►
From YouTube: Kubernetes SIG Testing 2018-01-16
Description
Meeting notes: https://docs.google.com/document/d/1z8MQpr_jTwhmjLMUaqQyBk1EYG_Y_3D4y4YdMJ7V1Kk/edit
A
A
A
Meeting
last
week
about
where
the
best
place
to
roll
out
and
discuss
automation,
changes
was
feedback.
I
have
gotten
from
the
community,
as
I
have
been
trying.
The
one
by
one
enable
tie
for
certainly
poses
that
people
don't
like
being
voluntold
process
changes
things
that
have
been
done
in
the
past
that
it
caused.
Some
friction
for
people
are
things
like
having
80
plus
labels
added
to
their
repo
unannounced
or
having
a
bot
start
commenting
on
a
bunch
of
stale
issues
and
PRS
unannounced
in
a
way
that
they
had
no
warning
and
couldn't
really
like.
A
We
configure
the
block
or
disable
the
POC
against
their
repo.
This
thing
started
a
lively
discussion
about
what
it
means
to
be
a
part
of
the
kubernetes
project
and
whether
or
not
there
should
be
standard
policies
and
procedures
and
automations
that
are
expected
of
every
repo
in
the
project
and
I
think
that
discussion
is
still
ongoing.
A
It
needs
to
find
a
home.
My
suggestion
was
that
the
contributor
experienced
special
interest
group
would
be
the
best
place
to
have
that
discussion,
but
I
do
think
that
we
as
a
group
when
you
pull
out
things
that
could
essentially
affect
the
way
automation
works
on
the
project
or
some
the
high
traffic
repos
need
to
do
a
better
job.
With
giving
a
heads
up
to
you,
the
broader
community,
you
know
call
the
shot,
take
the
shot
certain
things
and
we
have
gotten
better
about
it,
but
I
still
think.
A
C
C
You
know,
as
well
as
how
the
policy
will
be
enacted,
because
we're
super
huge
project
now
and
having
sort
of
a
tiered
pilot
model
as
like
Michelle
likes
to
reference,
it
asked
helps
to
sort
of
eliminate
friction
having
like
known
opt-in,
folks,
who
would
want
to
be
the
guinea
pigs
to
volunteer
to
do
quick
feedback
on
whatever
process
changes
we
want
to
make,
and
then
you
know
then
do
a
stage
two
and
a
stage
three
but
I
think
that
I,
don't
think
that's
even
control.
Beck's
is
your
responsibility.
C
I
think
that
needs
to
be
driven
by
the
student
can
because
it
affects
the
entire
things
of
the
entire.
The
kubernetes,
which
is
a
lot
of
things,
and
they
tell
kubernetes,
is
like
an
actual
finite
define
thing
like
because,
right
now
the
org
houses
like
everything
in
his
brother
I,
think
that's
a
little.
It's
cart
before
the
horse.
I
think
people
can
absolutely
opt-in
to
whatever
they
want
to
opt
in
that's
their
choice,
but
I,
don't
think
policy
can
be
defined
by
contributes
something.
A
Clear
I
wasn't
tired,
although
I
do
think
substance
of
lis.
That's
the
place
to
talk
about
that
contributor
experience,
but
more
around
how
we
actually
decide
to
roll
out
changes
to
our
automation,
because
I
think
in
the
past,
we've
just
moved
them
out
in
the
name
of
agility
and
innovation
and
and
to
prefer
to
ask
for
forgiveness,
rather
than
you
know,
giving
even
an
advanced
heads-up
a
lot
of
loan
asking
for
permission.
A
So,
like
specifically
for
me,
trying
repos
with
tide
I
have
been
asking
in
the
form
of
issues
and
for
requests
which
Rico
owners
are
free
to
refuse
to
merge,
or
you
know,
I'm
a
dialogue
there.
That
to
me
has
seem
to
be
the
most
direct
place
to
have
a
conversation
with
people
who
have
purview
over
that
repo.
C
And
that's
a
total
opt-in
mechanism.
The
the
point
I'm
trying
to
make
is
that
the
enforcement
of
tooling
and
labels
is
a
policy
change
and
we
can't
enforce
policy
on
what
is
kubernetes
until
we've
actually
defined.
What
is
kubernetes
because,
right
now,
it's
many
things.
It's
too
many
things
yeah,
so
so,
and
then
then
we'd
also
have
to
figure
out
what
is
what
does
it
mean
to
be
enforcing
these.
A
A
Just
hardly
make
this
project
work,
but
it's
also
pretty
cool
to
get
recognition
for
things
that
you
know
pointing
out
when
things
change
for
the
better
anyway,
that's
something
I
tripped
and
fell
over
with
tide
attempting
to
have
that
to
helm.
It
was
really
fun.
I
also
got
a
bunch
of
feedback
that,
like
we're
not
documenting
what
it
is
we
do
or
how
we
do
it
I'm,
not
sure
that's
entirely
fair,
but
I
do
think.
I
can
stand
to
do
a
better
job
of
attempting
to
broadcast
market
and
communicate
what's
happening.
A
To
implement
rather
than
just
implementing
it,
but
I
will
get
there
on
that.
In
the
meantime,
what
I'm,
concretely
working
on
right
now
is
updating
the
documentation
around
owners
because
it
has
fallen
out
of
date
and
we
have
added
at
least
one
new
field
to
it,
but
taking
away
one
with
Ian
from
them.
So
it
looks
for
a
pull
request
from
me
shortly
on
them.
Next,
up,
I
wanted
to
hand
off
to
st.
Clair
talked
about
sub
projects,
so.
C
I
just
got
feedback
from
Roberts,
because
I
was
poking
him
earlier
today
about
where
he
did
the
PRS
and
he
did
the
PR
s,
4
cluster
lifecycle
as
working
group
elements
and
crated
them
as
separate
working
groups
sponsored
by
the
main
thing,
even
though
they're
more
like
sub
projects.
So
we
haven't
really
defined
this
space
very
well.
C
There's
there
there's
stratification
for
it,
there's
the
saying
which
is
the
umbrella.
Then
there's
working
groups
which
are
usually
cross-functional
that
kind
of
crossed
sig
binaries.
Then
there
are
sub
projects
which
are
pretty
much
like
owned
by
the
sig,
but
they
require
they're
focused
enough
to
require
their
own
separate
thing
right.
C
This
kind
of
falls
into
the
category
of
sub
project.
What
we
were
bantering
on
on
the
mailing
list
about
which
is
basically
just
having
folks
focus
on
just
the
test
and
the
test
frameworks.
Oh
and
but
there
is
a
really
good
community
pulls
for
that.
But
I
could
easily
community
pull
as
a
working
group
and
then
put
a
note
inside
there
that
this
there's
no
documentation
for
how
this
is
done
formally.
A
A
That
I
just
won't
have
the
time
to
implement
in
the
next
two
weeks.
The
first
stab
that
I
wanted
to
take
was
updating
the
safe
thought.
Gamal
file,
that
was
inside
of
the
kubernetes
community,
repo
from
which
all
of
the
stake
documentation
is
generated
and
I
wanted
to
add
the
concept
of
a
sub-project
feeling
to
that.
A
So
as
called
out-
and
they
spoke
to
me
at
the
idea
of
a
sub-sub
project-
is
it
something
that
has
one
or
more
voters
spot
as
a
main
that
has
one
or
more
owner
spots,
as
defined
by
a
list
of
URLs
to
those
other
spots?
The
idea
here
is
that
some
sub
projects
may
be
able
to
map
cleanly
to
a
single
repo.
A
Come,
for
example,
is
probably
its
own
sub
project,
whereas
other
sub
projects,
like
the
workload
ApS,
are
spread
across
a
number
of
different
sub
directories
inside
one
or
more
Rica's,
and
so
those
would
need
to
be
call
that
message.
The
community
ladder
also
calls
out
the
concept
of
eight
owners
sucker
for
a
maintainer.
There
is
no
technical
difference
from
either
of
those
from
an
approver
as
far
as
our
automation
is
concerned,
but
it
might
be
useful
to
have
that
as
a
separate
field
to
owners
files
for
us
to
make
some
differentiating
behavior
down
the
line.
A
D
So
I'll
just
jump
in
on
that
real
quick.
This
is
Matt
when
we
talked
about
the
difference
between
a
maintainer
and
an
owner.
The
maintainer
has
right
privileges
to
the
repo,
and
one
of
the
things
that
was
suggested
was
that
if
you
knew
who
the
maintainer
--zz
were,
you
could
have
automation,
go
ahead
and
manage
that
list
in
github
and
that
right
access
automatically.
So
that
way
you
have
it
not
only
documented
who
will
has
the
right
access,
which
we
don't
in
many
places
today.
A
B
A
Can
add
automation
to
create
github
teams
based
on
those
files,
those
things
something
else
we
could
enforce
that
somehow
those
files
are
linking
back
to
you
and
are
correctly
populated
with
destinations
things
like
that,
but
the
goal
for
right
now
I'm
trying
to
see
the
satisfy
as
I
guess.
Tim's
request
is
just
to
list
the
sub
projects
so
I'm
taking
later
on.
Implementing
that
and
I.
C
C
I
think
what
I'll
do
is
I'll
read
through
that
doc
in
more
detail
and
I'll
create
the
PR
for
it
and
I'll
outline
whose
people
are
who
are
the
listing
in
the
owners
files
and
the
location
is
the
under
files,
none
of
which
have
not
been
done
yet
because
we're
still
waiting
on
the
repo
for
test
frameworks,
but
that
that
directly
relates
to
this
work.
So,
on
the
test
frameworks,
repo
I
have
created
that
repo
and.
A
I
have
added
the
two
individuals
who
initially
proposed
for
me
and
that
repo
for
their
integration
test
frameworks
as
admins.
They
have
created
an
admins
team
for
that
or
I'm
blocked
right
now,
as
we
had
talked
about
having
testing
decades
science
frameworks
redirect
to
that
repo
which
matters
to
them
just.
A
A
A
B
A
B
A
E
So
last
meeting
we
had
talked
about
having
a
co-working
day
in
Seattle
I
volunteered
the
Amazon
office,
so
we're
working
on
getting
the
space
secured
for
that
I
didn't
know.
If
we
wanted
to
talk
about
like
specifically
start
times,
there's
also
talk
of
having
a
kind
of
an
after
social
I.
Don't
know
the
people
in
Seattle
I.
We
can
do
this
over
email
or
I.
Don't
know,
I
just
wanted
to
call
it
out
and
it's
gonna
be
the
26th
and
which
probably
go
into
it.
E
A
E
A
F
Kind
of
a
pretty
big
stuff
to
say
this
is
the
canonical
way,
but
like
most
of
the
builds
and
infrastructure
are
basil,
builds.
B
F
General
customers
having
quite
a
discussion
around
depth,
even
not
necessarily
considering
basil,
just
because,
like
you
know,
DEP,
is
shining
out,
but
does
a
very
poor
job
not
pulling
in
the
world
into
your
repo
and
that
causes
problems
for
various
things.
Basil
included
also
just
bloating
your
reboot,
so
that
kind
of
needs
sorting
and
as
far
as
kubernetes
concerned
for
being
the
canonical
build
I
think
we
need
like
full
cross-platform,
builds
know.
F
A
D
So
just
a
quick
question
here,
so
if
we
don't
actually
check
so
we
checked
the
whole
vendor
world
in
already
today
in
something
like
kubernetes
right.
So
what's
the
difference
between
def
updating
it
when
we
update
it
versus
go
DEP,
updating
it
and
checking
all
the
vendor
in
any
way,
or
are
we
looking
at
a
different
process
for
handling
dependencies.
F
D
D
D
Yeah,
in
fact,
I've
had
a
whole
bunch
of
talks
because
when
you
go
start
stripping,
files
out
of
repos
or
making
changes
to
files
does
that
impact
licenses
and
what
do
different
companies
think
about
that
and
I,
don't
know
where
everybody
stands
on
it
now.
The
last
time
I
checked
on
it.
I
got
multiple
answers
to
that
question,
and,
and
so
that's
why
I
can
glide.
There
was
an
opt-in
tool
that
lets
you
went
and
clean
things,
yeah
glide
doesn't
prune,
but
there's
actually
a
tool
that
somebody
created
to
allow
you
to
opt
in
prune.
D
It's
called
like
bender
cleaner
and
it
has
different
Flags
depending
on
what
you
can
do.
So
your
organization
can
opt
into
certain
kinds
of
pruning
and
that
because
different
people
will
have
different
interpretations
and
different
code
with
different
licenses,
and
you
don't
want
to
automatically
do
anything
you
want
to
be
smart
about
altering
code
is
stripping
files
out
altering
it
or
not.
Yeah.
Let
lawyers
deal
with
that
and
I
just
kind
of
step
away
and
deal
with
options
and
that's
part
of
what
is
going
on
here.
Yeah
we're.
F
Also,
having
some
issues
upstream
with
and
with
with
basil,
maybe
some
of
the
go
tooling
will
smarter
about.
We
eric
has
hacked
something
together
for
a
tester
and
further
lets
you
kind
of
like
auto
strip
it
from
the
basil
side
of
things.
But
it's
shell
code
believe
we
can
like
make
that
feature
official.
C
C
F
Don't
have
a
particularly
great
handle
on
it
right
now,
but
I
think
Jeff
has
because
he
just
went
through
and
marked
a
bunch
of
binaries
that
don't
have
si
go
as
purely
for
the
basil
build,
so
those
can
actually
be
if
you're,
using
a
relatively
recent
tooling
in
a
recent
checkout,
you
can
build
between
platforms
now,
okay,
so
there's
a
PR
somewhere
that
goes
through
and
identified
some
boast
of
the
binaries
that
don't
but
they're.
Definitely
things
like
I
think
the
cubelet
lots
about
my
head
that
need
to
use
some
Seco
dependencies.
F
Basil
has
pretty
good
concepts
now
around
like
platforms
and
architectures
and
things
and
tooling
for
that
and
having
cross
platforms
and
will
like
Auto
manage
your
go
tool
chains,
but
it
doesn't
have
a
good
story
for
native
tool
chains
or
a
java
tool
chains
to
get
so
I'm,
not
sure
we're
on
on
that.
F
F
Also
think,
even
if
we
do
get
these
things
sorted
out,
there
definitely
has
to
be
like
a
bigger
discussion.
Like
you
know,
if
people
want
to
start
using
this
and
the.
C
Only
thing
I
really
care
about
is
that
these
artifacts
have
two
different
ways
of
being
created
and
you
cannot
produce
a
canonical
binary,
build
from
the
main
repository
you
go
through.
Their
leads
repository,
which
then
lunges
through
and
creates
the
artifacts
that
are
then
used,
but
those
artifacts
are
not
tested
as
part
of
the
test
cycle.
Great
so
like
we
only
create
the
RPMs
and
the
Debs
at
the
end
of
the
cycle
and
they're
not
continuously
constantly
just
data.
C
A
Mean
I
get
that
there
are
a
lot
of
people
here
that
work
on
basil,
but
I
also
feel
like
that's
a
discussion
that's
best
suited
for
those
things.
That's
focused
on
releases
like
the
beta
and
listed
there
in
these
bases
totally,
and
if
it
turns
out
that
there
are
two
ladies
in
using
the
right
way
to
build
kubernetes,
we
should
we
should
discuss
that.
They're,
like
you,
know,
heck.
A
F
A
David,
like
I
I,
share
the
enthusiasm
for
that
like
I,
said
I.
Think
more
so
far.
Nobody
really
pays
attention
to
this
until
the
very
last
release
is
built
when
we
should
be
building
at
least
is
every
two
weeks
and
calling
them
alphas
and
betas,
and
those
candidates
seems
like
we're
not
exercising
all
of
the
release
process
on
time
and
the
release
thing
probably
stand
to
hear
about
that.