►
From YouTube: 2017-06-20 17.02.31 SIG-cluster-lifecycle 166836624
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
B
Blockers
sure
yeah,
so
there
are
three
open
issues.
There
were
two
last
night,
but
one
PR
got
reverted
so
now
we're
back
to
three.
There
are
two
issues
in
the
main
repository,
the
first
of
which
is
about
the
IP
mask
agent
I'm
chatting
with
Daniel
about
that
right
now,
I
get
that
fixed
guessing
at
all,
get
fixed
today.
A
second
one
is
an
upgrade
test,
failure
which
a
number
of
people
are
working
on
done,
something's
going
on
that
yesterday,
I'm
not
too
sure
about
that
one.
B
The
third
one
is
in
the
Cuban
Rico,
which
is
her
documentation
for
operating
it
on
self
post-acute
I'm,
a
cluster
or
I
just
want
to
highlight
the
fact
that
these
things
are
being
tracked
by
the
release
team.
The
released
candidate
is
scheduled
to
be
cut
tomorrow
to
play
first
in
the
final
build
is
scheduled
for
next
Wednesday.
It's
not
clear
to
me
whether
or
not
we're
on
track.
There
was
some
discussion
yesterday
about
postponing
that
slightly,
but
we
do
need
to
make
sure
we
burn
down
our
issues
here.
B
C
So
there's
a
peer
out
for
this
I've
done
a
bit
of
feedback
and
addressed
it.
I'm
really
looking
for
more
people
to
actually
test
the
instructions,
so
I've
created
pre-release,
rpms
and
Deb's,
and
they're
they're,
hosting
Google,
Cloud
storage
of
instructions
on
how
to
fetch
them
so
Rio,
cross,
different
distributions
and
maybe
different
point
releases
of
1/6
it'd,
be
great.
D
B
D
E
D
A
D
E
D
C
Yeah
and
I've
been
thinking
about
hosting
a
session
of
anyone,
who's
interested
in
learning
just
to
get
that
a
little
bit
cheaper
from
my
time,
Christian
if
I
can
brew,
Q&A
and
Code
walkthrough,
and
things
like
that.
A
lot
of
that,
though,
is
that
the
test
info
keeps
changing.
They
keep
putting
more
layers,
they
keep
refactoring
everything
and
changing
things
to
scenarios
to
other
files
that,
yes,
like
I,
can
barely
keep
up
with
it
as
well.
Every
time
they
break
I
have
to
go
in
and
relearn
everything
from
scratch.
C
D
And
I
think
that
was
the
problem
like
love.
The
the
latest
week
have
been
exciting
because
of
that
three
refactoring
yeah.
So
but
basically
we
have
a
couple
of
issues
in
the
Cuban
people
that
would
be
cool
to
get
take
a
pre-summit
job
Jacob
a
superior
for
that
one
which
I
think
is
going
to
be
much
today.
Oh
yeah.
C
C
D
I
think
we
could
initially
have
it
falls
in
DQ,
venom
and
I
mean
I,
expect
this
cycle
to
have
pretty
much
a
couple
of
POS
every
at
any
given
point
in
time.
So
if
anything
else
are
like,
if
it's
starting
to
block
that
by
when
when
if
someone
else
makes
makes
a
change
that
breaks
cubed
M,
but
it
wasn't
noticed
in
the
pre
submitter
well
at
least
notice
them
in
our
cube
am
related
pre-submitted
and
that
way
get
notified
about
it.
D
So
I
think
it
makes
sense
to
at
least
enable
gave
them
for
for
only
RPF
initially
listen.
Then
we
have
to
test
the
version
once
it
control/play
from
from
the
Cuban
one
7
binary
and
let
PR
for
that.
One
too,
thank
you,
then
use
images
from
head
instead
of
from
latest
exe.
D
So
basically,
now
we're
using
the
latest
alpha
or
beta
version
of
cabañas.
That's
released
the
this
means
we
are
behind
at
least
or
maximum
two
and
a
half
weeks,
something
from
from
the
from
master
and
that's
not
optimal.
It
would
be
great
to
use
the
latest
images
and,
let's
Robert,
said
and
pointed
out
in
the
issue.
It's
a
long-standing
thing:
I
create
an
issue
in
December
forest,
but
there
has
been
enough
resources
like
human
working
resources
to
implement
it.
D
Yet
so
I
mean
that
would
be
an
awesome
thing
to
contribute
from
from
all
side
as
well.
It
isn't
that
hard
I
mean
I,
checked,
check
the
code
and
maybe
30
40
lines
well-placed
but
anyway,
and
that's
what
but
somebody's
gotta
take
on,
and
so
that
would
be
cool.
Then
we
should
verify
those
notarized
it's
working.
D
F
Hi,
so
is
the
node
authorized
are
not
already
on
for
the
e
de
alpha
channels,
I'm,
not
sure
I'm,
good
MIT,
it
would
be
really
nice
of
it.
It
is
yeah.
I
can
go
and
investigate
that.
I
think
that
the
node
authorizer
work
was
blocked
on
the
TLS
part
of
the
TLS
bootstrapping
stuff
being
on
by
default
for
the
EB
tests,
because
you
need
to
have
sort
of
unique
credentials
on
each
node
in
order
for
they
have
no
dollar
I
should
to
work.
D
F
D
D
F
D
D
Well,
we
we're
testing
manually
like
Jacob
me
and
others,
but
just
just
having
something
to
pin
up
a
key
bed
and
faster
v16.
Then
just
do
the
same
thing
for
just
run.
This
Q
beta
minute
command
with
one
seven
one,
seven
five
and
then
run
conformance
that
that's
like
enough,
but
it
great
to
get
it
automated
as
soon
as
possible,
but
might
be
hard.
Then
we
have
an
easier
one
that
we
should
make
a
ticket
lifecycle,
tab
or
cubed
M
tab
or
something
I,
don't
know
just
in
the
last
cops
have
already
its
own
or.
D
G
B
Sort
of
completed
issue
which
looks
like
we're
chatting
about
and
one
of
the
issues
which
is
it's
one
thing
to
be:
sort
of
use
the
deployment
tool
to
run
the
various
edu
tests
against
the
provider
like
you're,
using
cops
from
the
US
test.
But
we
also,
if
we
make
like
a
cube
admin
tab.
We
want
to
use
that
to
test
variations
of
cube
admin
itself,
not
just
to
run
tests
on
a
provider
using
cube
admin
right,
so
those
are
sort
of
two
different
variants.
B
I
think
we
have
a
cube
admin
tab,
it's
really
to
test
the
variation
in
cube
admin
and
like
we
might
use
cube
admin
or
costs,
or
something
to
run
the
tests
against
other
platforms,
and
then
I've
been
chatting
with
Eric
theta
from
the
inspiron
team
about
making
like
a
dashboard
for
our
sig,
so
like
a
sig
plus
or
lifecycle
dashboard.
Because
right
now,
if
you
look
at
test
grid,
it's
sort
of
a
this
huge
mingled
mess
like
you,
don't
even
know
where
to
start.
B
You
don't
know
where
to
look
for
tests
and
if
each
sig
had
a
place
where,
like
here's,
your
landing
page,
you
look
here.
This
is
where
you
triage
test
failures.
It
would
be
really
really
helpful,
so
this
is
sort
of
conflated
with
that
I
think
it's
like.
We
should
definitely
just
make
a
cube
admin
tab
and
get
ourselves
out
of
misc,
because
it
always
takes
me
like
five
minutes
to
remember
that
render
misc
and
where
to
find
our
tests,
but
sort
of
in
a
longer
term.
B
All
right
so
I
wanted
to
give
lighting
last
week
during
the
two-day
meeting,
which
was
helpful
to
do
like
a
very
brief
recap
of
the
breakout
meeting
that
we
had
happen
in
the
Friday
before
people
can
weren't
able
to
attend.
So
I
was
going
to
do
that
again,
trying
to
keep
it
under
about
five
minutes
or
so
for
people
who
weren't
there
and
don't
want
or
don't
have
time
to
watch
the
video
one
thing
I
will
start
with.
B
Is
we
forgot
to
record
attendance,
which
Lucas
pointed
out
so
I
threw
some
names
in
there
of
people.
I
knew
who
are
there,
but
if
you
were,
if
you
were
there
on
Friday,
please
go
back
and
add
your
name
retroactively.
So
in
terms
what
we
talked
about,
we
talked
about,
we
started
with
a
doc
that
Bobby
runs
I
put
together
about
sort
of
which
configurations
of
cube
Babbitt.
B
We
want
to
support
whether
it's
single
node
multi
node,
with
a
single
master
or
multi
node,
with
multiple
masters
and
sort
of
talk
through
different
iterations
of
grade
paths
from
each
of
those
into
point
pass
and
where
we
want
to
start
that
dock
has
been
made
comment
about
for
everyone.
So
I
was
working
through
it
again
last
night
and
was
for
sure
that
I
couldn't
add
comments,
but
I'm
going
to
do
that
now,
so
people
can
go.
Take
a
look
at
that.
We
also
talked
about
daemon
sets
versus
deployments
for
master
components.
B
Core
OS
uses
deployments
mainly
to
support
single
node
self,
hosted
bastards,
and
we
talked
about
asking
the
cig
apps
group
to
create
like
a
rollout
policy
for
day
one
set,
so
we
can
change
the
rolling
configuration.
The
main
reason
for
using
deployments
is
that
you
can
have
two
deployments
of
the
control
two
instances
of
like
a
controller
manager,
pod
on
the
single
node
master
or
the
daemon
set
you'll
kill
yourself
and
never
recreate
yourself.
B
We
talked
about
checkpointing
and
sort
of
came
to
a
nominal
agreement
that
we'd
like
to
ask
sake,
node
to
add
checkpointing
either.
You
know
probably
from
their
side
doing
reviews
and
for
us
implementing
a
via
an
alpha
annotation
on
pods,
so
you
hadn't
helped
annotation
to
the
master
component.
Pods
I
knows
we
get
check
pointed
by
the
cubelet,
so
we
don't
have
to
run
and
out-of-band
check
pointer
core
OS
pointed
out,
but
this
is
only
useful
for
them.
If
we
also
checkpoint
secrets,
so
I
think
we
will
try
to
push
that.
B
Also,
we
may
make
that
a
separate
annotation
or
we
may
sort
of
bundle
that
with
the
Alpha
annotation
and
revisit
later,
but
I
think
for
the
initial
version
will
definitely
want
to
check
voice
secrets.
So
we
can
get
everybody
on
the
same
page
using
the
same
trick
home
there
and
then
so.
As
a
result,
there
were
a
number
of
action
items
that
came
at
the
end.
B
If
you
look
at
the
meeting
notes,
I'm
going
to
talk
to
say,
gaps
about
the
diamond
set,
upgrades
Tennessee
is
going
to
start
creating
the
checkpoint
proposal,
which
are
bunch
of
people,
are
going
to
jump
in
and
flush
out.
So
we
can
give
the
signal
and
then
Lucas
is
upgraded.
The
checklist
for
self-hosted
q
got
them
sweet.
That
was
like
two
and
a
half
minutes.
So
definitely
under
the
fagner
anything
else.
People
do
were
there
when
I
try,
min-width.
A
D
Yeah,
so
this
might
be
a
topic
for
next
week
in
in
more
detail
but
Joe
Nathan,
so
DC,
McCoy
and
me
and
Eric
talked
about
bootstrap
tokens
and
a
security
flaw
with
man-in-the-middle
that
possible
to
do
now.
If
you
have,
if
one
token
is
compromised,
you
can
set
up
a
man-in-the-middle
api
server
kind
of
thing
that
makes
you
so
the
client,
when
you
cube
areum,
join
you
can
unfold
it's
possible
to
join
a
master
that
isn't
secure.
D
That's
someone
someone
else's
set
up,
so
that's
why
it's
good
to
be
in
alpha,
so
to
not
not
have
that
that's
security
for
in
it,
but
basically
on
in
sigil
the
gold
Fleck
Channel
we're
discussing
how
to
fix
this.
If
you
have
to
litigate
it,
it's
not
a
huge
thing
it.
Basically,
you
have
to
get
one
token
have
to
be
compromised
and
then
you
know
can
top
things,
but
still
we
want
to
lock
working
that
down,
even
if
a
token
was
compromised
at
some
point.
D
So
and
probably
it's
going
to
be
varying
like
different
tokens.
If
you
have
a
live
a
couple
of
alternatives,
so
you
can
use
a
really
secure
token,
which
is
longer
or
you
use
less
secure,
something
which
is
shorter
or
you
just
copy
the
file
out
of
band
so
which
is
the
most
secure
alternative
of
course.
So
there
will
be
a
couple
of
alternatives,
but
I
think
we'll
go
into
this
in
more
detail
next
week
when
I
think
Joe
is
on
a
plane,
it
said,
and
the
time
couldn't
make
it
either
so
yeah.
B
Other
folder
we
can
do
which
consider
that
we've
been
looking
into
on
the
GCE
side
and
I
know
the
AWS
supports.
The
same
thing
is
using
like
the
assigning
attributes
of
the
virtual
machine
infrastructure
to
do
this
as
well
right,
instead
of
pre
distributing
a
shared
key,
we
can
use
for
the
PKI
infrastructure
built
into
the
underlying
cloud
platform
to
prove
identity.
Also,.
G
G
So
it's
like
the
race
to
see
who
gets
it
first
and
the
theory
is
the
init
process
will
get
it
first,
because
then
anyone
else
can
also
read
it,
but
it
will
be
worthless
after
first
use
interesting.
So
it's
sort
of
a
one-time
use
things.
What
are
some?
That's,
how
that's
why
they
make
it
tolerably
secure
in
volt,
2,
volt
standards,
okay,.
G
D
So
the
the
problem
may
be:
oh
I
guess
we
should
just
write
yet
another
doc,
for
that
is
more
cubed.
M
specific.
This
is
proposal
for
pocus.
It
is
more
more
for
the
validation
of
the
master
from
the
node
side
and
not
as
much
the
TLS
bootstrapping,
which
is
like
when
I
actually
went,
Achatz
and
authenticate
to
the
master
and
get
my
things
signed,
so
we're
using
both
TLS
bootstrapping
we're
using
the
same
token
for
both
the
like
bootstrap
talk
and
validation
flow
and
TLS
food
shopping,
so
yeah
I
know
yeah.
D
F
D
D
D
D
We
have
to
somehow
generate
a
serving
certificate
for
the
new
API
server
will
go
and
which
includes
the
right
host,
name
and
IP
address
or
whatever
that
that
it
has
to
include.
So
then
we
have
to
get
some
mechanisms
to
get
that
to
sign
that
certificate
and
I
distributed
the
right
place.
So
one
possible
solution
would
be
to
basically
use
the
same
mechanism
for
for
API
service
as
we
use
for
the
cubelet
and
maybe
just.
D
The
CSR
approver
right
now,
I
think
it's
a
diamond,
dynamic,
dynamic
rule
that
you
can,
if
you
add
the
true,
it
will
approve
all
or
see
a
thousand
specific
kind,
and
that
way
we
could
get
automatic
API
services,
which
is
individual
and
maybe
use
the
certificate
rotation
stuff.
Jacob
Simpson
is
working
for
the
qubit
reuse,
some
of
that
before,
but
I,
don't
know
what
what
it's
it's
something
we
could
think
about,
or
and.
F
So
we're
we're
considering
using
sort
of
a
similar
policy
for
cephalus
it
FTD
to
ensure
that
they
have
ensure
that
they
have
unique
to
life
certificates
when
they
go
ahead
and
talk
to
one
another.
So
that's
the
pure
communications
inside
of
the
self
of
the
cluster.
For
that
to
be
the
FTD
cluster
created
by
the
FTD
operator
that
is
sort
of
self
hosted
and
running,
everything
is
properly
secured.
F
H
On
a
pretty
bad
internet
connection,
but
that
sounds
very
reasonable
to
me.
I
think:
there's
talk
about
pot
identity,
so
we
should
once
that
is
developed.
We
should
try
to
be
a
connection
extension
of
pot
identity.
When
we
authenticate
against
API
servers
s.
It
is
slightly
different
than
the
API
service
club,
because
the
API
servers
are
behind
a
service,
but
that
would
that's
all
something
we
have
to
work
it
worked
out.
D
Yeah
thanks,
then,
if
it
sounds
at
least
somehow,
within
a
with
a
scope
for
I,
think
I'll
go
ahead
and
write
something
a
little
bit
longer
than
three
sentences
and
we
can
work
on
it
across
cross-cutting
effort
between
to
go
up
and
pick
up.
Lifecycle
I
think
it's
pretty
easily
achievable
but
well
see.
D
See
ya
new
contribution
ideas
for
the
week,
I
thought
as
part
of
the
the
contrib
X
thing
we
talked
about
last
week.
It
would
be
cool
to
highlight
a
couple
like
from
three
to
seven
things
or
something
on
each
meeting.
That
would
be
helpful
for
us
that
we
don't
necessarily
have
the
resources
to
implement
for
for
the
coming
two
weeks
or
something
that
would
be
really
helpful
tastic.
I
D
Exactly
yeah,
so
they
are
mocked
helpful
them
did
you
know,
because
we
had
some
looking
better
as
a
quick
comment
on
them.
Cubed
invalidate
would
be
cool.
To
have
something
I
mean
running
idiots
right
now
or
just
market
is
a
pain
you
have
to
download
things
from
from
GTA
the
GPS
buckets
somewhere
will
compile
yourself.
Then
you
have
to
setup
some
different
credentials
stuff
and
and
invoke
the
large
binary
with
five
different
command
line
flags.
Otherwise
it
won't
work.
D
D
This
DNS
resolution
doesn't
work
with
flannel
IP
mass
configuration
doesn't
work
which
I
so
recently
today
and
and
then
we
could
spot
thesis
to
what's
actually
the
problem
with
such
a
command
and
have
like
five
five
different
and
then
we
could
have
a
both
mode
or
something
the
circulating
actually
set
or
informs
us
what's
the
problem.
What
the
problem
is
is
that
over
line
when
the
user
report
to
deprive
that
information
would
probably
help
our
issue
again
as
well,
get
the
things
try
as
possible
if.
D
I
I
D
Yeah
so
I
think
this
is.
This
is
lightweight
yeah
for
sure
it
takes
3-4
seconds
to
to
do
I
think
what
you're
talking
about
was
something
heavier
that
takes
at
least
like
five
minutes.
Ten
minutes
I,
don't
know
I.
I
D
So
yeah
that
would
be
a
good
contribution
of
well,
probably
so,
first
hash
out,
which
which
are
the
real
1
1
and
then
implement.
Yes,
a
qubit
invalidate
we
can
have
cubed
invalidate
know
that
well,
then
work
all
of
these
node
validation
tests.
From
that
it's
a
sig
output
from
signal
then
cubed
M
verifies.
This
is
mostly
about
making
preflight
checks
in
vocables
the
first
way,
so
it's
basically
adding
a
command
and
hooking
that
command
up
to
the
pre-flight
package,
then
proof
I
check
for
bad
EDC
records
wrestle
cone
there.
D
There
have
been
a
lot
of
users
that
have
reported
issues
with
qbm
recently
that
wrote
where
they
have
local
health
defined
in
EDC
health
or
a
falcon.
So
something
like
a
strange
IP
somewhere
that
is
coming
from
the
VPS
provider,
something
I
don't
know,
and
in
that
case
cubed
a
21-6
fail,
because
it's
trying
to
resolve
that.
Oh,
it's
trying
to
call
that
IP
somewhere
that
isn't
reachable
and
Q
beta
minute
tanks,
so
that.
H
D
D
I
B
D
I
think
or
I
think
we
can
work
around,
even
if
they
don't
sign
up
for
it.
The
only
thing
I
don't
know
if
it's
more
fragile
or
not,
but
we
hypothetically,
we
could
detect
the
single
master
or
single
node
case
and
then
just
spawn
a
deployment
with
controller
controller
manager
that
would
like,
for
a
very
short
period
in
time.
Do
the
right
scheduling,
controller
manager
run
the
controller
mention
routes
and
then
remove
it
right
after
the
demon
set
up
greater
I
mean
yeah.
D
Look
we
could
work
around
it
in
the
cubed.
M
upgrade
come
on
Oh
like
controller,
that's.
Does
it
even
but
I
think
it
will
be
possible
for
things
to
stick
up
stir
to
do
it,
because
it's
generally
a
feature
I.
A
few
people
would
like
to
have
so,
but
in
in,
like
worst
case
scenario,
I
think
we
will
be
able
to
just
work
around
it
pretty
easily.
So.
D
G
D
D
D
It
gets
good
idea,
then
another
one
that
has
been
discussed
back
and
forth
if
using
secrets
for
storing
the
certificate
without
pasta,
pasta,
pasta
and
this
mean
between
using
secret
or
some
custom,
CTR
method,
but
I'm
I
think
we
should
use
secret
and
a
child
pointed
out
well
because
it's
cool
and
a
native
feature.
Otherwise
we
have
to
hack
around
with
I,
don't
know,
tempis
s
mounting
and
a
lot
of
such
such
things.
But,
as
joe
said
somewhere
of
like
maybe
on
an
issue,
we
might
yet
have
have
a
tepee
our
custom
type
for
women.
D
Basically
well,
the
drawback
with
with
the
current
secret
approach.
Is
that
all
ingress
controllers,
or
at
least
the
most
of
them
just
bind
closer,
always
get
secret,
which
means
they'll
have
access
to
everything
which
is
not
ideal.
You
can
just
pack
your
ingress
controller
and
you
have
root
access,
the
hole,
the
hole,
cutter
or
Federation
controller.
Also,
this
has
access
to
all
secret
and
whatever
use
applications
does
not
care
to
use
a
role
and
be
like.
D
Let's
you
think
about
what
namespace
I'm
going
to
use
or
such
things
can
and
I
mean
it's
totally
possible
that
they
just
have
a
class.
The
role
with
all
the
resources
and
I
want
to
get
secret.
Okay,
then
it's
in
my
my
consecration
and
suddenly
I
have
access
to
everything
which
is
what
we
want.
So
so
that's
a
drawback,
but
it's
I
think
the
little
latest
discussions
around
this
first
lean
towards
using
that
method
anyway,
and
then
maybe.
B
I
There
are
benefits
to
using
TBR's
from
a
security
perspective
because
you
can
quarantine
who
has
access
to
that
area
a
little
bit
better
than
coop
system,
because
there
would
be
less
people
accessing
it
or
less
actors,
but
the
there's
still
the
there's
still
the
general
discomfort
that
exists,
but
I
think
Lucas
and
I
chatted
on
slack
saying
that
we
can
always
upgrade
into
an
official
strategy
if
we
want
to,
and
this
for
the
time
being,
is
no
worse
than
what
we
are
doing
today
is
slightly
worse.
I.
Guess
you
quickly
a
security
person
but
I.
G
Mean
it
feels
like
this
is
the
next
thing
that
will
get
fixed
right.
The
English
controller
should
not
have
access
to
every
secret
in
the
system.
It
should
only
have
access
to
the
ones
that
have
been
tagged
as
like
certs
and
keys,
and
that
will
get
fixed
I
am
sure
and
I,
don't
think
we
should
work
around
someone's
going
to
get
fixed
right.
G
I
We
should
probably
create,
like
a
checklist
that
says,
quote
unquote
production-ready,
because
we
have
to
put
that
clause
at
the
top
of
Canadian
and
eventually,
once
we
have
oxy,
Goths
and
other
folks
have
a
answer
that
they'd
like
to
pursue.
You
know
we
could
eventually
have
that
as
part
of
the
checklist
I
think
you
know,
we
could
always
put
the
disclaimer
that
says
at
the
beginning
of
a
DM
that,
like
you
know,
these
are
the
things
that
we
use
are
known,
knowns
that
exist
and
go
with
God
on
the
other
stuff
yeah.
B
I
wouldn't
say
it's
not
production-ready
I
think
it's
depending
on
what
you're
using
it
for
in
production
right.
If
you
are
trying
to
strongly
isolate
your
cluster
administrators
from
your
cluster
users,
then
maybe
it's
not
production
ready.
But
if
they're
one
in
the
same
group
of
people,
then
it's
actually
not
that
bad,
because,
as
you'll
already
have
access
anyway,
sure,
let's.
I
G
B
A
B
I
D
B
Brandon
Phillips
sort
up
a
little
while
back
sort
of
the
definition
of
self
hosted
for
kubernetes
and
basically
proposing
is
building
one
two
for
self
hosted
system
which
is
at
CDN
up,
but
not
the
cubelet
or
if
we
decide
we
can't
make
a
city
itself
self
stood
it's
a
two
to
four
self
as
a
system,
so
I
think
in
terms
of
terminology.
B
B
G
B
D
E
F
So
one
of
the
things
we
were
looking
into
is
a
potential
internship
project
or
the
one
eight
timeframe
was
looking
into
the
API
server
counts
lie
API
server,
endpoint
issue
kind
of
wanted
to
to
gauge
possibility
of
doing
that,
and
if
there's
you
know
kind
of
figuring
out
a
scope
of
a
solution.
You
know
that
might
fit
and
go
on
eight
timeframe
because
there's
some
changes.
F
There's
some
other
changes
that
just
might
be
better
behavior
emails
in
the
near
term.
I
think
the
plan
that
Yvonne
and
Abhinav
were
going
to
move
forward
with
was
to
bring
this
up
in
API
machinery,
but
it
kind
of
affects
us
as
well.
So
it's
at
this
point
it's
a
little
bit
exploratory
like
the
seem
like
I
know.
We've
wanted
this
resolved
for
a
while
or
I
think
other
people
have
as
well
so
just
kind
of
reaching
out
to
see.
You
know
what
people
think
about
that.
So.
H
When
I
was
working
on
open
shift,
we
actually
implemented
this.
It
open
shift
and
swapped
out
the
default
implementation
that
kubernetes
has
so
it
does
use
the
at
CDV
to
TTL
implementation
and
I
know.
That
was
something
that
I
don't
remember
who
it
was,
but
somebody
for
Wes
didn't
think
that
was
a
good
idea
because
of
changes
coming
in
with
sed
3,
but
there
is
an
implementation
that
is
fairly
minimal.
That
basically
solves
the
issue
at
hand.
H
Although
I
know,
Jordan
Liggett
recently
had
posted
that
he
would
find
it
useful
to
have
a
fuller
definition
to
what
is
an
API
server.
What
features
does
it
support?
What
endpoints
does
it
have,
which
I
think
is
reasonable,
but
if
somebody
is
looking
for
kind
of
a
quick
and
dirty
solution,
what
was
written
for
shift
could
work
as
long
as
you
use
that
in
the
DVB
to
load
well,.
F
H
H
H
H
D
A
Be
would
be
not
yeah
I'm
expecting
the
parents,
probably
thinking
a
proposal
would
be
part
of
the
work
the
enzyme
would
do,
but
exam
yeah
okay,
so
we've
got
one
minute
left.
So
does
anyone
have
an
urgent
topic
for
the
last
minute.
E
E
E
B
I
think
I've
been
talking
to
throwing
a
little
bit
about
this,
and
he
also
is
is
happy
to
have
people
not
live
in
the
kubernetes
org
or
the
short
excuse
name
space.
So
we
need
to
sort
of
expand
where
things
go,
so
we're
I
think
part
of
what
sloppiness
is
unclear.
What
the
path
forward
is
in
terms
of
where
to
put
this
code
like
I,
know
where
we're
planning
on
moving
some
of
the
the
gke
code
out
and
we'll
probably
just
stick
it
under
the
Google
platform
or
instead
of
under
the
kubernetes
org
cool.