►
From YouTube: Kubernetes 1.10 Release Burndown Meeting 20180320
Description
A
Already
welcome
everybody
to
the
kubernetes
1.10
release
team
burned
down
meeting,
and
it
is
Tuesday
March
20th
2018.
If
you
want
to
follow
along
after
the
fact,
the
agenda
and
what
we're
talking
about
it.
No
cheating
here
is
available
at
bit:
dot
Lee,
/,
gates,
110
burn
down
so
without
further
ado,
I'm
gonna
go
ahead
and
share
my
screen
and
we're
gonna
run
through
where
we're
at
and
process.
Here.
A
So
we
have
Jennifer
for
a
half-hour,
we'll
probably
do
the
quick
rundown
on
this
risk
assessment
and
then
we'll
skip
the
docs
and
then
won't
go
back
to
the
rate-limiting.
So
so
I
put
a
link
here.
Jordan
has
a
PR
which
I
would
love
for
him
to
talk
about,
and
ideally
for
us
as
a
release
team
to
come
to
consensus
on
so
without
further
ado
Jordan.
You
want
to
talk
about
that
pull
request.
B
It
doesn't
seem
like
a
very
valuable
feature
to
have
in
110,
even
in
alpha
state,
so
I
discussed
with
Klaus
Bobby
wasn't
online
at
the
time,
but
when
he
comes
online,
I'm
happy
to
catch
him
up
as
well.
We
discussed
just
neutralizing
the
feature
even
in
alpha
state,
and
so
these
are
the
three
places
that
the
feature
gate
is
checked,
checked
and
controller
and
it
is
tested
in
one
unit
test,
and
so
the
proposal
is
to
completely
remove
this
feature
from
110.
B
Klaus
was
amenable
to
that.
Given
the
the
current
issues,
and
so
this
has
the
LG
TM
from
him.
It's
already
picked
to
the
110
branch,
because
we
fast
forwarded
and
lifted
freeze
and
so
we're
getting
green
CI
right
now,
but
the
goal
is
to
unblock
and
stabilize
the
110
release
and
get
signal
back
on
the
CI
jobs
we
care
about.
B
A
B
Given
that
the
all
the
issues
we've
seen
related
to
this
have
been
in
the
Alpha
job,
which
is
what
we
would
expect
for
a
feature
that
is
only
on
when
this
gate
is
turned
on,
I
am
pretty
comfortable,
saying
the
issues
are
limited
to
the
code
guarded
by
this
gate.
So
this
PR
is
not
a
revert.
It
is
a
a
neutering
of
of
the
code,
and
so
this
was
as
targeted
as
I
think
we
can
make
it.
B
E
Can
speak
to
Docs
because
the
PR
for
the
dock,
for
this
feature
has
been
a
little
difficult
to
wrangle,
also
possibly
reflecting
some
of
the
issues
with
the
implementation,
so
I
called
it
out
yesterday,
in
fact,
and
as
Jason
pointed
out
in
the
slack
channel,
Jordans
PR
would
moot
the
docks.
Pr
and
I
am
ready
and
waiting
to
close
it
unmerged.
E
A
B
B
E
A
It's
it's
not
in
the
blog,
but
maybe
somewhere
else,
but
I
think
I
think
we're
safe
I
mean
it's
the
one
dog
PR,
the
the
thing,
just
as
a
release
manager.
If
I
put
that
hat
on
deeply
for
a
moment
having
cruft
code
sitting
out
there,
that's
basically
dormant
and
not
accessible,
it
makes
me
always
makes
me
nervous,
but
given
that
there's
no
way
to
cleanly
revert
the
PRS
themselves
we
just
have
to
whoever
is
leading
111
needs
to
lean
into
the
scheduling
sake
and
really
get
them
to
clean
this
up
and
fix
his
stuff.
G
A
Well,
it
sounds
like
sounds
like
disabling
this
feature
for
110
makes
a
lot
of
sense
for
getting
the
test
back
green
for
eliminating
any
possible
churn
in
the
community
with
people
trying
to
use
this
thing
and
having
it
blow
up
their
clusters,
so
I'm
in
favor
of
the
the
PR
getting
merged.
Anybody
not
know.
C
And
I'll
say
based
on
my
discussion,
my
meeting
with
six
storage.
They
were
ok
to
go
ahead
with
the
feature
based
on
the
idea
that
it
was
completely
ready
to
go
alright
Wow
last
Thursday
and
and
obviously
that's
not
the
case
so
I'm
sure
they're
gonna
be
okay
with
reverting
on
that
on
the
basis
that
there's
little
bugs
Technic
scheduling,
I
mean
you're
gonna,
give
them
that
suit
storage.
Sorry,
okay,.
A
Okay,
all
right,
then,
let's
well,
these
are
the
two
PRS
that
are
actually
out
there
right
now
in
the
Winton
milestone
the
PR
and
the
cherry
pick.
So
we
will
get
these
through
and
Jordan.
Thank
you.
This
is
yours.
You've
been
tremendously
helpful
here,
honestly,
I,
don't
know
what
we
would
do
without
you
in
this
instance.
Thank
you
all
right,
so
I'm
running
the
in
the
notes
that
this
is
approved
by
release
team
and
it's
recorded
to
video.
A
E
You-
and
this
is
quick
I
added
content
before
I
realized
I
was
free
for
this
half
hour
and
so
just
to
follow
up.
You
know
we
just
discussed
one
of
those
items
and
I
hope
by
tomorrow.
What
I'm
gonna
do
I
bet
that
I
want
an.
I
downgrade
issue
at
this
point
is
merged
the
yet
they're,
currently
not
really
quite
right.
Pr,
it's
technically.
Okay,
it's
just
not
really
user-friendly
and
then
submit
another
PR
for
review
to
make
sure
that
I
haven't
broken
anything
technically
I'm.
E
E
A
E
E
A
A
C
B
C
C
A
Okay,
this
and
my
brains
train
to
mush
with
all
these
PRS
okay.
So
this
this
got
merged.
That's
good,
I'm,
glad
that
we
got
that
all
squared
away.
So
that's
very
documented
in
the
prior
meeting,
okay
I'm
gonna
write
an
email
after
the
meeting
today.
Just
everybody
know
in
coordinating
step,
hey
it's.
D
A
A
H
A
C
C
C
A
C
B
Think
it's
more
a
matter
of
these
are
the
scalability
targets
and
what
we
previously
achieved
and
deciding
at
the
last
minute
to
say
well
we'll
ship,
even
if
we've
surpassed
scalability
targets
like
this,
is
very
reminiscent
of
conversations
with
the
1/8
time
for
him
and
that,
like
moving
the
goalposts
at
the
last
minute,
is
not
something
we
should
do
lightly.
I.
A
Agree
you're
exactly
right:
this
is
identical
to
what
we
did
it
1/8.
It's
like
deja
vu
all
over
again
the
thing
that
I
the
thing
that
I
want
to
understand,
though,
is
I
mean
there.
There
are
certain
cases
where
there
will
be
increased
latency
and
it
cannot
be
avoided.
I
mean
operations.
Argit's
are
set
given
a
certain
set
of
assumptions,
and
if
those
underlying
assumptions
change
in
a
way,
that's
not
not
retro
actively
actionable
and
we
we
have
to
adjust
those
expectations.
I
mean
right.
B
I
think
the
difference
is
we
need
to
understand.
Why,
right?
Why
would
why
this
won't
happen?
So
in
1:8
we
saw
a
similar
bump.
We
figured
out
what
the
cause
was
and
because
we
knew
the
cause
and
we
knew
the
impact
we
were
able
to
make
a
more
informed
decision
making
decision
here
that
we're
okay
with
missing
scalability
targets
without
knowing
the
underlying
cause
doesn't
seem
like
a
good
idea.
No.
A
I
and
I'm
absolutely
not
advocating
for
that
in
any
way
what
I'm
advocating
for
it.
So
we
escalate
this
to
try
and
get
more
eyes
on
this
or
what,
if
we
need
to
do
to
figure
that
out
yeah
and
the
last
resort,
as
as
it
wasn't.
One
eight
because
remember
I
was
a
very
strong
proponent
of
holding
off
the
release,
because
that
last
time
the
last
resort
is
is
if
we
have
to
do
some
sort
of
documentation
or
whatever.
That
is
to
do
that.
So
that's
that's
planet
like
plan
R
or
Plan
F.
A
A
C
The
and
then
the
other
part
of
it
is
to
you
know,
frankly,
throw
this
out
there
in
a
common
forum
on
death
in
the
community
meeting
thing:
hey
we're
just
gonna,
keep
delaying
the
release
until
we
know
what's
causing
us.
So
if
you
care
about
kubernetes
continuing
to
release
and
maybe
a
little
help,
yeah,
that's.
H
A
Yeah
does
I
mean,
is
this
I
don't
understand
enough
about
that?
The
regression
to
just
speculate,
but
is
it?
Is
this
something
where
we
hit
a
tipping
point
at
a
certain
number
of
nodes,
or
is
this
something
that's
an
anchor
did
across
the
nodes,
or
you
know
like
latency
associated
with
writes
a
net
CD
or
something
like
that
that
are
that's,
causing
it
to
slow
down
or
where,
if
we
identified
as
a
contention
or
a
cliff
or
what.
B
The
patterns
were
saying,
don't
point
at
a
contention
or
rate
limiting.
If
you
look
at
the
graph
like
there's
a
there's,
a
range
of
variability
over
times,
but
what
we
see
is
a
bump
in
the
overall
times
and
so
on.
Much
smaller
clusters
like
the
variability
still
exists
and
the
bump,
so
the
variability
is
smaller
in
the
bump,
is
smaller
and
I
think
the
variability
starts
to
overwhelm
the
bump.
It
makes
it
really
hard
to
figure
out
the
the
signal.
B
B
A
F
B
In
the
meantime,
while
that
was
in
apparently
something
else
came
in
as
well,
and
that's
what
they're
trying
to
track
down.
So
it's
it's
overlapping
regressions.
So
you
can't
just
look
at
when
the
bump
happened
and
say
well,
the
problem
must
be
there.
Let's
revert
that
that
happened
already
and
now
they're
bisecting
again
to
try
to
figure
out
when
a
different
regression
came
in
yeah.
H
H
H
H
So
so
that
was
one
the
other
one.
Is
we
probably
need
to
request
Wojtek
to
help
I
haven't
seen
him
active
in
this
er?
Maybe
he's
helping
behind
the
scenes
he
shot.
The
shaman
Wojtek
I
haven't
seen
them
pick
up
from
Andrews,
so
that
might
be
another
thing
to
do
because
they
have
been
running
this
for
a
really
long
time.
Now,.
A
G
I
was
just
looking
on
just
there,
the
cigs
info
that's
published
and
we're
seeing
that
they
say
99%
of
POD
start
within
5
seconds
and
depending
on
where
we're
looking
and
some
of
these
bad
ones
are
coming
up
above
6
and
the
good
ones
are
below
4.
So
it's
as
much
as
a
50%
decrease
at
this
scale,
which
is
pretty
significant
and
it
it's
definitely
past
what
they
say
is
their
number
and
it's
not
like.
They
clarify
that.
This
is
only
a
number
insert
context
like
that's
appears
to
be
a
broad
goal.
H
G
H
H
C
C
J
H
H
C
D
Jiki
looks
like
it
cannot
bring
up
clusters
yeah
I'm
nagging
some
like
internal
folks,
see
if
they
can
find,
and
we.
H
D
H
H
C
H
A
A
H
A
D
C
C
That
should
be
getting
close
now
that
we've
made
a
decision
on
the
PVC
stuff,
but
in
take
a
look
at
it,
building
that
got
recently
opened
not
a
blocker,
but
we
found
yet
another
bug
in
the
sub
path:
security
fixes
Council
and
the
P
for
PR.
For
that
comes
through
before
we
release
110,
we
probably
want
to
include
it
in
one
fix
in
110.
Sorry,.
B
C
C
B
The
cherry-pick
is
just
pending
milestone
or
branch
approval,
so
I,
guess
that
is
to
see
them.
Caleb
yep,
so
I
will
assign
him
and
the
other
one
is
pending
a
green
cops
test
which
was
flipping
for
a
little
while
on
a
bunch
of
yours.
A
D
A
A
A
A
I
Lee
so
the
last
meeting
we
pegged
a
few
of
the
same
networking,
no
dates.
Why
store
visit
in
there
just
to
make
some
reviews
and
some
addition,
if
you
haven't
gone
in
already
you're
part
of
this
group
and
or
you're
working
on
a
feature
that
is
important
to
communicate
through
the
release
blog,
please
go
in
there
today.
I
You
are
a
lot
of
work
on
it
yesterday
as
well,
and
then
I
sent
it
over
to
Aparna
and
Brian
grant
this
morning,
just
as
a
FYI
here's
the
progress
we've
made
on
the
blog,
please
give
it
a
review,
and
let
us
know
if
you
have
any
feedback,
but
there's
any
changes.
What
flag?
You
know
things
like
that
for
the
blog,
so
in
terms
of
next
step
seems
like
there
are
still
quite
a
few
things
that
we're.
A
I
A
I
Know
what
we'll
does
people
is
you
coordinating
with
everybody,
making
sure
that
everybody
is
up
to
date
and
on
the
same
page,
and
that
no
one
is
getting
any
surprises
on
if
the
release
pushes
and
or
features
are
not
included,
we're
just
making
sure
to
just
overly
communicate
this
time.
Just
because
a
lot
of
changes
have
happened.
A
And
we've
brought
up
this
key
questions.
We
need
to
answer
this
still
needs
some
attention.
I'm
gonna,
try
and
give
this
some
attention
today,
and
hopefully
others
can
do
so
as
well.
Having
the
input
into
the
blog
from
networking
and
storage
is
incredibly
helpful,
though
I
mean
in
some
ways,
storage
is
really
taking
the
the
spotlight
this
time,
which
is
great
because
that's
one
of
the
the
most
vexing
problems
for
cluster
operators
for
sure.
A
I
And
just
to
go
back
to
the
document
that
you
are
on
those
answers,
don't
have
to
be
robust,
you
don't
have
to
feel
like
you're,
you
know,
you're
writing
several
paragraphs,
I
mean
even
a
couple.
Sentences
are
completely
fine.
This
is
just
these
are
questions.
We
get
every
single
release
and
if
we're
going
to
be
limited
on
time
from
when
we
can
send
things
to
media,
to
win,
the
release
actually
publishes
I
want
to
make
sure
that
we're
pre
answering
these
questions
for
them.
So
that
way,
our
response
is
making
it
to
the
story.
Okay,.
A
A
I
F
G
F
Killed
Noah
yeah
he's
been
dying
for
the
last
little
while
anyway
anyway,
so
he
he
ran
another
poll
last
night,
just
get
the
sort
of
remaining
things
that
were
committed
after
we
did
the
last
poll,
so
we've
got
about
22
new
issues
down
at
the
bottom
to
run
through,
but
that
shouldn't
take
very
long.
So
we
have
been
hearing
back
from
most
of
the
people
that
we've
asked
for
clarification
from,
but
there
are
still
people
that
we
haven't
heard
from,
but
nothing
is
really
a
showstopper
everything.
That's
in
here.
You
know.