►
From YouTube: Kubernetes SIG Testing 2019-01-15
Description
A
Right
hi,
everybody
today
is
Tuesday
January
15th.
This
is
the
weekly
stink
testing
meeting
that
is
being
recorded
and
will
be
publicly
posted
to
YouTube.
So
please
remember
that
what
you
say
here
is
held
to
the
kubernetes
code
of
conduct,
which
basically
boils
down
to
you
to
be
a
jerk.
I
am
aaron
of
sagir
today
on
our
agenda.
B
Okay,
thanks
Aaron
and
just
for
your
benefit
and
for
everyone
else's
benefit.
It's
pronounced
for
them
is
the
way
we
usually
pronounce
it.
So
the
FS
Island
accent
is
on
the
second
syllable
and
it's
Norwegian
if
you're
curious,
so
I'm
told
so
yeah
thanks
for
giving
me
some
time.
I
know
this
is
a
short
meeting
and
I'm
going
to
be
I'm.
Gonna.
B
I
know
I've
gotten
some
feedback
on
previous
calls
and
in
slack
and
I,
really
appreciate
that
it's
all
been
very
helpful
and
very
encouraging
and
I
just
want
to
so
I
just
I
kind
of
wanted
to
get
a
little
bit
more
in
depth
into
that
before
I,
keep
you
know
continuing
before
I
continue
working
on
this
stuff.
So,
let's
see
here
I'm
trying
to
share
my
screen.
B
But
I
decided
against
that,
because
I
I
know
there's
kind
of
this
tradition
of
going
with
a
nautical
names.
I
tried
to
think
of
what
was
a
nautical
way
of
calling
somebody
you
know,
get
his
British
line
if
you're
not
familiar
get
his
British
slang,
it's
kind
of
an
insult,
so
I
thought
what's
the
equivalent
kind
of
nautical
insult
and
you
know
scallywag
seemed
seemed
appropriate
anyway,
not
married
to
that.
If
anybody
has
better
ideas,
but
it
just
seemed
kind
of
catchy
and
fun
so
and
I
see
I'm
getting
some
laughs
so
I'll.
B
Take
that
as
a
good
sign
Wow.
So
github
is
good,
I'm.
Sure
I.
Don't
need
to
tell
you
folks
that,
especially
if
you've
been
working
on
prowl
prowl
is
very
much.
You
know
senator
Don
on
github,
but
you
know,
there's
been
some
talk
in
I've
heard.
I
know
it
furred
in
some
of
the
past
conversations
before
I
kind
of
became.
A
part
of
this
conversation
on
some
past
calls.
You
folks
have
talked
about
supporting
other
get
providers
and
when
I
brought
it
up,
it
seems
like
there
is
an
appetite
for
this.
B
My
interest
in
doing
this
is
that
I
work
primarily
I
work
for
a
company
called
cloud
bees
and
I
work,
primarily
on
at
Jenkins
X,
which
I
take
it.
Some
of
you
are
familiar
with,
and
I
won't
get
into
what
Jenkins
X
is
just
for
now,
for
our
purposes
here
know
that
we're
using
prowl
for
something
that
we're
calling
serverless
Jenkins
but
essentially
proud
but
proud
as
an
essential
component
of
that,
and
so
what
we
would
like
is
we
would
like
for
proud
to
be.
You
know
flexible
enough
to
support
other
good
providers.
B
Besides
the
site's
just
get
not
because
github
is
bad,
but
just
because
people
use
other
stuff
and
it's
night
it
would
be
nice
to
to
be
flexible.
So
you
know
that
could
include
github
enterprise,
which
is
it's
an
obvious
first
choice.
You
know
bitbucket
server,
get
lab
and
then
the
world
of
dogs
and
get
tea
and
whatever
you
know
whatever
other
get
provider
is
your
favorite,
so
so
scallywag
is,
is
what
I've
is
it
sort
of
a
thing
I've
created
to
kind
of
to
answer
this
problem?
B
So
the
idea
right
now,
all
every
get
action
that
proud
performs
is
very
tightly
coupled
to
github.
Now
there
is
a
little
bit
of
abstraction
in
that
there's
some
there's.
Like
a
you
know.
There
are
some
methods
you
can
invoke
like
update
issue,
which
kind
of
abstract
away
that
the
mechanics
of
connecting
to
the
github
API
and
updating
an
issue,
for
example,
but
it's
not
generic
enough
to
just
plug
in
another
different
get
provider.
B
So
the
idea
is
that
we
have
a
free-standing
highly
available
service,
G
RPC
server
that
implements
a
some
well-defined
interface
of
git
commands.
So
this
could
be
things
like
you
know,
create
issue,
update,
pull
requests,
create,
pull
requests
whatever.
That
is,
are
the
kinds
of
things
that
proud
does
and
really
what
the
the
goal
here
would
be
to
completely
pull
to
completely
rip
out.
B
All
of
the
specific
all
of
the
get
provider
specific
logic,
so
that
all
other
components
of
proud
that
want
to
do
get
operations-
all
they
have
to
know
is
I
just
want
to
update
an
issue.
Here's
the
update
I
want
to
make
here.
Maybe
here's
some
data
about
the
issue-
that's
very
generic,
that's
not
specific
to
the
git
provider
and
to
kind
of
build
a
set
of
abstractions
and
actions
that
allow
us
to.
B
You
know
to
invoke
this
interface
that
will
then
go
and
go
and
just
use
whatever
get
provider
is
configured
to
to
use
it
so,
and
the
idea
here
is
that
if
we
use
G
RPC,
we
can
actually
define
that
interface
and
in
these
dot
proto
files,
and
then
you
know
we
could
sort
of
have
like
a
that.
My
initial
idea
was
that
there
would
be
kind
of
like
a
kitchen
sink
get
provider
that
supports
github
and
get
enterprise
and
bitbucket
and
get
lab,
or
you
know
whatever
else
makes
sense
whatever
folks
want.
B
But
then,
if
there
is
other
things
other
stuff
that
folks
want
to
support,
they
could
take
these
protobuf
files
and
just
generate
their
own
server
and
kind
of
implement
it.
However,
they
want-
and
so
I've
been
just
doing,
some
initial
prototyping.
On
my
you
know
my
fork
of
the
test.
Infra
repo
I've
also
been
working
a
little
bit
on
kind
of
re
pulling
some
of
the
you
know.
B
We
have
a
lot
of
this
generic
get
provider,
logic
and
Jenkins
X
and
I'm
working
actually
working
on
pulling
that
out
into
an
external
library
that
could
be
used
by
Jenkins,
X
and
prowl,
and
you
know
whoever
else
wanted
to,
but
basically
it
would
just
be
a
set
of
implementations
of
of
a
generic
get
interface.
You
know
so
there
were
in
this
go
gets
repo.
There
will
be
implementations
for
github
and
get
lab,
etc.
As
the
the
kind
of
specific
implementations
that
could
be
used
by
you
know
an
implementation
of
scallywag,
for
example.
B
So
if
you're
interested,
you
know
this
is
here's.
My
info,
including
links
to
these
repos
I'm,
happy
to
share
this
deck
and
that
this
is
really
the
end
of
its
just
four
slides
so
and
let's
see
I'm
still
10
minutes.
Ok,
so
that's
that's
kind
of
what
I
have
on
again,
like
I
said
I
wanted
to
keep
it
short,
I've
kind
of
wanted
to
just
let
you
folks
know
what
I
was
doing
and
get
some
thoughts.
So
if
there
are
any
questions,
comments,
snide
remarks,
you
know,
let
me
have
it
I
suppose
it
seems
really.
C
B
D
Okay,
well,
this
is
Andrew.
So
this
like
been
said,
this
looks
really
interesting.
So
I
was
involved
with
CSI
two
years
ago
for
about
a
year
and
I
know
that
that
can
be
real
fun,
trying
to
create
an
abstraction
layer
on
top
of
different
providers.
That
kind
of
do
the
same
thing,
but
have
their
own
little.
D
And
the
second
question
is
so
today:
I
mean:
where
would
the,
for
example,
the
github
server
run
that
github
server
provider,
rather
since
github
I
mean
I'm
about
to
make
a
statement
that
could
be
completely
wrong,
so
somebody
can
correct
me:
github,
doesn't
support
G
RPC
directly
for
its
api's
am
I
wrong.
Last
I
looked
they
didn't,
and
so
if
this
is
sort
of
an
abstraction
layer,
sort
of
middleware,
fortunately,
unfortunately
middleware
seems
in
my
life
for
about
ten
years
now.
B
B
One
of
those
components
like
hook
or
you
know
some
of
the
other
I
mean
hook-
is
the
big
one
I
think
about
a
lot,
but
you
know
the
I'm
blanking
on
the
names
of
the
other
ones,
but
the
idea
would
be
that
well,
so
that
an
obvious
pattern
would
be
to
run
it
like
that,
where
it's
just
in
a
pod,
but
really
the
idea
is
that
you
would
tell
you
would
tell
book,
or
you
know
whatever
other
component
figure
out
exactly
how
that
should
work,
but
basically
you
would
specify
okay,
this
is
the
endpoint
to
my
get
provider.
B
Maybe
it's
in
the
cluster.
Maybe
it's
somewhere
else
doesn't
really
matter.
It's
a
gr
PC
server
somewhere
that
implements
some
interface
and
that's
all
I
really
know.
But
you
know
we
could.
The
idea
is
at
least
initially
to
make
it.
So
it's
just
something
that
you
can
configure
and
specify
and
prowl
and
deploy
with
the
other
components.
B
Cluster
and
so
yeah
and
you're
right
that
it's
you
know,
the
G
RPC
piece
is
really
just
a
way
of
making
it
it's
just
a
tool
for
creating
a
generic
interface
that
you
know
that
I
can
predict
and
know.
Okay,
there's
going
to
be
this
G
RPC,
interface
and
I.
Can
I
can
connect
to
it
in
well-known
ways
without
knowing
the
underlying
implementation
details,
but.
D
D
May
be
some
features
not
supported
on
some
get
providers,
so
I
would
suggest.
Maybe
yeah
I
think
yeah
I
really
like
this
idea,
but
CSI
is
kind
of
it
just
hit
one
point
on
Jen
in
December.
You
might
take
a
look
at
that
because
they
do
have
the
ability
to
advertise
features
from
the
provider.
So,
let's
the
consumer
know
what
features
are
supported.
I
think
perhaps
that
would
be
a
good
idea
for
this.
Okay
I'm.
So
you
know
the
the
feature
set
could
be
dynamic.
B
B
A
C
Just
super
quickly,
so
I
know
if
you're
familiar
with
it
short
version,
is
kind
of.
What's
us
to
run
local
tributaries
clusters,
you
stock
containers
I've
had
a
few
people
at
cube
con
on
the
more
since
then
reach
out
requesting
that
we
get
a
sub-project
meetings.
We
have
a
place
to
discuss
this
stuff.
There's
you'd
want
to
do
everything
and
anything
with
it.
So
if
you're
interested
I've
put
out
a
win
is
good
trying
to
figure
out
how
we
can
arrange
a
meeting
with
all
these
different
time
zones.
C
A
Cool
okay,
thank
you.
So
next
I
wanted
to
talk
a
little
bit
about
what
our
plans
are
for
B
114.
This
is
going
to
be
super,
rarely
because
I
don't
actually
know-
and
this
is
my
sixth
meeting
of
the
day,
so
I
actually
haven't
had
much
time
to
look
at
it.
Historically,
what
I've
tried
to
do
is
to
push
everything
that
this
group
wants
to
accomplish
within
a
quarter
into
a
milestone
in
the
testing
for
repo.
Now,
that
kind
is
branching
out
into
its
own.
Some
projects,
I'm
gonna,
have
to
go.
A
And
do
within
114
and
then
what
we
has
a
segment
like
utilize
time
for
I
know.
There
have
been
discussion
of
like
moving
kind
to
a
pre
submit
and
doing
potentially
making
their
pre
submit
blocking.
Although
we
probably
want
to
quantify
what
merge
blocking
criteria
should
be
for
participants,
things
of
that
nature,
so
I
know
and
has
put
together
a
project
board
or
something
for
kind.
D
C
Now
that
we've
made
the
big
multi
node
changes,
we
should
be
able
to
kind
of
stop
breaking
the
command
line,
guys
so
that
everyone
can
build
Cir
they
do.
Everyone
will
also
be
collaborating
what's
a
closer
lifecycle
as
much
as
they
still
want
to
make
it
a
pretty
scimitar,
not
I'm,
not
going
to
push
that,
but
if
they
are
pushing
it,
I
will
help.
D
C
There
is
one
more
thing
where
they
are
looking
to:
GA
high
availability
clusters
with
key
Batum
and
kind
is
looking
like
the
most
viable
short-term
route
to
potentially
get
actual
CI
up
for
that
and
giing
that
is
probably
pending
on
having
CI.
So
we
will
attempt
to
do
that
this
cycle
as
well,
because
otherwise
it
should
probably
be
good
enough
for
da,
but
there's
no
testicle
right
now.
D
Hey
Ben,
you
might
want
to
touch
base
with
me
architect,
base
with
you,
I'd,
been
doing
some
work
and
standing
up
high
availability
clusters
successfully
with
yakety
internally
and
I
was
toying
around
with
a
cap
to
introduce
to
something
like
cube
ADM.
It
uses
an
external
discovery
service
like
@zd
discovery
to
do
it,
but
it
it
can
stand
up.
You
know
100
nodes
in
a
few
minutes,
and
so
maybe
some
similar
logic
could
be
used
in
kind.
C
C
All
these
problems
for
the
no
clouds
but
yeah.
If
you
have
something
like
that,
getting
it
NCI
would
be
great
because
Q
Batum
right
now
does
not
have
CI
for
H
a,
and
if
you
can
get
a
che
CI
that
uses
Q
page
names
a
che
mechanisms.
They
have
a
few
people
working
pretty
hard
on
making
sure
it
works,
but
like
we
need
CI
signaler,
we
can't
get
properly.
Okay,.
A
So
then
the
other
thing
is
over
in
the
testing
for
repo.
This
is
the
the
114
milestone.
There
are
24
of
an
issues
in
it.
These
are
all
of
our
milestones,
so
we
have
V
114
as
this
quarter.
V
115
is
a
quarter
after
that.
2019
phones
would
be
anything
just
for
2019
in
general
and
someday
someday
someday.
A
We
will
get
to
these,
so
my
concern
is
that
the
issues
that
we
have
that
don't
have
114
milestone,
they're
like
250
of
them,
which
I
kind
of
feel
like
if
these
issues
aren't
getting
planned
to
be
done,
a
milestone.
It's
a
sign
that
they're
not
happening,
which
may
be
just
representative
of
this
critics
bandwidth,
but
I
feel
like
they're
kind
of
needs
to
be
more
in
here.
A
If
there
is
but
I
might
be
wrong
if
I
just
kind
of
scan
through
this
and
I
know
that
some
of
the
stuff
that
I
am
personally
motivated
to
work
on
as
the
release
team
lead,
is
to
focus
on
like
job
consistency
for
release
and
things
like
making
sure
that
we're
starting
to
actively
deflate
kubernetes
tests.
There's
a
wide-area
deflate
label
that
I
can
use
for
this.
It's
also
tied
to
a
deflating
board
but
I
view
like
plowing
forward
on
this
board
as
orthogonal
or
tangential
to
state
testing.
A
I
just
wanted
to
call
it
out
there
lately,
but
I
do
think
so.
Testing
like
we're
gonna
improve
the
alert
configuration
to
help
out
the
release
team
I'm,
definitely
going
to
finally
remove
any
of
jobs
that
don't
spark
joy
and
have
been
continuously
failing
for
over
400
days.
At
this
point,
I'm
doing
it
right
right
and
then.
A
So
there's
this
eks
provider
and
I
think
Senna
had
talked
about
like
it
could
test
v2
at
one
point
I
that
we
probably
there
is
a
conformance
image
that
is
built
as
part
of
the
upstream
kubernetes
repository,
but
it's
not
built
by
default.
I,
don't
know
why
and
you
don't
have
a
job
that
continuously
runs
conformance
tests
using
that
conformance
image.
I
would
look
for
us
to
get
to
that
point.
A
So
this
is
me
just
sort
of
trying
to
reverse
engineer
what
a
set
of
consistent
priorities
or
some
kind
of
narrative
based
on
the
issues
that
we
have
in
here.
If
people
feel
like
there's,
nothing
isn't
being
represented
in
here,
please
please
add
it,
because
what's
going
to
happen
is
when
it
comes
time
to
share
with
the
community
like
what
our
plans
are
I'm
going
to
be
going
more
or
less
off
of
what's
in
here,
so
it
makes
sense
to
everybody.
A
A
I
will
share
briefly.
It
will
go
out
of
screen
sharing,
for
this.
I
have
tried
over
time
to
see
if
this
group
of
people
is
can
get
any
better
about
calling
their
shots
before
they
take
their
shots
and
I
have
found
it
staggeringly
difficult
to
do,
and
I
have
actually
kind
of
come
to
accept
that
the
right
stuff
generally
gets
done
anyway,
but
I
would
it
would
still
be
nice
to
kind
of
like
call
the
bigger
level
shots
out
first
and
so
a
lot
of
the
250
open
issues
that
aren't
milestones.
A
Maybe
requests
from
the
community
that
we
just
kind
of
don't
have
the
resources
to
prioritize,
which
is
maybe
why
they're
not
milestone
or
it's
kind
of
unclear.
If
there's
a
need
to
address
that
particular
thing,
but
I
feel
like
oftentimes,
many
of
them
valid
requests
and
they
do
sound
nice.
But
there
are
other
things
that
just
seem
a
little
more
important.
Given
the
resources
that
we
have
a.
D
That's
who's
math
suggests
me
and
I
I
don't
mind
taking
this
or
somebody
else
can
but
either
like
a
Google
sheet
or
something.
So
we
don't
in
fact
the
issues
directly,
but
just
quickly
and
figuring
out
which
issues
are
ex-old
and
dumping
them
into
a
bucket,
and
then
all
the
other
issues
may
be
paying
back
to
the
author
to
see
if
they
still
find
this
to
be
important.
D
Its
I
know
Eric
has
the
automation
set
up
that
eventually
does
closeout
issues
or
freezes
them
I
mean
and
it's
effectively
when
we
use
beta
bot
for
take
care
of
that
for
us,
okay,
so
I
guess
what
I'm
asking
is?
What
are
you
asking
I
sense
that
you
need
help
right
and
so
I'm
trying
to
figure
out
what
kind
of
help
do
you
need?
So
if
I
can
be
the
one
to
help
or
somebody
else
can
but
I'm,
not
sure
I
understand
exactly
the
help
that
you
want.
I
think.
C
A
That's
like,
if
you
want
to
know
what
to
work
on
taking
a
look
at
the
114
milestone
and
anything
that
doesn't
have
a
person
assigned
probably
could
use
some
doing,
because
that's
a
sign
that
that's
an
issue
that
we
as
a
group.
Typically,
we
want
to
try
and
get
done
this
quarter,
but
nobody's
actively
signed
up
for
it
yeah
and
then
yes,
the
other
thing
is:
if
you're
trying
to
take
a
shot,
it'd
be
great
if
you
could
call
the
shot.
A
This
isn't
about
trying
to
impose
a
bunch
of
additional
friction,
but
just
trying
to
make
it
easier
to
sort
of
share
like
what
we're
planning
on
doing
from
a
community
perspective.
Okay,.
A
Are
you
are
welcome
to
go
through?
There
is
actually
an
open
issue
somewhere.
I
just
kicked
out
of
the
money
issue
for
issues.
It's
like
it's
a
meta
issue.
It's
one
of
our
goals
would
be
to
one
day
get
200
open
issues
in
this
repo,
so
I
do
from
time
to
time
go
through
and
sort
issues
by
at
least
recently
updated
and
see.
If,
to
my
mind
they
still
seem
applicable
or
not
just
so,
we
need
a
kind
of
a
triage
water
workload.
A
Is
it's
it's
whatever
makes
sense
to
you,
but
yeah,
I,
I,
guess
I
kind
of
feel
like
I'm
gonna,
go
ahead
and
add
the
tide
action
history
page,
for
example,
there's
a
tracking
issue
for
that
I'm
gonna.
Add
that
to
where
114
milestone,
because
I
think
Cole's
been
actively
working
on
that
yeah
that'll
definitely
be
done
this
for
ya.
Just
things
like
that,
if
you
know
you're
working
on
them
or
you
notice,
somebody's
working
on
something
adding
it
to
the
monster
will
be
helpful
because
then
we've
got
to
go
back
real
quickly,
just
see.