►
From YouTube: Kubernetes kops office hours 20190927
Description
Recording of the kops office hours meeting held on 20190927
A
Hello,
everybody.
It
today
is
Friday
September,
27th
2019.
This
is
cops
office
hours.
I,
am
your
moderator,
facilitator,
Justin,
Santa,
Barbara
I
work
at
Google,
a
reminder.
This
meeting
is
being
recorded
and
will
be
put
on
YouTube
probably
later
this
afternoon.
So
please
be
mindful
of
that
and
of
our
code
of
conduct
and
I
pasted
a
link
to
our
agenda
in
the
chat,
I'll
replace
it
because
I'm
never
sure
whether
slack
actually
keeps
it
or
not,
but
we
have
a
fairly
full
agenda
today.
A
A
Basically,
a
lot
of
these
issues
are
all
of
these
topics
are
things
which
I
put
in
the
agenda
but
have
come
up
in
slack
or
on
github.
So
please
interrupt
if
you
are
in
fact
the
a
better
person
to
talk
about
them
than
I
am,
but
one
great
conversation
that
came
out
the
other
day
in
slack
was
whether
we
should
have
sort
of
golden
configurations
like
this
idea
that
we
have
a
fairly
large,
probably
a
very,
very
large
permutation
set
of
configurations
for
cops
like
once.
A
There
was
discussion
of
having
some
form
of
like
survey
so
that
we
could
actually
understand
what
people
really
use
I,
don't
know
whether
that's
something
that
in
the
past
we've
done
surveys,
we've
had
trouble
getting
I
guess
significant
results
like
it's.
There
was
a
that's
quite
a
lot
of
work
to
like
provide
all
your
information
and
it's
hard
to
like
great
inaccurate
surveys.
A
One
suggested
a
great
suggestion:
if
people
will
do
it,
which
was
to
upload
a
sort
of
sanitized
versions
of
their
cluster
spec
of
their
yeah
Mille,
that's
a
fairly
large
amount
of
work,
though
yeah
both
to
sanitize
and,
on
the
other
end
to
collect
so
I'm,
not
sure
other
people
would
do
that,
even
if
we
did
all
that
work,
but
I
like,
for
example,
I
think
our
ETS
are
not
working
right
now
and
I
am
yeah.
I
definitely
think
we
should
get
that
going.
A
I
think
we
should
I'm
gonna
like
add
a
couple
of
like
combinations
that
are
important
to
me
personally
or
that
I
think
are
important
in
the
community
and
then
we
can
like.
If
people's
configurations
are
not
in
that
list,
then
we
can
presumably
add
more
if
need
be,
but
it
is
sort
of
it
would
be
great
to
get
a
list
of
which
ones
those
are
as
at
least
as
a
starting
point
same
as
anything.
They
want
to
add
I.
A
Yes,
I
imagine
I,
think
I
think
if
we
got
a
couple
of
like
C
or
R
a
cup.
If
we
put
a
couple
of
configurations
on
the
home
page
which
were
or
on
our
top-level
readme,
which
were
like
these
configurations
are
the
best
tested
I.
Imagine
that
set
will
grow
rapidly,
as
people
want
to
get
their
particular
favorite
configuration
in
there,
which
I
think
it's
a
good
thing.
A
So
do
it
all
right?
Well,
let's
keep
going
on
the
agenda.
We
can
always
come
back
to
anything
that
if
anyone
has
any
other
thoughts
on
these
things,
so
Peter
I
see
you
are
here.
You
have
a
PR
which
looks
wonderful
to
introduce
static
check,
which
is
a
if
you
want
to
describe
what
it
is.
I,
don't
know
whether
you're
able
to
talk
right
now.
A
I'll
assume
not
but
do
pop
in
if
you
want
to
so
it's
basically
a
lot
of
static
analysis.
A
against
our
go
code
should
catch
a
lot
of
errors
like
incorrect
format,
strings
or
differs
in
block
basic.
It's
like
a
go
vet
greatly.
Enhanced,
so
basically
should
give
us
much
higher
quality
and
give
people
sort
of
automated
code
reviews
on
the
sort
of
basics
that
are
automatable,
which
is
always
helpful
like
cuz
I
miss
stuff
and
like
certainly
like
we
come
across
these
ones
who
are
like.
A
We
are
masking
an
error
or
not
even
checking
an
error,
and
it's
like
how
did
that
get
through.
But
yes,
unusually,
it's
me
that
writes
tell
myself
and
I
had
it
like
that.
How
did
I
do
that?
So,
yes,
that
would
be
super
helpful
and,
like
the
approach
that
Peter
has
taken
is
to
it
is
Peter.
Isn't
it
their
first,
video
I
think
it's
Peter.
A
But
if
there
are
no
objections,
I
think
we
should
try
it
and
see
if
there
are
tons
of
false
positives
from
that,
then
we
can
deal
with
that,
but
otherwise
we
should
probably
do
it
and
it's
also
used
company
in
in
KK.
So
so
that
feels
reasonable.
Peter
I
see
you
have
switched
zoom
devices.
Do
you
once
you
say
anything.
B
Carry
on
I
would
I
was
just
gonna
real,
quick,
sorry
doesn't
so
yeah
I
fully
support
this?
Would
it
make
sense
for
us,
you
know
if
this,
when
we
get
this
merged
for
us
to
pull
it
in
as
like
a
non-blocking
test
like
in
the
short
term,
I
mean
I,
guess
if
it's
passing
right
now,
that's
cool
then,
but
I
don't
know.
If
maybe
we
should,
you
know
do
something
like
that.
That's.
A
It's
actually,
why
don't
we
start
with
it
in
blocking,
because
otherwise
you
have
to
create
another
one.
I
don't
know
man
we
can
be
like,
but
I
think
it's
a
good
idea.
This
is
restructure
it
maybe
as
a
non-blocking
separate
test,
but
don't
because
it
might
be
much
slower,
I'm,
not
actually
sure
I
have
not
run
it
myself
directly,
because
I
think
right
now
it's
in
the
I
kind
of
wish
my
car
targets.
It
is
in
right
now
it
is
in
the
verify,
static
check,
which
is
I.
Guess
it's
a
new
target
entirely.
A
A
C
A
A
Next
on
our
agenda,
who
do
endtime
we're
doing?
Okay?
Next
on
our
agenda,
so
this
week
or
this
two
week
we
did
not
have
where
were?
We
did
not
release
one
14-0,
primarily
because
we
had
a
little
snafu
with
CentOS,
where
I
guess
sent
us
did
a
new
release
and
they
have
a
slightly
different
package
archival
strategy
than
other
Oasis.
So,
for
example,
Debian
has
an
archive
of
all
packages,
both
current
and
historical,
and
so
we
pulled
from
their
CentOS
appears.
A
To
only
is
how
a
set
location
for
new
ones
and
for
archived
packages
doesn't
have
the
new
packages
in
the
archive
and
doesn't
and
also
seems
to
have
either
moved
or
we
were
pulling
from
the
wrong
place.
You
were
putting
from
like
a
non
fully
qualified
version,
I'm,
not
entirely
sure
what
happened
there,
but
anyway.
We
therefore
had
to
do
a
pretty
urgent
fix,
which
I
said
on
for
a
while,
which
I
apologize
for
to
get
to
change
the
default
locations
from
which
we
pulled
those
packages.
A
We
should
probably
bring
out
two
oldest
rows
and
not
not
have
like
different
configurations,
because
in
theory,
docker
is,
can
be
configured
completely
differently.
I
think
is
very
differently
on,
like
CentOS
and
on
Debian,
and
there
could
be
like
configurations
that
are
specific
to
each,
and
in
this
case
it
was
app.
Armor
policy
I,
believe
that
was
the
package
in
question.
A
D
A
A
One
of
the
challenges
that
we've
had
in
the
past
is
that,
like
it's,
been
a
little
tricky
to
track
down
where
the
where
the
packages
come
from,
like
Debian,
kicked
out
docker
and
then
docker
stopped
shipping
docker
and
all
this
sort
of
stuff.
So
it's
a
little
tricky
and
it's
another
reason
I'd
rather
get
to
tar.gz
that,
like
we
had
a
similar
problem
with
like
the
container
de
sécurité,
the
Runcie
security
issue,
where
I'm
like
it
would
be
much
easier
if
we
could.
A
Great
and
then
I
think
someone
posted
a
link
to
the
fact
that
the
ete
jobs
are
still
fighting
to
create
clusters,
which
I
think
is
actually
blocked
on
me.
So
there's
a
the
e
to
e
jobs,
the
first
one
that
runs
is
called
cups,
AWS
up/down,
and
you
can
see
that
we've
I'm
trying
to
get
so
cups
a
toss
up
down
runs,
are
very
quick
up
and
down
of
the
cluster
but
doesn't
run
the
et
tests.
A
But
yet
not
least,
the
note
labels,
but
there's
also
another
fix
or
another
PR,
that
I
sent
in
that
I
changes
the
name
of
the
image
when
you're
reloading
a
saved
talker
save
tar
file,
which
is
what
we
do
in
CI
I'm
in
a
specific
to
CI,
but
was
another
blocker
on
the
on
the
up
down
test.
So
it
looks
like
we're
closer
now.
A
We
just
have
something
going
wrong,
which
I
will
I
will
look
at,
but
then
yes,
I
would
love
to
get
all
that
you
tease
I,
don't
forget
that
is
basically
running
again
and
then
we
can
sort
of
start
to
inspire
our
goal.
Configurations
I
also
want
to
get
us
running
on.
We
previously
pinned
to
115,
which
was
great
because
we
had
a.
We
were
incompatible
with
116
because
of
the
label
break.
A
A
A
Don't
think
he
wants
to
say
anything.
Sorry
on
the
previous
topic.
We
were
just
talking
about
about
CentOS
and
about
eatery,
and
all
that,
okay,
on
calico
I
think
there
are
three
issues
which
we
would
like
I
think
we
did
a
new
version
of
the
calico,
manifest
in
cops,
114
and
I.
Think
it's
closing
or
we've
picked
up
three
problems.
A
There's
a
log
spam
problem
where
every
three
minutes
it
logs
our
coop
controller
manager,
logs
some
errors
about
a
new
bit
of
new
bit
of
calico,
which
is
doing
it's
using
exact
volume.
So
I
can't
wear
the
co
volume,
plugins
I,
guess:
flex
volumes
flex
volumes.
Thank
you!
Yes,
so
that
logs
every
that
that
logs
some
spammy
looking
message
are
quite
frightening,
but
actually
look
benign,
but
there's
that
there
is
a
fix
in
the
newest
calico
component,
which
we
could
advanced.
So
me
to
the
other
two
issues.
A
A
That's
merge,
that's
not
that
one's
done.
We
just
have
the
other
two
problems,
which
is
the
log
spam
and
this
potentially
scary
we
kind
of
put
up
with
it
or
we
can
update
to
a
newer
version
and
then
the
other
issue
is
that
other
directories.
But
let's
talk
about
the
logs
fine
one
I,
don't
have
anyone.
Yes,
I!
Think.
A
A
B
A
A
Okay,
similar
issue,
no
longer
a
114
blocker
with
fingers
crossed,
but
then
the
this
is
the
problem
about
the
Flex
volume
directory.
The
volume
plug-in
there
and
also
the
C&I
bender,
both
need
to
both
are
written
to
by
calico
and
also
exact
from
there
in
the
Flex
volume.
And
so
it's
hard
to
find
a
or
the
default
directories.
Don't
work
on
all
os's
oldest
rows
and
we
currently
have
some
logic
in
node
up
which
switches
the
locations
based
on
some
based
on
the
distro.
We
don't
know
the
distro
when
we're
building
the
manifests.
A
A
Next
topic,
we
have
some
AM
eyes
that
went
out
or
hopefully
we'll
go
out
c1
into
the
Alpha
Channel
and
I
wanted
to
make
sure
that
everyone
realized.
Thank
you
to
Peter
for
emerging,
that
I
want
to
make
sure
that
everyone
realized
or
Peter,
and
my
former
Jeanette
I,
want
to
11
realize
that
this
those
a.m.
eyes
are
now
coming
from
this
image
builder
unification
project.
So
the
if
you
want
to
know
where
they
come
from,
please
click
through
the
link.
A
It's
something
like
Ranieri,
SIG's
image
builder,
on
github
image
builder
on
github
and
I
will
update
the
older
location
to
point
to
the
newer
location,
but
we
basically,
we
did
a
full
import
of
the
old
repo
into
the
subdirectory
of
the
new
repo
and
we
are
building
in
there
and
doing
things
like
updating
in
4k
log
and
hopefully
we'll
get
all
these
different
projects
or
like
pulling
together
in
the
same
direction,
and
we
have
a.m.
eyes
for
115.
Now
there
are
currently
just
I.
A
Don't
think
there
was
any
change
in
115
versus
114,
even
on
the
darker
version,
I
think
in
116
now
that's
released.
Hopefully
they
have
documented,
which
docker
version
they
are
using
and
we
can
do
a
fairly
big
docker
bumping,
I.
Guess
at
this
point
and
then
maybe
container
D
optionally
in
117
I'll
see
how
that
goes.
B
Right,
that's
what
I'm
here
for
alright!
So
as
you
can
see
on
the
agenda
over
there,
I've
been
working
on
the
docks
site
and
just
doing
some
general
docks
cleanup.
So
if
you
look
at
that
PR
it's,
it
continues
to
grow
a
bit,
but
the
main
idea
is
to
really
get
us
some
actual
cops
docks
site
rather
than
having
people
dig
through
the
docks
in
our
repo.
B
You
know
those
are
great,
but
you
know
that
we
have
so
many
and
they're
so
spread
out,
so
I've
been
trying
to
kind
of
refactor
them
a
bit
and-
and
you
know,
kind
of
just
start
us
with
a
site
here-
I'll
actually
share
real
quick.
You
can
obviously
click
the
link,
but
this
is
what
we're
looking
at
so
far,
I've
tried
to
organize
it
a
bit
under
you
know
some
getting
started
operations.
B
Some
of
this
still
needs
to
be
cleaned
up
even
more,
but
I
was
really
trying
to
just
do
a
first
major
pass
to
kind
of
make
this
somewhat
manageable,
so
we
can
start
publishing
it
and
everything
so
feel
free
to
poke
around.
This
is
obviously
just
a
temporary
domain,
but
you
know
I
just
wanted
to
have
it
kind
of
end-to-end
working,
so
I
can
show
people.
You
know
a
few
of
the
options
so
feel
free
to
check
this
out,
throw
some
thoughts
on
the
PR.
B
As
for
next
steps,
this
was
there's
been
a
number
of
people
that
put
work
into
this
over
the
past
couple
years.
Really
that
you
know
PRS
have
come
and
gone.
So
thank
you
to
everyone
that
has
previously
worked
very
hard
on
our
docks
and
so
I.
Think
next
steps
will
be
I,
gonna,
I'm
working
on
looking
into
getting
us
an
actual
official
sub
project,
site
and
I.
Think
Justin.
A
A
B
Through
some
photos,
okay,
yeah
and
then
we
get
a
necklace
I
site,
which
is
a
little
more
automated
than
using
github
pages.
I,
don't
know
if
I
has
a
strong
opinion,
but
that's
what
I
was
using
for
this
and
it
works
really
well
so
I
was
considering
going
that
route
rather
than
doing
all
the
weird
github
pages
stuff.
B
Search
works
on
this.
How
does
it
work
I
think
it
generates
like
a
static,
JavaScript
manifest
or
something
I.
Don't
know
it's.
It's
that's
yeah
yeah
yeah!
So
it's
it's
pretty
neat
I
was
also
going
to
explore.
It
looks
like
the
rest,
like
kubernetes,
kubernetes
and
most
official
sub
project
sites.
Ingress
doesn't
have
an
official
sub
project
site,
so
it's
trying
to
just
figure
out
what
the
pattern
is
in
the
community
and
most
of
them
have
actually
gone
with
Hugo
for
their
generation,
but
this
uses
make
pages
from
Python.
B
So
you
know,
I
was
just
gonna,
give
it
a
shot
just
to
see
you
know
comparatively.
If
we
can
match
the
community
in
some
form,
but
still
be
our
unique
cop
selves.
So
I'll
keep
everyone
updated
and
feel
free
to
throw
any
comments
on
there,
but
I'm
hoping
maybe
we'll
get
this
merged
in
at
some
point
soon,
and
then
we
can
iterate
on
it
and
then
once
it
looks
you
know
relatively
good.
We
can
actually
then
start.
You
know
pushing
it
as
a
I.
B
A
Us
do
this,
like
as
part
of
a
well
yeah
I
wants
to
figure
out
how
we
do
tie
it
into
the
release
strategy.
Right
like
we,
don't
sorry
want
new
features
showing
up
there.
I
know
that's
cause
some
people
challenges
in
the
past,
but
we
want
like
typo
fixes
getting
in
there.
So
I
yeah,
we'll
figure
it
out.
Yeah.
A
A
Will
move
on
to
the
next
topic?
If
if,
unless
anyone
wants
to
say
anything
more
z-table,
are
you
here,
sorry
I'm
we're
still
seeing
mics?
Sorry,
that's
right!
I,
don't
see
you
here.
So
it's
a
couple
of
pprs
that
need
review
and
I
will
look
at
those.
If
other
people
don't
beat
me
to
them,
looks
like
one
for
ED
CD
manager
and
some
for
OpenStack,
which.
A
A
A
A
A
All
right:
well,
that's
so
we
did
in
the
past
two
weeks
or
two
weeks,
ish
the
cops
2013
one
113
two
with
CentOS
fixes.
We
got
115
zero
alpha
one
out,
I,
think
just
before.
That's
this
week's
meeting
or
two
weeks
meeting
and
we
have
some
new
a.m.
eyes
in
the
Alpha
Channel,
including
the
115,
Am,
I
and
so
little
thing
we
really
need
to
get
for
the
next
few
weeks.
The
thing
we
really
need
to
hunt
is
the
one
14-0
blocker
I
was
thinking.
A
We
had
to
do
all
these
things,
but
Mike
may
have
like
made
my
day
by
suggesting
that
we
don't
do
that
until
1:15.
So
we
can
kick
that.
Can
one
little
step
down
the
road
which
would
be
great
but
yeah
I
think
it's
time
to
do.
114
and
calico
looked
like
the
only
blocker
or
the
Calico
related
issues,
and
so
hopefully
we
can
now
do
that,
and
then
we
should
do
one
14-0
beta,
1
I,
don't
yeah
I'll
have
a
look
at
I'll
debate.
Whether
like
the
Calico
update,
is
a
blocker
for
those
for
beta.
A
A
There
was
a
blocking
PR,
which
I
think
has
been
either
some
blocking
PRS,
but-
and
we
should
point
out
that,
like
not
all
cloud
work
or
some
car
providers
want
people
to
add
labels,
but
the
that
should
be
basically
ready
to
go
out
at
alpha
stage
and
also
hopefully
we
can
get
that
under
ete
testing
with
kubernetes,
116
and
117,
and
we
probably
should
and
then
sort
of
just
general
like
promotion.
We
should
promote
the
a.m.
eyes
from
alpha
to
stable
and
the
community's
versions
from
alpha
stable.
A
A
B
If
you're
curious,
where
things
are
some
things,
we
won't
back
board.
Typically,
we
won't
back
port
to
like
113,
unless
we
really,
you
know
unless
it's
a
critical
fix
like
the
stuff
we
mentioned
at
the
beginning
of
the
call.
But
you
know
we
many
things:
we've
been
back
porting
to
114,
because
they've
been
relatively
small
for
the
most
part.
You
know
like
fixes
that
it's
like.
Well,
it's
actually
really
better.
We
get
that
out
now,
you
know,
but
then
typically
things
will
go
into
master
and
then
we'll
cherry-pick
into
115.
D
A
Would
be
great
if
we
can
get
there
I'd
like
to
say
is
I
feel
like
this
has
been
not
me
that
has
done
this
I've
occasionally
pushed
the
release
button,
which
I
would
like
to
like
make
sure
I
am
NOT
the
bottleneck,
but
like
big
shout-out
to
everyone
actually
on.
This
call
like
thank
you
Mike,
and
we
go
and
sit
who
have
like
done:
wonderful
jobs
and
others
of
I'm
Peter
you
as
well.
Thank
you.
Everyone
on
the
call
I
should
name
you
each,
but
yes,
it
has
been
like
really
great
to
get
like.
A
Most
of
those
cherry
fixes
are
not
done
by
me.
Mostly
approvals
are
no
longer
done
by
me.
So
thank
you
for
all
of
that
and
yes,
we
will
hopefully
get
onto
the
Cates
artifacts
working
group,
Cates
infra
train,
so
that
we
may
be
able
to
release
things
without
having
to
push
buttons,
which
would
be
great
yes
exactly.
We
were
at
least
we
released
I.
Think
there's
a
PR
out
there
to
like
add
artifacts
as
another
mirror,
artifact
stop
gates.