,
you
can
have
access.
You
can
have
at
least
access
to
the
east
of
instances
of
only
service
about
money
on
the
cluster
here,
as
you
can
see
most
of
the
services
that
are
running
our
services
that
are
backing
up
at
the
back
end
of
elastic
cube.
C
Alright,
so
once
you
have
done
all
this
administration
is
in
dec
that
loan,
you
can
see
all
the
charts
that
the
users
would
be
able
to
see
and
deploy,
and
this,
as
you
can
see,
you
can
also
hear
it
says
the
food
this
is.
The
name
is
picture
that
I
have.
I
can
easily
switch
between
namespaces
under
staying
at
the
phone
and
space
and
I'm
going
to
also
switch
to
the
instance.
The
screen
I
just
guessing
this
is
the
screen.
C
You
see
all
of
the
resources
that
are
currently
running
on
that
namespace
have
a
couple
of
services:
replication
control
it
on
a
pot.
If
I
click
on
the
replication
controller,
I
can
see
the
events
of
thermal
amplification
controller,
the
selector
and
the
instances
of
the
parties
messages
that
are
part
of
that
replication
controller
I
can
see
the
much
of
the
tags.
I.
Can
click
on
that
replicate
pot
and
can
get
more
details.
C
C
C
You
can
also
manage
the
lifecycle
of
those
resources,
so
I
can
select
those
resources
and
delete
them.
Basically,
they
would
be
destroyed
and
also
I
can
launch
new
instances
of
new
charge.
So
these
the
release
of
charts-
it's
led,
let's
say
fri
this
cluster.
I
can
provide
the
additional
labels
and
click
the
boy
and
the
orchestration
is
happening.
As
you
can
see,
the
images
were
already
from
so
he
was
really
fast
to
create
those
instances.
You
can
see
the
dance
everything
that
happened
I
compared
to
the
containers.
F
C
C
C
We
also
that's
one
of
the
request.
We
are
going
to
launch
a
kind
of
a
dedicated
decides
to
the
commendation
and
extra
information
on
how
to
abrade.
Although
we
are
trying
to
make
the
experience
of
elastic
you
as
simple
as
possible,
such
as
anybody
that
has
never
use
urban
areas,
can
go
into
elastic,
cube,
perform
the
deployment
and
even
perform
complex
operations
like
a
rolling
update
or
a
canary
deployment.
C
We
also
simply
find
the
ability
to
diagnose
that
we
create
night
Gnostics
and
see
if
anything
wrong
like,
for
example,
if
hipster,
even
in
compression,
that
we
may
not
be
compatible
being
able
to
show
you
that
information.
We
also
create
additional
information
like
dashboards,
such
as
when
you
jump
into
a
name,
is
Paige
to
get
complete
historical
information,
see
what
is
going
to
happen
in
I've,
been
able
to
drill
into
the
details
of
the
resources
that
running
on
that
name
is
place.
We
also
that
provide
an
additional
capabilities
of
historical
you
CPU
and
memory
usage.
C
So
that's
pretty
much
what
we
want
to
share
with
you
guys
if
you
have
any
questions
information,
so
you
can
contact
me
at
a
lot.
Alberto,
an
elastic
box
calm
as
I
mentioned,
is
open
source.
You
can
go
to
github
com,
alas,
the
box
plastic
cube
and
follow
us
not
to
start
the
project
or
sarah
has
been
generous
to
generate
to
create
a
slap
channel
and
the
Cooper
net
is
slag.
Is
elastic,
cube,
I'm,
going
to
post
a
link
to
the
leg
right
after
this
meeting
and
if
you
have
any
questions
will
be
there?
G
Yeah
one
question,
though:
so:
how
do
you,
how
do
you
integrate
with
the
I'll?
Ask
it
this
way,
I
guess
Cuba,
Nettie's,
trunk
or
cuban
eighties
releases
or
you
are
you
doing
modifications
or
you
lay
it
layered
directly.
On
top,
do
you
have
like
a
high
flow
or
something
you
used
to
to
integrate
we're
very
rapidly
evolving
projects.
C
Into
what
you're
doing
our
goal
is
to
make
sure
that
you
can
deploy
elastic
cube
in
any
cluster?
So
we
don't
ask
you
to
modify
your
existing
configuration
of
your
cluster
right
now.
The
authentication
and
authorization
happens
on
the
layer
above
the
cluster
that
is
provided
by
elastic
cube.
That
being
said,
we
are
our
first
contribute
contribution.
We
hope
is
over.
First
countries
include
the
coolest
called
project
will
be
to
integrate
with
the
open
ID
system
and
even
providing
implementation.
C
For
example,
authentication
such
as
the
tokens
that
we
generate
for
the
user,
authentication
and
elastic
cube
can
also
be
used
from
the
console
with
cube
CTL,
so
right
now
its
layer
above,
but
we
are
looking
at
ways
of
how
to
authenticate
thick.
Does
that
authentication
authorization
writing
to
the
core?
Oh,
and
what,
with
you
guys
to
make
it
a
standard.
H
H
C
Not
I've
been
talking
to
the
Red
Cloud,
guys
and
I
know
that
have
met
with
the
friend
authentication
system
as
well.
We've
been
working
into
the
community
for
in
the
community
in
the
last
few
weeks
and
we
are
looking
forward
to
build
these
connections
and
make
sure
that
we
have
a
concerted
effort
to
do
it.
Mom
I
standardized
authentication,
authorization
of
resources
in
the
cluster,
so
looking
forward
to
look
into
that,
there.
I
A
A
J
Hi
everyone.
How
are
you
thank
you
so
much
for
joining
the
community
meeting
as
always
we're
going
to
spend
the
next
call
it
20
minutes
or
so
doing
a
post-mortem
on
Cooper
Nettie's,
1.2
launch
and
a
little
bit
of
the
1.3,
with
let's
really
trying
to
stay
focused
on
1.2.
For
those
of
you
who
have
done
a
post-mortem.
This
will
be
old
hat
for
you
for
those
of
you
who
haven't
I
like
to
do
them
in
a
fairly
agile
way,
which
is
we
open
up
the
space
for
everyone?
J
All
voices
are
equal
to
comment
on
questions
and
things
that
they
have.
That
transpired
is
not
a
place
for
value
judgment
or
a
discussion
about
the
particular
thing.
I
may
ask
a
follow-up
question,
but
by
and
large
it
should
just
be
people
staying
or
typing
in
the
chat
whatever
worked
and
whatever
didn't,
and
I
do
want
to
really
stress
that
it
should
be
positives
and
negatives.
Please
say
whatever
you'd
like
us
to
keep
doing
more
of,
and
please
say
what
you'd
like
to
see
happen
next
time.
J
So
we'll
spend
maybe
e
about
10
minutes
doing
that,
and
then
perhaps
we
can
look
at
the
list
and
decide
if
there's
some
particular
items
that
we
want
to
dive
into
a
little
bit
more.
So
that
should
take
the
time
any
questions
before
we
begin
okay,
so
this
is
a
like
I
said:
all
voices
are
equal,
no
value
judgments
feel
free
to
have
at
it.
So
what
went
well
and
what
could
have
gone
better,
just
start
voicing
in
typing
whatever
you
like.
G
Hey
Bob
wise
here,
Allah
I'll
start
with
just
an
anecdote
from
one
of
the
one
of
a
fellow
CTOs
in
Seattle
who
is
running
Cooper
Nettie's
and
production,
which
was
released
notes
for
one
dot
to
this
is
a
real
pain
point.
We
ended
up
having
a
pretty
lengthy
discussion
about
it.
They
did
a
upgrade
from
11
to
12.
Pretty
aggressively
and
release
notes
were
apparently
not
actually
available
on,
or
they
were
hidden
behind
some
they
were,
they
weren't,
published
publicly
I.
G
J
A
F
Just
sort
of
totally
speaking
I
think
that
the
concept
of
a
post-mortem
that
only
lasts
20
minutes
for
a
release
that
took
a
month
seems
like
sort
of
a
short
amount
of
time
and
in
terms
of
soliciting
feedback
from
the
community.
I
think
maybe,
like
I
missed
the
mailing
lists
post
that
was
like
sent
out
a
week
prior,
but
this
seems
sort
of,
like
short
notice,
to
get
the
right
parties
involved
to
really
discuss
the
post
mortem
effectively,
not
saying
we
shouldn't
talk
it
over.
J
A
K
F
Sort
of
think
that
moment
where
a
future
development
stopped
and
we
froze
the
world
for
a
second
while
we
took
care
of
testing
stability
and
sort
of
head
to
triage
out,
which
was
flaky,
which
was
disrupted,
which
could
be
run
in
parallel,
probably
helped
the
project
quite
a
bit.
It's
sort
of
unclear
to
me
whether
folks
felt
like
that
was
sort
of
the
appropriate
way
to
deal
with
stability
issues
going
forward
or,
if
there's
there's
something
we
could
do
to
sort
of
integrate
testing
best
practices
a
little
bit
more
continuously.
F
And
also
just
you
know,
this
is
sort
of
what
I'm
trying
to
help
bring
to
fruition
with
sleep
testing,
but
once
again,
release
12
got
cut
with
not
that
much
coverage
on
alternative
cloud
providers.
I
saw
a
lot
of
you
know
coverage
for
every
single
PR
for
GCE,
but
there
was
a
gap
of
like
a
week
or
so
where
the
AWS
tests
weren't
even
passing,
and
they
weren't
even
running,
and
nobody
really
noticed
that
I
sort
of
really
like
to
see
alternative
cloud
providers
beyond
GCE
and
gke.
F
These
sort
of
the
gate
releases
going
forward
sort
of
on
that
front.
I
know
there
are
some
folks
who
are
trying
to
contribute,
as
your
cluster
stand
up
and
testing
and
Justin
Santa
Barbara
made
a
significant
impact
on
the
quality
of
e
2
e
testing
in
the
AWS
for
the
one
dot
to
release.
But
again
these
weren't
used
to
actually
gate
release.
It's
just
sort
of
nice
to
haves
we'd
like
to
see
that
taking
a
little
bit
further.
L
L
Those
are
pretty
massive
shifts
that
ultimately,
before
the
1.2
release
got
cut,
I
had
to
dig
through
the
code
and
docs
and
whatnot,
but
even
in
the
release,
notes
just
kind
of
like
a
blip
on
on
the
on
the
laundry
list
of
stuff
I
changed,
which
is
kind
of
an
understatement,
because
it's
such
a
radical
shift
and
how
you
interact
and
use
community
as
of
1.2,
with
deployments
kind
of
being
the
standard
rebel.
L
I
Think
in
general
there
were
some
pretty
big
changes
around
sort
of
the
model
like
you
know,
in
terms
of
some
of
the
core
concepts
and
I
in
you
know,
and
the
questions
things
like
you
know
ingress
and
is
it?
Is
it
as
part
of
Korres
it
out
of
core?
You
know,
I,
don't
feel
like.
We
got
great
sort
of
roll
up
in
summary
into
these
types
of
meetings
along
the
way
yeah,
and
so
at
least
with
the
community
in
general.
I.
I
Don't
think
there
was
a
lot
of
great
communication
of
that
stuff
and
I
think
some
of
that
translated
in
terms
of
the
way
that
we
message
it
I
think
that
there's
a
certain
amount
of
like
well.
Everybody
knows
this
because
everybody's
been
talking
about
it
and
if
you're
not
in
the
room,
all
the
time,
it's
it's
easy
to
miss
that
stuff,
and
especially
for
the
end
customers
again,
you
know
just
reiterating
that
point.
You
know
they
weren't
on
the
journey
with
us
through
1.2,
so
they
didn't
see
this
stuff
gradually.
D
The
perspective
those
had
was
that
you're
gonna
be
able
to
keep
these
replication
controllers
as
long
as
you
want,
and
so
you
can
learn
about
the
new
things
when
you're
ready
and
we're
saying
here's
a
new
thing
pretty
for
you
to
learn
about
you,
you
people
think
as
long
as
you
want,
it
is
going
to
be
the
new
thing
so,
but
I
mean
I.
Also
see
your
point.
I
just
wanted
to
share
the
other
perspective.
There
Thanks.
I
And
I
think
it's
important
to
remember
that
you
know
there's
going
to
be
a
ton
of
folks
who
are
you
know,
as
the
project
grows,
there's
going
to
be
a
ton
of
folks
who
are
using
kuber
natives
for
the
first
time
and
when
there
is
this
confusion
around
okay,
what's
deployments
wise
replications,
that's!
What's
replication
controllers,
how
do
these
relate?
I
You
know
the
sort
of
like
you
can
keep
doing
what
you
were
doing
when
everybody
you
know
to
a
statistical.
You
know
approximation,
everybody
is
new.
You
know,
I,
think
I,
think
making
sure
the
best
practices
are
clear
and
and
that
we
actually
update
and
communicate
those
best
practices
as
they
evolve
and
help
people
sort
of
transition
into
it.
It's
going
to
be
useful
for
both
existing
youth
in
new
users
like.
D
I'm
really
happy
with
the
changes
would
happen
with
the
new
docs
repo
that
make
it
much
lighter
weight
process
for
anyone
to
contribute.
Doc's
changes
to
do
just
what
Joe
said
it
used
to
be.
You
had
to
like
get
your
bill
going
and
run
a
bunch,
the
doc
generation
stuff,
and
now
you
just
can
edit
in
github,
don't
have
to
have
your
own
get
client
boom.
So
please
check
out
how
easy
it
is.
Contribute
your
dogs,
now
one.
K
I
You
didn't
break
backwards,
compatibility
I
mean
that's,
that's
actually
a
really
good
point.
The
cube
control
is
an
API,
it's
a
user
interface
I
mean
people
will
script
against
it
with
bash,
and
so
we
should
be
viewing
it
as
such,
and
so,
if
we
change
the
behavior
of
cube
control
run
that
should
be.
You
know,
under
a
flag
with
deprecation
messages
and
sort
of
you
know.
Maybe
we
don't
have
to
be
as
hardcore
with
the
as
with
the
API,
but
I
think
we
want
to
be
very
careful
about
changing
behavior
there
I
know.
M
Joey
we
have
a
generator
flag
that,
like
let
you
keep
the
old
behavior
but
I
think
we
didn't
do
a
good
enough
job
first,
telling
people
that,
if
they
expected
stable
output
for
our
commands,
that
that
should
have
been
using
the
generator
flag
in
their
scripts.
So
maybe
that
means
better
documentation
on
our
part
that
just
say
that
moving
forward,
but
like
I,
think
we
should
be
able
to
free
to
change
free
to
be
changing
the
behavior
of
Q
control
run
in
the
future.
M
I
Mean
I
mean
maybe
I
mean
I,
think
you
know,
look
at
the
get
tool
said
I
mean
they
don't
change
output
right,
I
mean
it's
like
they
just
don't
change
it
and
they
add
new
features
in
backward
compatible
ways
and
they
view
that
output,
as
as
as
part
of
the
API
I,
don't
know.
So
it's
a
it's
a
deeper
discussion,
but
you
know
I.
G
I,
just
could
not
agree
with
that
statement
anymore,
Joe
and
I.
Think
that,
like
my
opening
anecdote
is
a
good
example
of
that
11
I'll,
say
typical
way
to
try
to
address
this
is
to
think
about
having
something,
that's
more
like
a
formal
beta,
so
that
effectively
it's
released,
but
potential
users
have
from
a
little
bit
of
warning,
to
be
careful
and
I.
Think
that
helped
set
up
the
patients.
D
I
Would
expect
opens
up
a
couple
of
things
there
I
mean
so
as
features
move
into
beta.
We
should
expect
full
documentation
going
into
beta,
but
you
know:
there's
Anna,
you
know,
there's
an
assumption
that
anything
in
beta
is
going
to
be
a
little
bit
really
rough-and-ready.
So,
but
you
know
the
idea
is
that
by
the
time
you
actually
make
a
feature
GA,
it
should
have
been
in
beta
for
quite
a
while
and
and
all
that
stuff
should
be
smoothed
out
and-
and
you
know
the
documentation
should
be
there.
I
I
think
you
know
getting
more
formal
about.
You
know
what
we
need
when
and
where
and
how
I
think
is
going
to
be
critical
and
and
making
sure
that
we
review
documentation
and
tooling
as
a
part
of
actually
calling
a
feature
ready
for
beta,
or
you
know
being
able
to
slap
that
type
of
label
on
it
and
then
the
other
thing.
I
I
It's
in
pre
release
or
something
like
that
and
then
just
like,
encourage
the
community
to
actually
sort
of
look
at
the
experience
with
fresh
eyes
as
if
everything
was
public,
and
you
know,
and
then
at
some
point
a
week
later
we
actually
make
noise
and
that's
a
good
chance
for
everybody
to
sort
of
roll
up
their
sleeves
and
help
fix
the
documentation
and
make
sure
that
that
everything
looks
smooth.
You
know
and-
and
it's
just
a
matter
of
messaging
right-
we
just
you
know
we
can
you
know.
I
J
Why
we
released
1121
and
wants
you
to
so
quickly,
because
we,
you
know,
heard
things,
but
instead
it's
just
it's
almost
just
a
different
kind
of
lens
on
it
and
it
makes
everyone
in
the
community
very
aware,
hey.
This
is
out
there,
but
then
all
our
things
like
the
five
days
of
communities
and
other
things
like
that,
we
time
for
when
it
actually
goes
GA
rather
than
that
RC.
Is
that
roughly
what
you're
thinking
yeah.
D
I
N
A
one
point
to
point
you'd
like
to
make
yes
I
I
had
at
least
a
couple
users
hit
hit
a
case
where
they
had
upgraded
the
API
server,
I
inadvertently,
not
upgraded
a
node,
and
then
they
reported
an
issue
used
against
X
or
Y.
New
feature
doesn't
work,
and
it's
because
they
were
getting
scheduled
on
to
a
cube
that
didn't
support
the
feature.
So
Justin
Santa
Barbara
created
an
issue
for
this.
N
The
number
escapes
me
right
now,
but
I
think
we
should
consider
adding
a
condition
for
a
node
that
lags
the
version
of
the
API
server
that
it's
connected
to
you
to
make
it
a
little
bit
more
obvious
when
these
issues
happen,
because
it
definitely
confused
at
least
a
handful
of
people
that
we're
upgrading.
J
Please
can
ping
me
a
offline,
there's
actually
been
quite
a
bit
of
thinking.
Sorry
I
didn't
need
to
respond,
there's
actually
quite
a
bit
of
thinking
around
both
version
skew
as
well
as
potentially
helping
folks
to
have
better
visibility
into
when
there's
finally
skew,
but
when
there
is
drift
off
of
like
master
or
last
head
or
something
like
that
and
I'd
love
to
talk
to
you
more
about
that
yeah
sure.
I
But
having
a
way
such
that,
when
somebody
has
a
problem
or
a
bug
report,
they
can
easily
get
like
a
block
of
text
that
they
can
copy
and
paste
into
a
gist
that
lists
the
versions,
the
command
line,
flags,
all
the
tunable
parameters.
The
OS
like,
like
all
the
crap
that
we'd
want
to
collect
to
try
and
say:
oh
you're,
hitting
the
journal
d
hue
with
large
clusters
or
oh
look.
A
F
I
So
the
fact
that
you're
getting
silence
here,
Sarah
I,
think
you
know
I've,
never
shipped
a
software
product
and
been
happy
with
it
at
the
time
that
I
shipped
it
I
think
you
need
a
little
bit
more
distance
to
be
able
to
see
the
good
things
mom.
It's
just
you're,
always
too
in
the
weeds.
When
you
actually
have
just
shipped
something
I.
P
L
L
G
Make
I
guess
I'd
also
like
to
say,
having
been
involved
in
open
source
projects
of
one
sort
or
another
for
a
long
time.
This
is
the
most
fun
I've
had
in
an
open
source
project
in
quite
some
time,
and
it's
really
great
community
and
thanks
everybody
for
making
this
such
a
great,
a
great
community
to
work
with.
Really
it's
been
just
phenomenally
great
I.
J
I
The
community
is
starting
to
look
towards
wider.
You
know
problems
to
be
solved
instead
of
actually
just
noodling
on
the
existing
system
that
we
have-
and
you
know
in
the
past,
I've
said
that
you
know
our
goal
should
be
coo.
Burnett
is
boring
right
so
that
actually
can
be
stable
and
people.
People
can,
you
know,
predictably
use
it
and
I
think
I
think
we're
definitely
moving
in
that
direction.
I
think
that
that's
that's
really
good.
J
All
right,
so
in
the
spirit
of
using
a
hard,
stop,
I
think
this
is
extremely
valuable.
We
will
also
circulate
in
the
show
notes
the
a
survey,
so
people
can
do
this
offline,
ongoing
and
I
will
take
the
time
and
summarize,
let's
plan
on
maybe
two
weeks
from
now.
I
will
have
some
summary
things
and
I
will
also
be
reaching
out
to
folks,
either
on
the
call
or
elsewhere
to
own
improving,
given
summary
areas
and
notify
everyone
that
they
will
be
taking
that
on
so
again,
our
goal
is
continuous
improvement.
J
A
Q
Thanks
Sarah
yeah
I
just
wanted
to
let
everyone
know.
Brian
grant
has
passed
the
torch
for
the
maintenance
of
the
1.2
branch
over
to
myself
and
Zach
Marilyn,
so
brian
has
done
a
great
job
to
date.
Coordinating
the
large
cherry
picks
after
we
got
the
branch
to
to
make
sure
we
got
a
stable
release
out
now
that
the
number
of
cherry
picks
has
diminished
and
we've
gone
back
to
the
sort
of
one
PR
/
cherry-pick
model
we're
going
to
sort
of
distribute
that
responsibility
out,
so
Brian
can
keep
focusing
on
the
1.3
work
going
forward.
A
F
I
make
just
a
quick
comment
on
the
cherry
pick
process,
so
this
wasn't
for
the
1.2
branch.
It
was
for
the
1.1
branch,
but
I
went
through
sort
of
a
bit
of
a
trail
of
woe
of
bots,
adding
and
removing
a
wide
variety
of
labels
from
cherry
pick
approved,
cherry-pick
not
approve,
no
lgtm
release
notes
needed
released,
it's
not
needed.
It
seems
like
there's
a
that.
F
That
process
is
still
somewhat
in
flux
or
maybe
not
quite
as
tightly
documented
as
need
be,
and
the
the
point
that
I'm
trying
to
drive
to
is
once
again
I'm
a
community
member
who
does
not
have
the
ability
to
add
labels
to
a
PR.
So
it
means
that
I
need
to
be
extra
noisy
to
the
reviewer
assigned
to
my
PR
to
properly
assign
these
labels,
but
I
don't
know
where
to
point
them
to
to
make
sure
that
they
do.
They
apply
the
correct
set
of
labels,
so
my
stuff
gets
cherry-picked
appropriately.
Does
that
make
sense?
F
Q
And
I
totally
agree:
Jack
and
I
fought
with
the
labels
getting
flipped
back
and
forth
by
the
bots
for
a
couple
hours
yesterday,
because
we
were
trying
to
understand
what
they
were
doing
and
I
think
next
week.
Dave
McMahon
is
scheduled
to
give
a
presentation
about
the
release,
process
and
she's
written
up
what
this
is
supposed
to
look
like
and
the
bot
is
not
quite
caught
up
to
the
expectations.
Yet
so
Eric
Paris
is
working
hard
to
catch
that
up
and
it
should
be
a
lot
smoother
going
forward
and
I.
Think
there's
two.
R
Things
here
as
well,
one
is,
we
should
have
a
state
machine
like
we
should
just
draw
that
out.
Hopefully
that's
what
David
will
show.
Oh
she
get
that
in
the
docs
of
like
here's,
the
various
things
and
where
you
go
through
and
I
think
you
know,
part
of
the
problem
is
also
the
github
permissions
or
not
fine-grained
enough,
and
we
should
potentially
consider
I
hate
to
say
it,
but
adding
some
sort
of
automation
that
enables
us
to
allow
end
users
to
change
certain
labels
without
requiring
that
we
give
them
full
on
permissions
on
the
project.
F
F
R
I
mean
I
think
we
could.
We
could
create
Lee,
create
an
app
engine,
app
that
had
permit
had
the
right
permissions
on
get
that
only
allowed
community
members
that
you
know
used
your
get
used,
github
Oh
auth,
to
get
your
credentials
and
only
allowed
you
to
manipulate
labels
on
your
own
PRS
and
stuff
like
that,
and
only
certain
labels,
obviously
not
like
the
lgtm
label.
It's
not
a
hard
thing
to
do.
I
would
definitely.
This
is
definitely
one
of
those
like
PRS
welcome
kind
of
situations.
R
A
What
is
vision
for
the
project
and
how
do
we
appropriately
communicate
all
of
this
with
with
people
externally,
and
how
do
we
set
expectations
when
people
people
want
to
be
maintainer?
Zor
want
to
have
these
permissions
to
add
labels,
or
you
know,
work
their
way
up
the
the
contributor
hierarchy
for
lack
of
a
better
description
and
what
the
expectations
are
that
tie
with
that.
A
So
last
week
we
sent
you
all
home
with
homework
to
get
communities
that
do
this
well,
communities
that
you're
familiar
with
and
are
happy
with,
and
you
know,
would
love
even
if
it's
just
a
piece
of
the
way
they
do.
Something
and
I
saw
a
PHP
proposal,
a
link
in
Joe's
document
and
something
from
the
PHP
community.
I
know.
A
Sorry
not
hate
me,
but
it's
the
Python
something
it's
p,
something
p
right.
Peps.
I
A
A
We
saw
the
pups
process
in
documented
in
the
nerd
referenced
in
the
stock.
Are
there
other
communities
and
processes
for
getting
efforts
tracks
having
people
grow
through
community
grow
through
and
into
community
leadership
positions?
Who
does
this?
Well?
What
can
we
learn
from
because
you
all
did
your
homework
right?
A
R
I
E
I
Frustrating
for
the
community,
and
so
I
think
some
of
the
stuff
here
is
to
actually
like,
like
I,
think
you
know,
Quinn
says
Oh.
Something
that
happens
is
that
you
want
to
do
something
across
this
three
different.
You
know
components.
However,
we
define
components,
one
of
those
hasn't
signed
up
to
actually
do
it
in
the
release
right.
What
do
you
do
then?
Right?
It's
like
that's
a
tough
problem,
but
at
least,
if
you
see
that
earlier,
rather
than
later
you,
at
least
you
know,
aren't
going
to
be
surprised
after
you've
written
a
top
coat
yeah.
R
S
P
S
What
it
seeks
to
achieve
I
think
it
does
it
very
well
and
it
definitely
frustrates
some
developers
and
and
companies
sometimes
when
they
want
to
get
a
feature
in
and
the
process
as
it
is.
Today,
it's
evolved
over
about
six
years
to
be
a
lot
of
early
design
and
discussion,
and
if
you
don't,
if
you
were
writing
some
code
and
you
don't
go
participate
with
that
community
in
the
design
process,
if
you
just
show
up
with
code,
it's
typically
rejected
out
of
hand
whether
it's
good
or
not.
It's
because
you
didn't
participate
in
the
design.
S
S
D
D
Maybe
he
wants
to
write
a
Google
style
design
bakken
that
can
be
proceeded
as
heavy
actually
mission
kept,
not
as
a
governance
process,
but
just
like
you
can
have
a
successful
open
source
project
will
write
thoughtful
design
docs
as
a
prerequisite
for
for
coding,
and
my
personal
views
I'd
like
us
to
be
design
doc
first,
but
that's
got
to
be
something
a
cultural
choice
we
make
as
a
community
will
be
hard
to
do
sort
of
both
the
same
time.
That's
all
mm-hmm,
my.
A
S
When
I
things
you
in
OpenStack
that,
because
there
are
so
many
subprojects,
each
sub
product
is
its
own
community,
they
each
have
slight
variances
in
how
they
want
that
community
work.
So
there's
there's
an
overarching
technical
committee
that
sets
the
policies
for
all
of
the
projects.
I
can
point
to
that.
But
then
the
end
each
project
has
I
would
eat
the
nuances.
There's
significant
differences
in
how
much
design
or
the
the
templates
they
want
to
use
for
the
design,
the
kinds
of
decisions
that
need
to
get
teased
out.
I
E
I'm,
just
going
to
aside
from
code
reviews
and
design
reviews
what
I
like
about
both
the
Django
and
the
Python
community
is
their
inclusivity
and
their
work
towards
getting
new
beginners
enrolled
in
the
community.
That
you
know,
linux
kernel
is
what
it
is.
It
has
its
own
reputation
for
lacking
of
that
sometimes,
and
one
of
the
things
that
I
think
I
see
the
Cooper
Nettie's
is
you
know
the
more
complex
it
gets.
The
murder
it's
going
to
be
to
get
people
on
boarded
to
contribute
code
from
square
one
in
it.
I
I
J
E
I
A
A
Eric
wants
a
benevolent
dictator
for
life.
Aaron,
though,
asked
about
the
submit
Q,
and
this
is
why
I
want
to
give
us
a
couple
of
minutes.
The
Mickey
was
currently
blocked
and
do
we
have
any
update
for
anyone
on
the
call
that
can
update
the
plan,
the
path
of
the
pain
at
the
moment?
I
know
there
were
conversations
this
morning
about
the
the
testing
and
pushing
testing
pushing
against
testing
again
I,
don't
know
who's
on
the
call.
M
Q
Just
just
looking
at
Jenkins,
we
see
the
correlated
failures
across
many
runs
starting
at
about
8am
this
morning.
So
there
may
be
some
underlying
issues
with
GCE
that
need
to
be
debugged,
Jeff,
just
notice
that
he's
booked
up
for
the
day.
So
no
it's
been
paying
close
attention
this
morning,
but
that
will
change
very
shortly.
A
Q
R
And
to
be
clear,
the
right
way
to
fix
this
and
all
right,
maybe
go
for
it
is
go,
find
the
tests
that
are
marked
flaky
in
github
and
fix
those
issues
right
like
the
fact
that
the
bill
gets
broken,
is
representative
of
a
underlying
problems
in
the
code
that
we
should
fix,
right
and,
and
so
I
mean
sometimes
this
infrastructure.
Sometimes
that
providers
have
a
bad
day,
but
the
image.
R
M
One
thing
bremen:
if
you
see
like
it's
not
merging
because
they
have
a
flake
and
then
you
go
and
tell
kpop
to
test
it
again,
give
your
flake
and
eventually
things
pass.
But
right
now
we're
at
the
point
where,
like
you,
just
look
at
a
PR
and
all
things
pass
and
and
the
bots
not
doing
anything
right
so
I
mean
window.
R
I
mean
that
the
merge,
the
things
that
it
blocks
on
on
the
merge
q,
those
r
into
n
test,
runs
to
right.
We
run
continues
in
the
end
up
head,
in
addition
to
the
Indian
that
we
run
on
your
PR,
and
so
if
those
tests
fail,
that's
when
that's
when
they
go
red
and
those
tests
fail
either
because
you
know
we
have
we
either
we
leak
resources
and
we
run
out
of
resources.
L
R
Like
that's,
those
are
the
causes
and
some
of
those
are
not
fixable
by
you
or
item.
If
we
run
out
of
quota,
obviously
you
guys
can
go
in
and
find
the
resource
leaks,
but
you
don't
have
access
to
the
projects
that
you
need
in
order
to
delete
the
things
that
we're
leaking.
If
our
infrastructure
is
broken,
then,
like
we
can't
again
because
of
the
permissions,
you
can't
let
people
do
that,
but
flaking
tests
or
something
that
everybody
should
be
owning
concretely
right,
yeah
I
could
write.
J
Like
to
take
this
opportunity
to
mention
yet
again
that
a
distributed
test
dashboard
is
a
blocker
for
1.3
and
we'll
be
coming
order
of
soon
so
just
FYI.
We
have
heard
this
loud
and
clear,
and
the
ability
giving
people
the
ability
to
have
better
visibility
into
tests
and
running
their
own
tests
and
submitting
those
test
results
is
of
a
key
importance
that
we
are
working
on
now.
So.
F
F
Yes,
yeah,
okay,
so
I've,
just
I'm
trying
to
sort
of
close
the
loop
here,
because
you're
talking
about
something
that
I
haven't
actually
heard
about
a
roadmap
or
commitments
for
okay,
I
swear
people
are
interested
in
testing
specifically
federated
testing
in
the
display
of
test
results
seems
like
the
special
interest
group
about
testing
would
be
a
great
place
to
have
an
in-depth
discussion
about
that.
Yeah.
J
Absolutely
I
thought
he
had
already
presented
that,
but
if
not
or
regardless,
maybe
you
can
revisit
it
and
give
an
update
and
will
encourage
him
to
do
so.