►
From YouTube: Kubernetes SIG Testing 2019-03-26
Description
A
A
A
C
A
A
B
C
B
Tm
or
slash
of
it,
slash
approve
just
there,
but
actually
the
clicking.
The
approve
radio
button
on
your
submitting
a
github
review
right
now
that
behaves
as
a
as
a
slash,
improved
command,
but
we
are
proposing
to
also
make
it
behave
as
they
/l
g
TM
command,
and
that's
because
right
now
pretty
many
people
are
forgetting
to
put
slash
approve
when
they
actually
mean
it
and
they're
they're
having
to
come
back
to
do
that
again.
If
they
haven't
used
to
get
up
review
or
explicitly
slash
proved.
B
D
A
Can
trigger
how
they
feel
about
it.
I
think
it's
cool,
because
it
simplifies
everything
to
one
button.
Click.
If
you
wait
to
prove
it's
gonna,
do
what
you
would
expect
as
a
normal
kid.
Perky
I
dislike
it
as
somebody
who
wants
to
try
and
keep
and
the
health
of
our
community
and
better
understand
how
many
people
are
approving
PRS.
A
So
it
sounds
like
everybody
here
is
cool
with
it
for
tests,
infra
I
would
still
run
it
by
contributes,
even
though
we're
doing
it
in
our
little
playground,
because
I
think
right,
it
is
kind
of
surprising
so
finding
a
way
to
communicate
that
to
people
would
be
helpful.
Maybe
a
heads
up
to
thank
you
differently.
Maybe
we
want
to
update
our
contributing
about
empty
box
to
describe
this.
C
B
E
A
A
A
F
Okay,
so
I
am
proposing
and
also
already
did,
but
we
turned
down,
tops
and
replace
it
with
sniffing
IDs.
The
upshot
being
that
we
will
no
longer
have
related
issues
or
passages
of
which
we
have
had
several
in
the
past
and
also
the
build
IDs
are
going
to
be
a
lot
longer
and
less
comprehensible,
although
in
many
cases
I
think
they
weren't
very
comprehensible
already
so
I
already
did
that
and
we
can't
undo
it.
So
it's
quite
it's
kind
of
a
thing
was
done:
yeah,
okay,.
D
D
F
Monotonic,
so
anything
with
the
pendulum
continuing
to
increase
will
be
fine
as
long
as
you
don't
expect
them
to
be
sequential.
All
of
our
tooling
handles
this
I,
don't
know
of
any
tooling
that
isn't
ours
and
are
we
are
we
officially
deprecating
saw?
What's
our
stance
there,
I
do
not
have
any
plans
to
deprecate
touch.
I,
just
don't
intend
to
use
it.
I
don't
know
if
anyone
else
has
big
plans,
and
would
you
suggest
anybody
else
using
it
I
don't
think,
there's
any
amount.
G
F
A
It's
a
bad
idea.
We
probably
ought
to
deprecated
it
I'm,
not
asking
that
you
necessarily
do
that,
but
I
think
we
should
have
been
initiated
to
talk
about
it
and
then
we
can
start
the
conversation
of
hey.
Does
anybody
actually
use
this?
That's
sort
of
deal
super
cool
that
that
happened
and
nobody
noticed
okay,
Cole
and
Cosman.
If
I'm
pronouncing
your
name
right,
would
you
like
to
talk
to
us
about
your
proposal
to
add
a
Tecton
pipeline
controller.
C
To
prowl
yeah,
so
you
pronounce
correct
my
name.
Yes,
this
is
a
initial
draft
design
to
add
a
new
controller
for
Tecton
pipeline
in
prom.
So
what
started
basically
in
Jenkins
X
project
and
we
try
now
to
kind
of
generalize
it
and
make
it
more
useful
for
the
entire
community
and
there
are
still
quite
some
questions
open,
which
I
would
try
to
answer
tomorrow,
I'm
based
in
Europe,
so
I'm
meeting
in
the
dark
right
now,
but
try
to
clarify.
C
My
generation
and
start
to
an
external
service-
and
this
was
a
service
developed
in
Jenkins
X
and
has
still
its
ongoing
work
as
its
hired
level,
syntax
definition
for
the
pipeline
and
which
still
need
to
document
staff
and
so
on.
So
but
the
main
idea
is
that
we
have
did
these
two
modes
and
we
try
to
sort
out.
What's
the
best
for
the
like.
Let's
say,
entire
community
I
think
the
embedded
mode.
F
C
It
yeah
so
yeah
we
work
I,
guess
directly
with
the
Tecton
team.
We
have
two
engineers
they
are
originally
from
Jenkins.
They
work
on
the
Jenkins
declarative
pipeline
and
they
who
are
closely
we
tacked
on
to
add
new
features
to
the
pipeline
like
Tecton
pipeline
itself,
but
in
the
same
time
they
are
also
developing
these
higher-level
syntax.
C
That
I
think
that
I
mean
you
can
use
the
pipeline
CID
as
it
is,
but
like
for
a
normal,
that's
a
user
which
is
not
very
familiar
with
kubernetes
and
all
the
you
know
schema
the
finishes
and
whatever
it
can
be
a
bit
cumbersome
to
go
like
it's
a
bit
lower
level.
Let's
say,
and
also
you
have
things
like,
let's
say,
build
bags
and
you
know
splat
for
a
specific
configuration
and
so
on
which
it
can
be
maybe
be
difficult
to
set
up
on.
C
Let's
say:
I
want
to
see
Rd
itself,
so
you
need
to
deal
with
docker
images
and
everything.
So
you
having
a
high-level
syntax.
Did
it
make
sense?
I'm
not
probably
the
best
to
discuss
about
this
because
I'm
not
I'm,
not
responsible
for
these
higher-level
syntax.
But
if
you
want-
and
you
want
more
details
I
can
we
can
bring
grind
you
into
the
discussion
and
we
can
give
you
all
the
reasons
why.
D
C
D
C
Think
it's
a
bit
too
early
to
discuss
about
this,
so
maybe
the
best
is
to
define
an
interface
on
brow
level
to
be
able
to
like
hook
up.
Also
other
I,
don't
know
see
ICT
systems
if
they
want
to
define
their
own
pipeline
syntax
on
top
of
tacked
on
so
I'm,
not
sure
if
I
mean
we
will
be
very
happy
to
use
Jenkins
X
in
taxi
in
proud,
but
we
don't
want
to
push
this
as
the
default.
C
You
know
it's
not
the
idea
that
we
push
this,
but
this
is
what
we
have
right
now
and
kind
of
try
to
show
what
we
did.
But
we
can
generalize
this
interface
that
we
also
allow
other
CIC
system
if
they
want
to
implement
their
own
seed
syntax
on
top
so
and
then
we
can
choose
which
is
the
best
at
the
end,
but
there
is
an
ongoing
effort
to
standardize
this.
Yes,
please
sorry,
I.
H
H
I
think
what
we
have
at
the
moment
is
kind
of
a
case
of
we're
going
for
as
simple
as
possible
and
no
simpler,
so
it
might
be
slightly
more
verbose
than
a
user
would,
like
so
I
think
we
are
seeing
a
pattern
where
people
often
do
write
some
kind
of
language
on
top
of
it
and
I
think
that
that's
fine
for
what
Tecton
pipelines
is
trying
to
be.
But
that
being
said
like
we're,
definitely
open
to
improving
it.
H
If
we
can,
if
there
is
some
complication,
we
can
remove
so
I
think,
maybe
seeing
what
the
higher-level
syntax
looks
like
next
to
the
generated
Tecton
pipeline.
Syntax
might
be
something
that
would
help
with
that
yeah.
So
I
think
it's
it's!
It's
not
surprising
to
us
that
users
have
trouble
with
it
and
it's
not
necessarily
a
bad
thing,
but
we
would
like
to
improve
it
if
we
can.
B
H
C
D
I,
this
is
a
like
a
parallel
problem
to
projects
today
targeting
kubernetes
we're
just
embedding
the
pod
definition
in
there
and
often
that's
not
really
what
users
want
so
downstream
of
that
definition.
There
is
like
the
q-kidz
image
and
the
test
runners
and
whatnot,
and
so
right
now
we
are
sort
of
completely.
D
We
don't
really
delve
into
the
details
of
that
second
level
of
configuration,
that
is
more
user
friendly
and
so
I
wonder
if
we
would
want
to
use
a
similar
approach
here,
but
yeah
I,
guess
what
you're
proposing?
Is
it
a
more
generic
interface
to
delegate
some
part
of
the
pipeline
resolution
process
to
a
third-party.
D
C
C
C
G
But
I
mean
the
yeah:
do
you
know
I
mean
I
haven't
had
a
chance
yet
to
review
the
design
doc
I
want
to,
but
I'd
sort
of
be
inclined
to
think
about.
You
know
if,
if
we
wanted
to
have
like
an
abstraction
layer
that
makes
it
an
easier
to
write
brow,
job
I
would
be
inclined
to
think
about
right.
You
know
adding
that
logic
more
into
hook
than
in
plank
right
in
the
sense
of
making.
The
logic
for
you
know,
did
whatever
having
some
language
that
makes
something
simpler.
G
Have
that
go
on
top
at
first
and
then
to
have
plink
continued
to
be
simple
and
read
the
more
verbose
language
that
I
think
is
similar
to
the
pattern
that
Tecton
is
using,
where
you
know
you've
just
create
the
pipeline
and
pipeline
Tecton
knows
how
to
deal
with
the
robot
or
post
pipeline
in
the
process
of
creating
the
you
know.
Details
of
Tecton
can
then
be
delegated
to
somebody,
but
I'm,
not
sure.
G
B
D
Controller
for
plank
yeah,
I,
yeah,
I,
guess
I'm,
like
resolved,
resolved
to
a
pipeline
definition
by
the
time
this
year
he's
actually
created,
and
then
the
pipeline
controller
you
know,
runs
through
the
rest
of
the
world
entirely.
I
think
it
was
one
of
the
questions.
I
had
I,
don't
think
like
the
data
flow
was
modeled
there.
Well.
B
Just
like
the
build
controller,
we
shouldn't
have
to
actually
touch
hook
or
anything
like
if
we
do
I
say
that's,
potentially
an
indication
that
we're
not
properly
generalized
yeah,
like
our
triggering
shouldn't,
have
to
change.
This
is
just
substituting
in
a
different
execution
platform.
Well,
I.
Think.
G
G
So
when
we
yeah
so
like
with
this
now
when
we
create
the
build
control
like
we
create
a
build
object
to
build
controller
is
responsible
for
running
that
object.
With
this
before
we
create
the
Tecton
object.
We're
now
having
the
build
object
have
to
be
delegating
to
somebody
getting
that
response
back
in
creating
the
actual
pipeline
thing
which
doesn't
to
me.
G
C
C
It's
for
me
this
multi
question.
We
want
to
embed
the
entire
Tecton
pipelines
here.
The
definition
inside
of
the
prow
job,
like
it's
done
today
for
build,
which
is
a
bit,
is
definitely
smaller
like
in
terms
of
resources.
I
mean
it's
also
here
possible.
You
can
embed
the
entire
pipeline.
Crt
definition
inside
of
router,
but
it
gets
probably
too
large.
I
would
say
well.
D
I
guess
that's!
That's
sort
of
the
question
is
to
like
the
data
flow
and
the
separation
of
concerns
of
who's.
Doing
what,
if
we're
writing
a
controller
that
asks
detect
on
to
execute
this
pipeline,
and
it
manages
how
that
pipeline
is
being
executed.
That
I
think
we
should
have
that
embedded.
The
same
way
that
build
is,
if
we're
writing
a
controller
that
asks
the
Jenkins
ex
to
do
something
for
us,
then
we
can
just
have
the
configuration,
the
name
or
whatever
identifier
would
help
us
tell
Jenkins
X
to
do
some
things.
D
A
Want
to
dive
in
here,
because
this
has
all
been
really
productive
discussion,
but
first
off
Cosmo
wanted
to
thank
you
for
showing
up
here
in
your
time
zone.
That
is
not
trivial.
So
thanks
a
bunch
for
that.
Second
I
think
also.
This
sort
of
discussion
is
why
we
wanted
to
try
and
have
this
discussion
before
investing
a
lot
of
time
in
the
implementing
code.
A
Sounds
like
now
that
we
have
the
dock
and
I
see
so
many
comments
on
the
dock,
that
that
would
be
a
good
place
to
iterate
and
if
we
need
to
have
continued
to
have
higher
bandwidth.
Discussion
like
this
might
be
worth
having
a
more
European
friendly
times
a
breakout
session,
which
we
are
happy
to
schedule
to
talk
this
through.
But
unless
there's
any
like
super
last-minute
thing,
you
think
would
help
us
get
to
consensus.
A
E
It's
it's
a
horrible
dad
joke!
It's
a
goal:
egg
replacement
for
Shelton
a
unit
as
Irish
like
it's
my
response
to.
Basically
it's
my
middle
finger
to
Steve
thumb,
standing
my
my
script
and
saying
he
prefers
going
so
I
guess:
I
proved
him
right.
He
decided
to
you
know,
tell
him
off
by
writing
a
cooling
version
of
it
and
it
you
know
the
shells
done:
Oh,
Sheldon,
Big,
Bang,
Theory,
bazinga
and
essentially
I
opened
a
PR
and
and
I'll
make
this
real
quick.
It's
my
first
substantial
PR!
E
That's
not
just
you
know,
test
job,
so
I'm
trying
to
learn
this
process,
I,
just
rebase
it
with
basil,
flying
it
and
everything
and
I
don't
even
know
if
it's
necessarily
appropriate
for
pest
infra.
It
seemed
like
something
that
could
be
useful
saw
that
Ben
had
done
a
lot
of
work
in
Cube
chest
to
I
stole
a
little
bit
of
that,
but
not
a
lot.
E
E
A
Feel
like
the
discussion
you
just
witnessed,
would
be
an
example
of
why
I
would
suggest
having
a
design,
doc
yeah
before
trying
to
land
up
massive
PR
on
us.
I
do
want
to
encourage
like
iteration
and
stuff
like
just,
and
this
is
me
making
stuff
about
a
whole
cloth.
I
feel
like
sure
landing.
Something
in
the
root
of
our
directory
may
carry
with
it.
E
A
Want
some
place
to
merge
and
iterate
and
experiment
with
a
tool,
I
think
moving
us
over
to
the
experiment's,
subdirectory,
okay
and
then
I
think
sort
of
trying
to
put
together
a
design
doc
that
scopes
out
like
what
is
it
you
are
trying
to
solve.
What
is
it
you
are
not
trying
to
solve?
Why
are
you
doing
it
this
way.
C
E
E
It's
just
mostly
saying
here's
something,
that's
real
that
I
can
think
about
when
it's
just
abstract
I,
it's
hard
to
bound
my
the
scope
for
me
so,
like
I,
said
in
the
PR
that
this
is
really
supposed
to
be
a
point
of
discussion
like
maybe
this
is
useful
and
again
I
only
situated
it
there
because
I
didn't
know
better.
Like
yeah,
there
was
no
intent
for
anyone
on
the
call.
There
was
no
intent
or
opinion.
It
was
just
it
didn't
fit
under
a
prow
command
and
didn't
know
about
experiment.
That's
all
cute!
There.
D
E
A
I
admire
and
appreciate
those
who
express
their
constructive
criticism
in
the
form
of
pull
requests
and
code.
That
is
super
helpful
and
so
I
just
want
to
make
sure
I'm
super
encouraging
that
kind
of
behavior
I
think
the
lesson
I
have
taken
away
from
today
is.
It
would
be
a
good
idea
to
document
some
of
these
expectations
about
the
way
we
operate
for
certain
things
of
certain
sizes
into
are
contributing
MD
file.
So
it's
a
little
clear
to
folks
what
a
smooth
path
to
landing
stuff
like
this
looks
like
yeah.
E
I'll
am
it
when
you
were
saying
we're
talking
about
the
the
last
item.
I
was
thinking.
Was
he
adding
all
this
knowing
what's
next
and
what
I
didn't
do?
Yeah
I
think
I
said
the
PR
that,
if
nothing
else
this
to
me
will
be
an
educational
experience
in
the
process
of
contributing
a
tooling,
if
you
know,
if
not,
hopefully,
actually
contributing
tooling,
but
for
me,
like
I,
said
this
I'm
learning
I'm,
you
know
learning
how
to
contribute
the
tooling
so
yeah
Stevenson,
since
you
were
so
it
made.