►
From YouTube: Kubernetes SIG Testing 2018-07-17
Description
A
Hi
everybody
welcome
to
the
state
testing
weekly
meeting
today
is
Tuesday
July,
17th
and
I.
Am
your
host
Aaron
crackin
burger
today
I
wanted
us
to
talk
a
little
bit
about
prowl
and
if
we
have
time
we
can
talk
about
sort
of
the
state
of
doctor
and
doctor
with
testing
and
the
concept
of
you
know
supporting
out
of
tree
testing.
A
B
B
B
A
Okay,
so
the
general
idea
here
is:
we
want
to
prowl
right
now
today,
like
I
described
prowl
as
a
thing
that,
given
a
github
event,
does
a
thing
or
it's
like.
If
this
then
that,
but
for
github
and
you're
talking
about
like
we
want
to
turn
it
into
if
this
than
that,
but
for
a
couple,
different
source
control
systems,
I
would.
D
Say
certain
parts
of
it
are
that,
like
hook,
other
parts
of
it
are
kind
of
just
kubernetes
controller
things
that
don't
really
care
about.
Any
of
that,
like
the
like,
the
the
job
agents.
Don't
know
that
much
about
that.
Don't
really
respond
revenge
for
that,
and
we
have
other
triggering
for
jobs.
That
has
was
like
no
connection,
but
then
you
also
have
like
the
the
robots
and
stuff
that
are
just.
If
this
is
that
forget
hub
yeah.
B
So
a
high-level
idea,
I
kind
of
want
to
propose
is
that
we,
for
example,
for
us
we
can
have
a
adapter
like
between
github
and
hook
so
hook
is
currently
respond
to
github
events.
We
can
make
something
like
Providence
instead,
so
the
github
adapter,
like
whatever
in
terms
interfacing
with
github,
will
still
receiving
a
web
hooks
but
translate
the
github
event
entry,
like
a
probably
meant
to
start
with.
We
can
even
make
those
two
event
the
same
and
for
Jarett.
B
We
can
write
on
your
controller
that
translate
any
shared
actions
into
a
product
and
similarly,
for
example,
for
other
plugins,
it's
using
a
github
client,
it's
caught,
yet
it's
called
github
client,
but
it's
actually
pretty
generic.
It's
doing
things
like
coming
back
to
github.
We
just
need
to
change
it.
To
your
like
comment
back
to
a
platform,
and
so
each
platform
came
then
this
generic
interface
I.
A
Mean
this
all
sounds
super
reasonable
and
makes
sense
to
me
I
guess.
My
question
is
like
relative.
It
sounds
to
me
like
the
sort
of
thing
that
you'd
want
to
see
like
a
design
proposal
for
mm-hmm
and
pushed
forward,
but
like
I,
don't
think
anybody
would
object
to
abstracting
out
sort
of
the
top
and
bottom
of
proud
to
make
the
GUI
bits
in
the
center
a
lot
more
reusable
right,
I
mean.
E
D
Think
there
will
be
some
we
discuss
something
like
this
in
the
past
and
Cole
and
Steve
had
a
little
bit
who
I
don't
think
our
here
had
some
concern
about,
like
you
know,
github
and
we
can
use
github
I
do
think.
At
least
some
level
will
be
hopeful
like
I.
One
of
those
I
think
it'll
be
helpful
with
some
point
of
view.
Just
because,
like
we
have
things
like
the
github
API
changing
I
like
the
idea
of
having
an
interface,
that's
like
I
need.
This
thing.
D
A
D
E
I
wanted
to
offer
my
suggestion
here,
cuz
that
actually
ties
into
the
stuff
later,
but
maybe
I
can
tie
it
back
up.
I
ignoring
this
stuff
for
Lily
to
report
get
to
it,
but
I
do
defy
the
provider
as
a
code
base.
It
provides
a
specific
implementation
of
a
general
interface
or
API.
Would
it
be
worth
defining
a
sort
of
I
mean
we
talked
about
providers
now,
cuz,
it's
all
the
rage.
E
E
A
I
guess
I'm
one
I'm,
not
sure
if
it
makes
sense
to
sort
of
unify
all
three
of
these
things
so
send
calls
out
the
three
things
that
kind
of
could
vary
right.
The
the
events
you
receive
and
then
I
guess,
like
the
things
you
might
want
to
report
to
or
round
to
and
I'm,
not
sure
that
those
necessarily
have
to
be
one
in
the
same
all
the
time.
D
D
Because
you
know
that
idea
sounds
kind
of
nice
and
abstract,
but
then
we're
kind
of
a
big
blocker
for
that
oh
I'd
like
to
be
less
of
a
blocker
where
at
least
we
can
point
to.
If
someone
wanted
to
do
the
work,
it's
it's
kind
of
possible.
It
was
like
right
now.
You
kind
of
would
just
rewrite
everything.
Yeah.
D
A
A
do
I
I
mean
to
me.
It
sounds
like
sin
is
sort
of
proposing
a
design,
and
maybe
you
were
trying
to
sort
of
get
some
interest
on
if
other
people
would
like
to
help
out,
because
by
no
means
is
this,
something
that
only
Googlers
can
do
if
anything
like
this
is
all
the
github
interactions
we
do
are
pretty
much
out
in
the
open
right.
A
B
A
Hate
to
take
us
off
on
a
tangent,
but
that
raises
a
thing:
that's
been
in
the
back
of
my
mind
like
what
what
moving
all
the
plugins
to
external
plugins,
basically
cuz.
That
seems
like
a
really
nice
loose
coupling
and
it
also
forces
us
to
think
about
the
events
that
communicate
in
between
hook
and
the
plugins
and
might
be
the
right
abstraction
for
the
stuff
coming
in
to
hook.
I'm
just
curious,
like
the
pros
and
cons,
I.
D
Think
the
main
reason
we
haven't
is
just
it
does
add
some
overhead
and
like
when
we
are
kind
of
kubernetes
project.
First,
besides,
like
you
know,
okay,
now
we
have
extra
binaries
and
containers
to
manage
and
deploy
on
your
now,
like
basically
proxying
the
web
hooks
again
for
the
end
of
time,
and
if
you
don't
need
to
do
that,
it's
going
to
be
faster
and
cheaper
to
just
stick
it
in
the
same
process,
so
I
mean
we've
been
building
a
similar
thing
for
artefact.
Viewing
I've
been
hoping.
D
We
have
an
intern
here
and
working
on
that
who's,
not
here
right
now,
but
we'll
probably
be
back
again
presenting
soon
and
we've
got
kind
of
a
similar
thing
going
on
where
there's
kind
of
like
a
baked
in
set
of
like
okay.
This
is
bare-bones
or
like
Coubertin
a
specific.
But
if
you
want
to
turn
on
your
own,
then,
like
you
you're,
just
gonna
use
an
external
deployment
just
so
that
you
don't
just
for
the
logistics
of
you,
don't
have
to
like
modify
it
while
the
binaries
and
things
so.
E
I've
been
working
on
something
for
way
too
long.
It's
an
essay
on
go
plug-in
architecture,
one
of
the
things
I
proposed
in
that
there
are
a
couple
of
reasons
to
use
them.
But
if
you
create
applications
that
are
gr,
PC
based,
you
could
technically
load
them
as
a
in
process.
Go
plug-in
with
a
standard.
Lib
adhere
to
function,
signature
passing
only
the
context
in
an
interface.
So
if
these
plugins
are
built
like
that,
you
could
load
them
both
in
process
and
still
stand
them
up
out
of
process.
D
E
D
A
F
Okay,
I
can
start
I
only
see
one
other
person
on
here
with
me,
but
basically
we're
trying
to
make.
We
have
an
existing.
F
I
was
trying
to
try
to
split
things
up
into
into
sort
of
lanes
that
we
can
tackle
as
far
as
I
can
see
it's
on
in
parallel,
because
there
was
a
question
yesterday.
Well,
you
know
how
can
I
help?
How
can
I
do
this?
How
can
I
do
that
and
I
didn't
really
have
any
any
good
answers
for
the
people
present
besides,
okay?
F
Get
get
get
this
done
and
be
done
with
it
and
not
not
suffer
the
you
know
the
failings
of
Travis
or
Jenkins
running
externally,
along
with
it,
pushing
pushing
chards,
pushing
and
release
artefacts,
etcetera,
etcetera,
because
the
only
thing
I've
started
is
I
made
a
proud
job.
I
barely
understood
what
I
was
doing
when
I
did
that,
but
it
does
work
and
we
could
probably
turn
on
reporting
for
it,
but
we're
gonna
need
more
than
more
than
one
job
and
we're
gonna
need
a
longer
running
job.
F
D
G
D
A
Our
our
overwhelming
desire
is
for
prowl
and
all
the
infrastructure
that
runs
the
kubernetes
project
to
be
as
self-service
as
possible.
Yes,
fortunately,
the
state
of
today
is
everybody
feels
like
they
need
to
come
and
talk
to
us,
because
we're
not
really
greatest
at
documenting
things
and
things
kind
of
move
really
quickly,
and
it
can
be
difficult
to
discern
between
the.
A
What's
the
word
here,
the
possible
options
you
can
use
versus
the
way
that
works
for
kubernetes,
kubernetes
versus
maybe
a
lighter-weight
way
for
other
projects.
Data,
yada,
yada,
right,
I,
think
like
by
getting
a
single
job
going.
That's
a
good
first
step.
I
would
personally
from
my
perspective
and
setting
instead
of
trying
to
do
everything
in
parallel
I'm
kind
of
interested.
A
If
we
could
crawl
walk,
run
so
figure
out
how
we
can
get
the
make
sure
like
we
have
all
of
the
everything
excluding
trigger
right,
so
everything
excluding
running
jobs
so
tied
and
all
of
the.
If
this,
then
that
things
I
was
referencing,
make
sure
you're
comfortable
with
all
those
you
have
all
the
ones
you
want,
and
then
the
next
step
would
basic,
probably
heavier
like
here
I
guess
it
runs
unit
tests.
We
can
start
to
use
that
as
the
first
sample
proud
job.
A
This
things
like
tied,
for
example,
that
automatically
merges
it
doesn't
necessarily
care
if
the
contexts
are
put
there
by
prow
or
by
Travis
or
by
Jenkins
or
by
circle
or
whatever
right
through
that.
I
can
do
cooler
things
if
it
knows
that
they're
proud
jobs,
because
then
it
knows
how
to
reach
Ritter
them
and
rerun
them.
I
think
that
could
be
like
another
example
of
something
that
would
be
nice
to
abstract
out
so
I
yeah,
maybe
maybe
a
breakout,
is
the
right
way
to
do
it.
F
My
I
would
go
back
to
you
know:
I'll
just
go
back
to
the
other
channel
and
say:
okay:
when
can
we
schedule
a
you
know
breakout
or
when
can
we
when
can
I,
have
somebody's
attention
to
sort
of
go
over
this
list
of
of
stuff
and
and
let
us
know
what
is
the,
what
is
the
next
best
step
here
or
what
is
the
replacement
for
Travis?
What
is
restore
secrets?
What
is
the
replacement
for?
F
You
know
different
jobs
with
what
what's
allowed?
How
do
we?
How
do
we
make
sure
we're
not
like
accidentally
spamming?
You
got
spamming
the
the
infrastructure
right,
what
what
is
like
what
is
allowed
and
then
what
is
like,
expected
and
and
all
there's
all
sorts
of
expectations.
So,
if,
if
it's,
if
it's
establish
a
I,
think
we
can
do
that,
that's
fine
any.
D
A
A
That
would
prevent
you
having
had
to
come
to
this
meeting
right
and
so
I'm
wondering
I'm,
looking
at
the
parallel
tracks
of
work
thing
in
your
linked
issue
on
the
dock
here,
yep
so
down
at
the
bottom,
like
up
at
the
top
you
sort
of
describe
the
way
you
do
things
now,
I'm
wondering
is
that
documented
in
your
repo
and
then
I'm
wondering
if
we
have
sort
of
a
document
of
that's
roughly
the
canonical
prow
way
of
doing
things
right.
So
that's
like
a
staple
workflow
using
all
the
bits
and
bobs
of
prow
I.
D
That
or
not
I
think
it's
broken
up
a
little
bit
between
discussing
some
of
the
pieces
and
then
separately
discussing
the
actual
term
days.
Workflow
I
think
contradicts
folks
have
done
a
much
better
job.
Writing
Docs
discussing
the
workflow,
but
they're,
not
next
Sara
Lee
linked
to
it.
Just
maybe
something
we
should
expand.
A
D
F
Best
practices,
you
know
the
frequently
asked
questions
of
I,
don't
know
if
anybody
else
is
doing
this,
it
seems
like
there
must
be
other
people
doing
this,
but
you
know
I
definitely
want
to
write
down.
Why
we're
making
the
decisions
we're
making
four
for
Service
Catalog,
with
guidance
from
whoever
sure.
F
I
think
that
that
was
that
was
my
next
because
basically
have
something
right
now
and
we
do
use
labels,
so
I
was
thinking.
Maybe
we
can
we
can.
We
can
jump,
start
it
and
then
sort
of
configure
it
dual
use
and
then
once
tie
it
is
actually
running
all
of
our
stuff
or
whatever
the
actual
term
is
then
then
yeah
we
don't
we
just.
We
stopped
doing
the
manual
way
right.
I.
E
Think
the
guidance
was
to
keep
keyword
for
me
like
it
and
that's
kind
of
why
I'm
here
for
other
things
too.
It's
yeah
I've
been
accused
of
doing
it
on
my
own,
but
it's
because
well,
there's
no
guidance
and
then
they
community
doesn't
wanna
get
involved.
It
was
developed
too
much
in
isolation,
so
you
try
to
get
the
community
involved
early,
but
there
p52
people
doing
ten
different
things.
Twenty
different
ways
right.
C
E
A
So
this
is
why
I
think
I
said
as
much
in
the
channel
when
Morgan
was
asking
like
people
coming
to
prowl
for
the
first
time
with
fresh
eyes
and
running
into
mistakes
and
then
documenting
how
they
got
through.
That
is
super
helpful
for
us,
because
we
don't
you,
don't
necessarily
have
fresh
eyes
right,
we're
sort
of
used
to
all
the
foibles,
and
some
of
these
conventions
are
just
like
in
our
brains
and
we
maybe
don't
know
what
we
know
in
some
sense.
So
that
is
one
of
the
most
helpful
things
that
people
can
provide.
I.
F
A
You
know
I
think
so,
I,
just
like
as
a
human
being
this
issue,
a
LinkedIn
chat,
is
long
and
large,
and
it
kind
of
scares
me
because
I'm
unclear,
like
what
sort
of
the
definition
of
done
is
I
know
you
have
lots
of
chat
boxes
and
there's
lots
of
things.
I
want
to
try
and
break
it
down
into
more
manageable
tasks,
and
so
for
me
right
now,
the
first
one
would
be
get
tied
up
and
running.
A
Okay
and
the
one
like
little
NIT
that
jumps
out
there
for
me
is
you're
talking
about
using
LG
TM
1
+
LG,
TM
2.
This
is
something
that
Doc's
folks
used
for
the
website
repo
and
we
eventually
had
to
convince
them
to
use
just
the
regular
LG
TM
and
approve
labels
that
are
used
everywhere
else
in
the
project.
Yes,.
F
F
There
I
just
thought,
writing
stuff
down
yeah
thumbs
up
for
me,
I
think
I
think
Jeremy.
How
do
you
feel
Jeremy
gave
us
also
a
thumbs
up,
so
I
think
I
think
we're
gonna
I
think
you
can
expect
some
traffic,
for
maybe
a
couple
of
us
just
as
individual
people
and
I'll
have
to
look
into
tide
and,
and
maybe
that'll
be
the
next
thing
we
immediately
turn
on.
A
G
Tracking
issue
that
we
can
follow
up
on
with
regards
to
like
what
you're
looking
at
in
particular
because
there's
there
are
many
layers
of
overlap
between
some
of
the
things
that
we
want
to
do
for
sequester
lifecycle,
as
well
as
other
consumers,
that
kind
of
all
they
all
converge
in
this
space.
So
you
mentioned
it,
you
were
working
on
something.
Is
there
a
tracking
issue
for
something
there.
D
Is
not
yet
what
I
was
working
on?
First
was
literally
just
looking
at
whether
all
of
the
ways
that
we're
leveraging
focusing
that
what
does
anyone
else
build?
What
do
we
have
in
test?
Intro,
we
had
a
prototype
for
a
while,
it
could
drop,
because
memory
was
working
on
it.
We've
gone
to
another
job
and
I
was
supposed
to
take
it
on,
but
I
had
a
couple
other
high
priority
issues.
I
have
been
evaluating
and
I
I
want
to
put
something
out
like
that.
D
G
G
D
Wanted
it
I
expect
to
have
something
by
sometime
next
week,
if
not
it's
not
okay,
I
just
wanted
a
little
bit
more,
but
I
would
also
add
that
that's
an
overlook.
A
thank
you
so
much
doctor
doctor
testing
now
for
various
levels
of
that,
but
not
really
like
clusters,
which
is
mostly
what
I'm
looking
at
right
now.
Yeah.