►
From YouTube: Kubernetes kops office hours 20191220
Description
Recording of the kops office hours meeting held on 20191220
A
Hello,
everybody
today
is
Friday
December
20th.
This
is
a
festive
edition
of
the
cop's
office
hours
meeting
I
think
we
have
fairly
light
agenda
and
probably
lighter
attendance
than
normal,
but
we
will
go
through
the
agenda
and
make
sure
that
this
is
put
on
YouTube
a
reminder
as
such
this
meeting
is
being
recorded
and
will
be
put
on
the
internet
and
Leasing.
Mindful
our
code
of
conduct,
I
put
a
link
to
our
agenda
in
the
chat
and,
please
feel
free
to
add
items
to
that.
A
Should
you
want
to
I
see
you
probably
doing
that
which
is
wonderful
and
there
are
a
half
dozen
plus
I
haven't
on
the
agenda
already,
so
we
can
go
through
those
I
guess,
hopefully
yeah,
let's
go
through
those
and
see
where
that
takes
us.
So
Eric
you
have
the
first
item
on
the
list.
Around
été
testing.
B
Yeah
so
I,
just
I
was
actually
reading
my
issue
because
I
could
assemble
what
it
was,
but
I
noticed
that
we
were
failing,
tops
post,
submit
push
to
staging
and
I.
Don't
know.
What's
going
on
with
that,
and
I
was
wondering
if
you
could
provide
a
little
bit
more
clarity
on
that
and
if
there's
anything
that
we
can
do
to
get
that
better.
So
that
we're
not
failing.
Yes,.
A
So
we
added
a
postman
job
that
is
supposed
to
work
towards
the
idea.
Well,
you
know
there's
this
like
working
group,
cakes
in
fro,
which
is
building
sort
of
providing
docker
or
container
registries,
and
a
non
container
buckets
to
all
the
projects,
and
the
idea
is
that
we'll
have
this
automated
promotion
process
or
we'll
have
a
structured
promotion
process,
and
ideally
we
get
to
no
automate
information
process.
A
The
first
step
is
that
on
every
merge
to
master,
we
should
build
our
images
and
push
them
into
what's
called
a
staging
bucket,
which
basically
is
it's
a
hosted
bucket,
but
is
sorry
staging
repo,
for
which
is
a
a
project
repo
from
which
we
could
then
promote
to
release
but
isn't
really
considered.
As
trusted
I
mean
it's
still
trusted,
but
it's
not
as
it
doesn't
it
all.
It
means
is
that
Simcoe
got
to
master
okay,
and
so
we
are.
We
added
that
job
I,
don't
think
we're
really
expecting
it
to
work.
A
It
really
didn't
work
when
we
first
started
it.
Okay
and
now
I
think
it's
closer
to
working,
but
I
need
to
look
into
it
as
I
think
anyone
could
look
into
it.
I
am
created
a
job.
I
have
more
contacts
on
what
it
is
should
be
doing,
but
as
far
as
I
know,
basically
it
should
be
that
we
have
a
job.
It
runs
as
a
post
submit
ie
after
a
commitment
on
master.
It
runs
it's
supposed
to
run
in
GCV,
Google
Cloud
Builder,
it's
supposed
to
build
our
well,
it
needs
the
end
goal.
A
Don't
ever
do
exactly
yeah
and
if
that's,
if
that
works,
that's
great,
it
is
even
if
it's
failing
it,
isn't
it's
a
new
functionality,
so
it
isn't.
It
is
another
test
and
looking
like
you
know,
so
we're
failing
failing,
but
it
would
be
great
to
get
the
great
could
make
some
progress
on
that.
I
cannot
have
a
look
at
and
see
if
I
recognize
what's
gone
wrong,
but
yeah
I
mean
just
that's.
That
is
the
state
of
what
that
is.
Okay,.
A
Aaron
all
right
man
I
a
question
which
I
may
be
moot,
but
people
don't
know
happen.
We
got
container
D
in
a
friend
contributed
support
for
container
D
into
the
master
branch,
so
that
will
land
in
one
eight
cops,
118,
I,
guess
yeah.
A
A
A
The
clever
thing
that
you
did
was
you
took
our
docker
install
and
you
said,
docker
is
actually
installing
docker
in
container
D,
and
so
we
should
treat
an
install
of
Dockers
for
beyond
a
certain
version,
the
ones
that
install
continually
separately.
We
should
actually
use
the
full
container
to
install
and
therefore
we
have
a
little
more
control
over
it.
I
just
wanted
to
like
check
that
that
didn't
break
our
sort
of
backwards
compatibility
guarantees
for
any
of
our
released
versions.
A
D
A
A
B
A
Don't
think
anyone's
gonna
accidentally
do
this
I
think
when
we
introduce
and
when
we
at
some
point
we
want
it
to
default
to
container,
do
you
believe
and
that's
we
would
want
to
default,
then
only
for
like
kubernetes,
118
or
greater
119
or
kubernetes
120,
and
normally
only
for
new
quest.
Well,
we'd
start
only
for
new
clusters
or
something
yeah.
D
A
D
D
A
Don't
know
so
I
mean
I,
don't
know
if
anyone,
if
you
have
a
view
or
if
others
have
a
view,
I
think
we
do
have
this
idea
that
we
don't
try
to
us
that
we
say
you
can
use
if
you
use
cups,
116
and
install
kubernetes
116,
and
then
you
use
cups,
118
and
still
use
your
existing
config.
We
try
not
to
break
anything
except
for
security,
fixes
or
we'd
also
try
not
to
change
anything
except
for
security
fixes
or
bug
fixes,
and
you
could
argue
that
this
is
like
in
that
sort
of
category
I.
A
It
makes
me
a
little
uncomfortable,
but
on
the
other
hand
we
haven't
necessarily
released.
We
no
way
yeah,
it's
pushing
them
a
little
bit
so
I,
don't
know
how
other
people
feel
about
this,
but
if
we
couldn't
make
it,
we
can
make
it
that
we
feel
like
there's,
no
actual
difference.
I
think
that
could
be
okay,
but
I
like
there
is
a
version
of
docker.
If
I
recall
correctly,
where
the
version
of
container
D
changes-
and
that
makes
me
a
little
more
uncomfortable,
I.
D
A
Okay,
yeah
I
mean
I,
think
so
I
think
that's
reasonable,
I
think
that's
good
I
think
we
can
put
it
in
the
release,
notes
and
maybe
flush
it
out
in
alpha
and
beta
I
guess,
because
I
think
I
think
that
sounds
like
a
good.
It's
the
right
thing
to
do.
It's
actually
a
better
way
to
install
container
G
I,
think
that
makes
sense
and
we
can
see
if
we
might
have
to
change
it
down
the
road,
but
I
think
we
should
pursue
what
you
did
there
like.
We
should
be
restarting
container
D
right.
It.
A
A
A
All
right,
yeah,
that's
that's
great
and
so
I
think
yeah.
That
actually
reminds
me
I,
think
I
think
I
found
out
that
we
might
not
be
doing
system
to
unify
all
restarts
correctly.
Like
I
read,
there's
some
other
shouldn't.
You
have
to
do
so
I'm
going
to
like.
Apparently
they
can
still
go
into
a
failed
state,
so
I'm
going
to
try
to
track
that
down.
Let
me
write
that
down
as
an
action
item
for
myself.
B
D
A
So
it's
certainly
fine
to
add
a
periodic
tests
that
might
not
pass
right
away.
We
certainly
had
some
of
those
tests
that
have
like
the
right
away
turns
into
like
years,
which
is
not
ideal,
but
it
is
sort
of
a
good
way
to
track
the
state
state
of
these
things
like
to
me.
That's
valid
signal,
so
I
think
I
guess.
A
When
we
talk
about
the
rel
8:1,
we
can
talk
about
like
adding
another
one
for
another
test
for
a
container
D
configuration
and
then
even
if
that
doesn't
pass
today,
it
will
be
interesting
to
understand
what
the
failure
is.
I
would
hope
it
passes.
But,
yes,
we
might
need
to
exclude
some
different
things.
Well,.
A
B
A
We're
not
gonna
use
the
side
entry,
we
should
not
do
it.
A
in
fact
should
I
keep
threatening
to
write
the
like
figure
out.
What
the
what
the
language
you're
implementing
in
like
is
this?
What
the
language
you're
implementing
is
like?
Is
this
Jason
path
and
like
write,
the
I
guess,
expression
evaluator
that
would
like
set
all
of
them
so
eventually
the
hope
is
that
set
will
always
be
supported
for
everything,
and
we
won't
have
to
do
flags
for
things,
but
in
this
case
it
feels
pretty
important
to
do
the
CRI.
So
we
should.
A
A
A
D
D
A
B
A
So
yes,
I
think
that
would
be
great
too,
and
the
script
does
pass
by
virtue
of
excluding
the
directories
ice
to
do
some
more
directories.
And
so,
if
we
continue
with
our
strategy
of
like,
let's
get
it
in
a
prow
job
and
we
can
then
prevent
any
regressions
because
I
think
like
there
was
one
that
crept
in
somewhere.
I
think
that
I
saw
and
so
I
could
getting
it
in.
There
will
stop
any
regressions,
and
then
we
can
like
continue
to
whittle
down
the
directories
and,
as
you
say,
I
think
like.
A
B
B
Notice
in
proud
that
they
added
support
for
defining
prowl
jobs
in
the
repo
being
tested.
That
was
a
recent
change,
Wow
and
so
I'd
make
these
kinds
of
things
easier
in
the
future.
That
would
be
really
cool
that
would
be
wonderful,
is
very
complicated.
I
saw
my
life
figuring
happen,
Danny's
gonna,
last
few
weeks,.
A
Crickets,
okay,
that
was
always
like
the
objection
in
the
past
is
like
okay,
whatever
uh-huh
I
guess
like.
We
just
have
to
like
to
take
responsibility
for
our
outcomes,
so
she
sort
of
do
anyway,
so
yeah
we're
gonna,
be
owners
in
different
repo.
Yes,
okay,
it's
not
safe
on
a
PR;
maybe
they
figured
to
run
them.
Maybe
they
forget
how
to
run
the
tests
from
the
master
branch
regardless
of
the
PR.
So
in
other
words,
even
if
you
change
the
test-
or
maybe
they
just
run
in
a
better
sandbox
yeah.
A
B
A
Would
say
we
should
look
at
putting
new
ones
into
our
me
bone
as
a
starting
point.
Isis
everywhere
we
go,
yeah
I
suspect
there
gonna
be
some
pretty
big
like
this.
One
is
a
great
one.
The
static
check,
yeah
I,
suspect
there
gonna
be
some
limitations,
like
I
suspect
we
can
put
things
like
static
check
in,
but
we
can't
put
e
to
e
yep
tell.
A
A
A
B
So
we
have
a
bunch
of
active
branches
and
cherry-picks
and
we
have
milestones
and
github
and
I've
found
it
helpful
to
tag
cherry-picks
with
milestones
so
that
we
can
easily
see
which
open
PRS
are
targeting
a
specific
release
branch.
There
is
a
pro
plugin
that
does
this
for
us
automatically,
so
I
opened
a
PR
to
enable
it
and
they
just
need
other
people
to
approve
it.
So
so
I'm
going
to
do
that,
that'd
be
great.
B
A
A
D
From
what
I
saw
there,
we
have
CentOS
7
and
Ubuntu
1604.
These
are
quite
ancient.
Is
there
a
limit
or
anything,
there's
no
practical
in.
A
That
it
does
cost
money
in
practice.
We
probably
should
turn
down
the
frequency
of.
If
we
were
butting
against
that
limit,
we
should
turn
down
the
limit.
The
frequency
of
some
of
the
existing
jobs
and
just
sort
of
like
spread
the
body
across
more
coverage,
so
I,
don't
think,
there's
a
limit
in
terms
of
the
number
of
jobs
or
in
terms
of
like
introducing
this.
A
We
and
we've
talked
in
the
past
that
it
would
be
great
to
have
more
people,
sort
of
I
presume
you
run
rl8
and
so
like
more
people
like
identifying
and
championing
and
looking
at
their
their
configuration
that
matters
to
them.
So
I
think
it's
a
good
idea.
One
maybe
challenge
is,
you
might
have
to
use,
sent
us
eight,
instead
of
which
I
presume
exist
instead
of
relay
just
because
I
don't
know
how
the
licensing
I
don't
know
if
there's
any
licensing.
That
is
sorry.
A
You
know
so
anyway,
I
don't
know
if
I
don't
know.
If
you
should
try
it
I,
don't
know
if
there,
if
we
are
able
to
use
the
rel,
8
am
I
just
in
that
account
like
obviously,
for
example,
if
I,
if
I
were
to
create
an
ami
that
shows
a
thousand
dollars
an
hour.
Presumably
criminally
cess
infra
would
not
be
very
happy
if
I
created
a
test
that
ran
it
their.
D
A
Sense
yeah!
So
yes,
if
you
want
to,
if
you
want
to
try
adding
the
r-la
test,
it
should
be
very
similar
to
the
CentOS
7.
You
just
change
the
image
and
that
would
be
great
I.
Think
I.
Imagine
the
event
some
will
say:
hold
on
you're,
like
you're
costing
too
much
money,
in
which
case
we'll
just
turn
down
the
frequency
a
little
bit.
Okay,.
D
A
B
A
B
If
we
think
that
we
should
log
a
warning
in
cops,
1:16
get
that
warning
added
and
cherry
picked
back
last
week
we
decided
I,
think
it
was
less
than
or
equal
to,
1/8
is
deprecated
and
I
was
going
through.
Some
of
our
you
are
code
where
we
compare
kubernetes
versions
and
I
noticed
that
we
map,
if
someone
doesn't
specify
a
docker
version,
we
mapped
their
kubernetes
version
to
a
darker
version,
which
made
me
think
about.
B
A
Feel,
like
the
darker
versions:
don't
cost
us
as
much
today.
I'd
say
I,
don't,
but
it
is
a
terrible,
terrible
idea
to
run
a
super
old
daughter
version
like
yeah
like
and
they
are
each
version.
Illuminati's
is
effectively
recommends
a
darker
version,
and
so
we
are
effectively
deprecating
those
darker
versions.
If
you're
saying
do
we
actually
remove
them
like
we,
never
we
never
supported
the
docker
version.
We
supported
the
communities
with
the
doctor
version,
I.
A
A
119
180:
let's
go
for
118
118
I,
can't
imagine
I
can't
imagine
any
one's
gonna
push
back
on
this,
but
it
would
be
like
let's,
let's
be
as
let's
be
as
aggressive
as
we
can
in
our
opening
statement
to
see.
If
any
complains,
if
we
like
what
it
feels
like
I
mean
microdata
itself
is
like
a
person
is
that
person
twenties
is
so
far
to
support
anyway,
like
that's,
yeah
should.
D
A
Think
that's
what
are
we
gonna
get
there
I
think
I
think!
Yes,
you
want
to
do
that.
I
think
I
think
we
all
would
like
to
we
sort
of
talk
to
us
a
little
bit.
I
guess,
I!
Think
we
don't
like
to
get
that
number
ten
down
a
little
bit
so
I,
don't
think
we
want
to
define
the
policy
is
being
ten
when
for
that
we're
for
kubernetes
a
number
is
either
three
or
four
I
can't
remember
what
like
it's
three
or
three,
it's
three
I
think
I
might
go
to
four
and
I.
A
It's
not
clear
exactly
how
you
count,
but
like
I
think
that
ten,
the
difference
between
ten
and
four
is
very
big
and
I.
Think
we'd
be
much
happier
if
that
was
like
around
the
six
or
seven
or
eight
mark.
So
I.
That's
what
that's
sort
of
that
I
think.
The
first
step
is
just
to
take
something.
That's
so
far
back
that
no
one
is
gonna,
object,
fingers
crossed
and
see
who
objects,
and
then
we
can.
But
yes,
it
should
be
rolling,
but
I
also
think
we
should
actually
try
to
accelerate
the
role.
I
agree.
A
We
can
we
can
I
like
what
you
said.
We
perhaps
when
we
write
this
Peter,
we
can
say
we
will
in
117,
deprecated
or
and
119
will
remove
support
for
one
nine
and
120
or
support
for
110,
which
is
nice,
because
it's
ten,
which
is
a
nice
round
number
and
then-
and
we
also
like
start
to
accelerate
that
or
we
will
start
to
accelerate,
will
start
to
like
narrow
the
gap
over
time
as
well.
C
A
I
I
think
it's
as
far
as
I
know
it's
separate,
but
yes,
we
should
I.
Think
you
brought
this
up.
I.
Think
we're
in
favor
of
okay
nobody's
been
reviewing
my
PR,
so
sorry,
yeah,
no
it's
I
I
know.
We've
said
last
time
that
we
were
going
to
focus
on
that.
Didn't
I
have
been
I
was
trying
to
get
the
container
d1
and
has
it's
like
the
biggest
and
then
I'm
gonna
I'm,
hoping
to
find
more
time.
This
two-week
break.
A
But
yeah,
like
you
know,
peer
review
is
always
a
it's
a
it's
a
difficult
balance
right,
so
people
balance
between
like
finding
the
time
and
trying
to
get
the
right
things
in,
but
this
does
rewrite
what
your
to
men
does
feel
super
important.
So
we
should
I
will
try
to
check
them.
You're
you're
are
you
are
next
on
the
agenda?
John
is
that
was
there
taint
all
notes
during
one
method?
This
is
a
new
feature
right
or
a
new.
Well.
E
A
All
right,
yes,
I
I,
will
try
it
also
we've
yeah
request
count
has
gone
up,
which
is
unfortunate,
I
think
it's
gone
up,
certainly
hasn't
gone
down
significantly
and
I
think
we
yeah.
We
want
to
get
back
down
and
I.
Think
me,
I'm
certain
gonna
try
to
get
some
reviews,
but
we
also
want
to
get
the
yeah
Rosamond
a
priority.
Yes,
yes,
okay,
I,
haven't
forgot
the
right
priority
function
for
PRS
to
be
honest
and
I,
pretty
much
don't
have
one
and
pretty
much
it's
random
so
often
see
me.
A
C
A
So
we'll
all
you
will
look
at
8:21
and
I
presume
we
have
some
others,
which
I
will
also
look
at
and
then
I
have
a
meter,
which
is
a
note
to
self
that,
like
I'm,
not
convinced
about
our
system
to
unit
file
restart,
follow
suit
cause.
Suddenly
I
learned
and
then
john
you
have
the
last
item
on
the
agenda,
which
is
CI
checks
on
release
branches.
Yes,.
C
A
I
think
we
should
I,
don't
know
why
I
went
on
Travis
on
the
other
ones.
That's
odd
that
we
do,
but
we
should
I'm
trying
to
see
like
I'm
from
put
cherry
pick
PR.
So
it
looks
like
just
thinking
one
at
random.
It
looks
like
we
run
e
to
e.
We
do
run
Travis,
which
is
good,
so
that's
the
that's.
The
slip-up
is
like
I've
started,
pulling
things
out
of
Travis
on
the
grounds
that
prowl
is
running
them,
but
that
isn't
necessarily
true,
because
not
all
the
projects
run
so
before
I.
A
A
Correct
the
only
reason
is
if
we
added
a
job
in
that
was
not
available
on
the
previous
one
yeah,
and
that
has
been
a
problem
in
the
past
that
we
can
always
override
the
merge.
We
always
like
get
the
signal
and
then
say:
I
will
choose
to
by
getting
the
signal
your
rights,
it's
like
there's
no
reason
not
to
have
the
signal
and
we
can
override
button
well,
I
at
least
kind
of
read
the
button.
I'm
not
sure
in
general.
I
should
have
that
permission,
but
I
do
have
that
permission.
A
A
A
A
E
Yeah
I
don't
need
like
an
answer
back
right
now
like
if
you
want
to
kind
of
mill
it
over
like
totally
acceptable
yeah.
If
we
could
just
get
some
conversation
going
about
that,
cuz
I
was
like
that
would
be
kind
of
in
line
with
how
like
we
do
things,
but
I
didn't
want
that
to
be
disruptive
for
other
people
who
might
be
using
manage
policies,
but
that
kind
of
saw
an
avenue
there
where
you
could
like
over
time.
It
was
just
sit
there
and
stack
these
manage
policies
over
time.
E
B
A
B
A
I'd
be
great
yes,
so
we
have
I,
think
it's
called
complex.
I
should
call
it
like
kitchen
sink
or
something
I
can
throw
it
in
the
in
the
complex
test
and
let's
just
have
a
look,
because
it's
sort
of
nice
as
a
preview
but
I
think
yeah.
It's
certainly
it's
also
interesting
to
think
about.
Like
yeah.
The
the
garbage
fashions
that
were
on
them
is
interesting
as
well
all
right.
Thank
you
for
the
context
on
that
and
I
will
try
to
take
a
look
on
that
as
well.
A
Sorry
about
the
delay
on
that
one
as
well
I
know,
if
there's
anything
else,
anyone
wanted
to
bring
up
or
oh
man,
you
have
your
hand
up,
but
I
don't
know
if
I
should
assume
being
zoom
yeah.
My
question
was
already
answered:
okay,
cool
sorry,
I'm
gonna
pay
attention,
I
know.
If
anyone
else
has
their
hand
raised
now.
No
you
don't
mean
having
to
check
I,
don't
see
any
other
hands
raised.
I,
don't
see
one
else
on
muting,
so
I
wish
everyone
a
very
happy
weekend
in
a
very
happy
holiday
break
season.