►
From YouTube: 20210816 SIG Arch KEP Reading
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
Okay,
I
think
we
should
be
good
to
go.
Recording
progress,
okay,
hi
welcome
to
another
cap.
Reading
club
session
today
is
august
16th
and,
as
usual,
we
follow
a
code
of
conduct
which
boils
down
to
be
excellent
to
each
other,
and
this
meeting
is
being
recorded.
So
please
don't
say
or
do
anything
that
you
don't
want
online.
A
So
with
that
we
have
two
caps
today.
The
first
one
is
sport,
cost,
which
is
the
the
responsible
seriously
gaps,
and
this
looks
like
a
moderately.
A
A
A
Okay,
awesome
thanks!
Okay,
so
I'm
starting
the
timer
in
three
two
and
start.
A
A
All
right,
I'm
assuming
we
can
start
discussing
so
does
anyone
have
questions
off
the
bat.
A
I
think
the
feature
gate
is
like
mislabeled
kind
of
like
so
in
the
feature
gate
section
of
cap.dml:
let's
replica
set
pod
deletion
cost,
but
then
I
couldn't
really
find
it
in
the
docs
or
anywhere
else.
But
then
in
like
the
actual
feature
gate
file,
it's
called
pod
deletion
cost
and
that's
what
is
reflected
in
the
docs
as
well.
I
think
so
I'll
probably
reach
out
and
ask
which
one
would
you
like
to
be
kept.
A
B
A
A
Also,
but
generally,
the
case
is
that
if
a
feature
is
in
beta,
it's
by
default
enabled
if
it's
an
alpha
it's
by
default,
disabled
and
you
need
to
like
explicitly
flip
the
gate
or
like
pass.
A
command
line.
Argument
saying
enable
this
sort
of
thing.
A
But
my
question
was
what,
if
let's
say,
you
have
five
faults
in
a
replica
set
and
there
is
deletion
cost
mentioned
for
two
of
them.
So
would
it
be
such
that
those
two
would
be
considered
for
deletion
in
accordance
to
this,
and
then
the
remaining
three
are
deleted
by
default
like
according
to
the
default
way
of
things,
or
is
it
just
all?
Is
it
like
an
all
or
none
kind
of
situation
that
happens
here?
C
I
guess
regarding
this:
it's,
like
you,
know
the
active
pods
with
the
ranks,
the
definitions.
I
think
all
the
definitions
make
it
clear
about
what
you
are
asking
about.
The
situation
which
you
asked.
I
think
that,
like
you
know
there
is
there
is
something
called
this
active
pods
with
ranks.
How
they
are
sorted
out
is
like
most
of
the
situations
with
two
parts.
You
know
how
they're
ranked
and,
like
you
know,
considered
those
are
already
written.
They're
like
you
know,
there
are
seven
rules.
C
I
think
what
is
there
in
the
proposal
is,
apart
from
that,
what,
like
you
know
the
following
criteria
that
is,
like
you
know,
comparing
two
bonds,
regardless
of
deletion
cost.
So
these
three
criterias
are
not
the
universal
one
rather
than
the
you
know
what
is
defined
in
that
active
pods
with
ranks
the
ranking
is
more
universal
than
what
is
mentioned
in
this
three
like
this
comes
after
that.
No,
this
is
not
a
primary
one.
D
A
All
right
so
just
like
before
we
start,
I
see
like
a
few
new
faces
so
basically
in
these
sessions,
as
you
might
have
now
noticed
what
we
usually
do
is
we
take
a
cap,
so
people
suggest
skips
that
are
of
interest
to
them
or
they've
written
them
and
they
would
like
feedback
on
it,
so
they
suggest
those
kepts
and
we
take
that
cap
spend
around
10
minutes
reading
that
together
and
then
discuss
that
cap
and
like
ask
questions
related
to
it
or
give
feedback,
or
things
like
that.
A
So
that's
basically
what's
happening
here.
So,
if
folks
join
in
the
middle,
then
that's
why
all
the
silence,
nothing
else.
Okay,
let
me
start
another
timer
and
we
can
look
at
the
next
cap.
A
So
this
cap
is
at
adding
grpc
probe.
This
cap
is
part
of
sig
node
and
it's
this
one's
targeted
for
123
because,
like
I
remember,
this
was
discussed
in
the
last
meeting.
So
this
one
is
for
sure
in
123..
So
if
folks
are
interested
in,
like
maybe
contributing
you
can
reach
out
to
signal,
okay,
starting
the
timer
in
three
two
and
one.
A
Also,
it's
just
a
reminder,
so
if
people
are
like
trying
to
access
the
doc
and
are
not
able
to
like
edit
the
doc
rather
to
get
edit
access,
you
will
have
to
like
join
the
city
architecture
meeting
list.
C
I
have
a
question
you
know
others
mentioned
about
this
now
as
specified.
The
cubelet
pro
will
not
allow
use
of
client
certificates
nor
verify
the
certificate
on
the
container.
So
I
want
to
understand
if
this
feature
you
know,
verification
of
a
certificate
on
the
container
was
already
a
default
feature
of
cubelet,
which
is
disabled
now
or
it
was
not
a
feature
at
all
enabled
in
cubelet.
Also
is
this
feature
available
with
grpc
health
checker
or
not,
because
verification
of
certificate
and
container
is
a
very
important.
C
B
Yeah
that
so
that
was
the
question
I
had
too
grpc.
Health
checker
is
actually
the
same
behavior
by
default,
so
you
can
enable
it
to
verify
certs
in
check,
but
by
default
it
will
not.
So
if
you're
using
grpc
health
checker
without
enabling
any
of
that
it
should
be
the
same
kind
of
behavior.
C
Okay,
but
here
it
is
mentioned
that
cubelet
pro
will
not
allow
so
once
the
grpc
is
integrated.
It's
called
this
readiness
proof.
So
will
that
also,
you
know
or
not
verify
the
certificate
on
the
content
like?
Is
it
a
means,
something
we
could
enable
and
disable
on
a
purpose
or
it
is
like
it
will
not
allow
the
use
of
this
verification
of
certificates.
B
Yeah
that
that's
how
I
took
it
as
it
won't
work
with
tls
on
the
with
the
feature
here
and
that's
probably
I'm
sure,
I'm
sure
that's
like
a
ease
of
implementation
thing
and
if
I'm
sure,
if
there's
enough
like
need
for
it,
then
you
know
it
can
be
added
in
a
future
cap
and
iteration.
A
C
Yeah
no,
like
you
know,
last
year
I
used
this
grpc
thing.
I
was
using
all
my
internship,
so
you
like,
I
worked
with
this
health
checker
and
you
know
just
just
a
bit
of
it.
I
used,
but
you
know
I.
I
was
not
that
much
acquainted
like
you
know.
Even
I
didn't
know
the
rpcs
and
open
source
or
something
like
that,
but
yeah.
I
have
a
like
my
very
skewed
idea
about
how
those
things
work.
A
So
I'm
still
a
little
unclear
on
the
fact
about
how
is
go.
How
is
the
cubelet
going
to
convey
the
status
back
like
in
what
format
so
is
it
going
to
be
like?
Is
the
readiness
readiness
probe
format
is
going?
A
Is
that
just
going
to
be
the
same,
or
is
that
going
to
differ
somehow
to
incorporate
more
verbosity
for
the
grpc
health
check,
or
is
there
going
to
be
like
an
option
to
maybe
like
provide
more
information,
if
necessary,
so
not
sure
yet,
but
I'm
assuming
like
since
they're
targeted
they'll,
probably
do
a
cap
iteration
now
and
add
more
details.
A
B
Yeah,
I
see
this
as
being
useful
for,
if
you're,
using
a
binary
to
exec
a
health
check,
that's
how
it
determines
so
yeah.
I
think
it's
just
checking
like
the
status
codes.
I'm
still,
I'm
not
sure
how
many
people
are
using
grpc
based
health
check
endpoints,
but
it's
cool.
A
But
I
mean
I
guess
it:
it
sort
of
makes
sense
if
you
have
an
increasing
number
of
services
them
itself
running
in
the
cluster.
A
So
if
you
would
like,
if
you
could
get
the
cubelet
to
do
all
of
those
health
checks
for
you
in
some
form
and
then
maybe
I
don't
know,
write
a
controller
which
checks
for
a
condition.
Maybe
since
a
lot
of
it
can
be,
you
could
abstract
away
a
lot
of
details
there,
that's
sort
of
nice.
I
guess.
B
C
Okay,
one
more
thing
that
is
mentioned
about
this:
you
know
like
in
the
grpc
port
action,
dot
port.
They
had
this
init
32
declared
at
like
they
are
saying
that
you
know
this
was
on
purpose
and
we
want
to
move
users
away
from
using
the
port
number
and
port
name
union
type.
I
didn't
get
that.
You
know
why.
Why
is
that
necessary?
Like
you
know,
do
you
mention
the
port
number
or
name
like
I?
I
like?
What
are
these
two
basic?
You
know
variables
or
functions
a
port
number
and
opening.
B
Yeah,
I
didn't
understand
that
either
especially
the
moving
away
from
part,
because
right
now,
when
you're
doing
when
you're
tying
a
deployment
to
a
service,
you
have
to
like
name
the
port,
and
then
you
can
reference
the
name.
So
I
didn't
realize
that
was
something
we're
trying
to
get
people
to
stop
doing.
I'm
not
sure.
A
I
think
it
also
might
help
looking
at
the
pr
for
this
cap,
because
I
thought
tim
hawkin
was
one
of
the
approvers
and
I'm
sure
like
he
would
have
had
some
like
nice
comment
on
that
particular
part
of
it
or
like
some
nice
discussion
would
have
been
this.
Let
me
see.
A
A
B
A
I
there's
already
one
cap
and
like
the
back
club
disc,
the
first
part
yeah,
so
I
I
tried
contacting
in
saucers,
but
she
couldn't
make
it
today.
So
what
she's
like
like
she
really
wanted
to
discuss
it
also.
So
she
asked
me
to
like
put
it
up
for
next
week
so
yeah.
So
we
already
have
one.
If,
if
we
can
get
one
more
that'd
be
awesome.
A
But
yeah
so
like
if
we
can
get
one
more
that
regret
and
like
preferably
like
early
on
but
lucky
for
us,
we
have
two
weeks.
So
it's
not
a
bad
time
period.
B
A
So
yeah
at
this
time
it
was
a
little
weird
because
I
remember
like
a
lot
of
gaps,
a
lot
of
implementations
of
keps
landed
in
the
same
week
so
and
because
they
all
landed
in
the
same
week,
things
looked
good.
Initially
they
got
merged
and
then
one
affected
the
other,
and
then
some
tests
started
failing
here
and
there
and
people
weren't
sure
like
what
pr
caused
the
failures.
A
A
C
B
D
A
I
actually
had
one
idea
which
I
was
like
hoping
to
bring
up
in
a
cli
meeting,
but
I
was
waiting
for
like
jordan's
approval
sort
of
thing,
but
like
basically
it's
on
cube
cuddle
debug.
A
So
someone
suggested
like
there
was
some
feedback
on
twitter
regarding
fml
containers
and
debug
in
general,
and
they
suggest
that
it'd
be
nice.
If
we
can
like
have
cubecoil
debug,
hyphen
knife
and
rm,
so
like
forecaster
administrators,
if
they
want
to
delete
it
and
since
123
like
in
122,
removal
of
fmer
containers
was
supposed
to
be
supported,
but
then
it
got
pushed
to
123
because
of
some
issues
happening
there
and
but
then
jordan
raised
some
concerns
about.
A
Do
we
really
need
deletion
of
fm
error
containers
in
the
first
place,
so
that
discussion
is
still
going
on,
but
and
if,
by
the
end
of
it,
there
is
a
conclusion
that
okay,
this
is
actually
like
a
nice
needed
feature.
Then
I
probably
like
bring
it
up
in
a
cli
meeting
and
like
see
what
folks
have
to
say.
B
Please
do
and
make
sure
you
talk
to
verb
elite,
lee
lee
verbing
yeah,
so
I've
I've
been.
A
Working
with
him
for
about
some
time
with
for
some
with
like
sentence,
is
grammar
what
is
happening
so
yeah.
So
basically,
I've
been
working
with
him
for
some
time
now
on
fm
memory
container,
so
that
seemed
like
a
nice
addition.
Also
so
yeah
like
there
was
quite
a
lot
of
feedback
on
twitter,
but
turns
out
all
of
them.
Most
of
the
feedback
would
be
solved
in
122.