►
From YouTube: Kubernetes SIG Testing - 2019-07-16
Description
A
Okie-Dokie
so
today
is
July
16th,
and
this
is
the
suggested,
meaning
that
this
is
being
recorded.
So
nice
we
posted
online
shortly,
and
so
we
have
a
couple
items
on
the
agenda
to
chat
about
first
being
the
spyglass
spyglass
lens
design,
doc
from
Katherine
and
then
potentially
talking
about
the
in
repo
configuration.
If
we
have
time
so
Katherine
are
you
are
there?
Can
you
start
us
off.
B
B
Okay,
so
this
doc
is
linked
in
the
agenda.
The
basic
idea
is
phat
spyglass
iev
thing
we
used
to
view
jobs
right
now
has
a
bunch
of
lenses.
B
We've
heard
from
a
few
people
who
actually
want
to
do
this,
but
in
practice
to
do
it
you
need
to
fork
trou
and
then
make
changes
to
Dec,
forget
LinkedIn,
which
is
a
fairly
high
barrier
to
entry
and
is
not
really
how
we
want
to
do
this.
So
this
document
proposed
is
actually
finishing.
The
original
goal
of
spyglass
and
Lessing
lenses
exist
externally.
You
can
read
this
whole
thing
if
you
like,
the
short
version
is
described
in
this
diagram,
Paulie,
I
guess.
B
The
short
version
is
that
currently
lenses
are
iframes
that
are
served
by
deck
in
the
future.
With
this
is
implemented
when
they
will
still
be
I
framed
for
surf
by
deck,
but
instead
of
the
thing
that
deck
does
being
handled
internally,
it
will
make
requests
to
external
lenses
and
then
bailed
feedback
through
deck,
which
will
essentially
reverse
proxy
to
res
lenses.
B
D
B
On
them
and
there
will
be
links
and
then,
if
you
were
to
go
to
this
link,
it
jumps
back
amazing,
that's
it.
How
does
it
handle
skipped
lines?
It
will
load
four
lines
and
bend
over
okay
cool.
A
Cool,
well
then
I
don't
know
if
Cole
is
don't
call
this
here,
I,
don't
know
if
you
planning
to
join
Alvaro.
Do
you
want
to
talk
about
the
repo
gig.
C
E
E
E
C
Yeah
I
think
so
just
being
given
the
caveat
that
anyone
who
isn't
super
interested
in
this
or
hasn't
read
the
PR
like
the
conversation,
that's
gonna,
follow
is
gonna,
be
very
specific,
so
I,
don't
think,
there's
anything
else
on
the
agenda
today.
So
if
that's
you
thanks
for
coming,
if
you're
interested
stick
around,
this
is
cool
stuff.
E
Okay,
so
this
off,
so
the
three
points
I
just
mentioned,
whether
once
I
went
as
a
feedback
that
we
like
the
more
high-level
concerns.
There
were
also
some
points:
I
want
how
like
tickets,
but
since
the
purpose
of
the
pull
request
was
more
to
determine
on
a
higher
level,
if
it's,
if
the
amount
of
complexity
is
okay,
I
just
picked
this,
is
this
the
complete
list
or
did
I
miss
anything?
That's
one
of
you
as
yeah,
more
concern
spot
like,
if,
mostly
if
we
were
to
find
okay,
a
solution
for
those.
E
E
C
C
Guess,
in
my
mind,
it
just
seems
like
we're:
that's
a
lot
of
stuff
that
already
gets
handled
for
jobs
and
reporting
and
I.
Don't
necessarily
want
us
to
have
a
bunch
of
different
ways
to
accomplish
this
activity,
but
I
don't
know
if
it's
any
less
any
more
clean
to
like
force,
repos
to
run
a
check
increase.
E
The
hook
and
the
presubmit
could
still
be
passing
like
no
connection
that
doesn't
work
out
or
whatever.
In
that
case,
the
user
would
just
see
where
senders
context
and
no
feedback
on
what
exactly
happens,
question
is
that
would
be
okay
or
how
we
should
handle
us.
I
mean
for
the
beginning,
I'm
also
fine.
To
just
say
we
we
put
the
whole
checking
into
into
a
piece
of
myths,
and
we
assume
that
if
the
piece
of
meat
passes,
the
passing
also
does
possible.
C
Essentially,
I
think
we
in
the
past
have
also
talked
about
like
shipping.
Other
jobs
is
sort
of
like
built-in.
You
know
anointed
jobs
like
the
record
sector
or
the
label
sink
and
stuff.
So
maybe
this
could
be
precedent
for
crowd
shipping
jobs
that
are
enabled,
but
at
the
same
time
then
I
think
we
need
to
be
careful.
What
how
he
identified
the
job
did
they
look
look
yeah.
C
E
E
C
C
F
C
A
F
F
C
This
case,
what
that's
why
I
was
asking
like
a?
Is
there
a
weird
race
condition
that
we
could
like
introduced
by
uploading,
the
version
of
the
config,
that's
in
Tokiko,
but
not
what's
been
uploaded
and
then
B?
If
we
could
just
look
at
what
was
uploaded,
then
we
can
just
mount
that
into
the
job
instead
of
having
to
you
know,
do
you
across
yo?
F
E
Yeah
and
what
folks
I'm
a
caught
happiness?
Oh,
that's
you
if
you
have
consequently
in
your
contact
currently
in
your
cluster,
which
doesn't
get
lowered
because
it's
invalid
on
the
checking,
it
will
just
say
there
are
things
in
minutes,
even
though
you
are
you
do
changes
to
apology,
I,
don't
I'm,
not
the
cause,
so.
F
The
chance
of
a
race
like
that
happening
is
pretty
low
and
it
would
the
validation
then,
would
just
start
do
we
just
have
like
broken
validation
on
the
master
branch
for
that
repo,
and
it
would
be
obvious
with
the
problem
is
so
I?
Don't
know
that
that's
a
check
like
it's
definitely
not
ideal,
but
I.
Don't
know
that
that's
a
huge
blocker
I
mean.
E
F
F
D
F
C
E
And
that
would
but
then
mean
that
people
basically
have
to
manually
create
this
piece
admit,
because,
depending
on
how
you
access
your
test
and
for
people
and
where
science
is,
the
arguments
for
the
config
checker
will
be
different.
So
it
would
be
a
bit
hard
to
generate
the
whole
thing.
Don't
worry
yes,
for.
F
F
E
C
Think
we
should
be
able
to
like
I,
think
the
external
facing
config
that
we
might
we're
definitely
going
to
need
is
the
version
of
the
check,
config
image
to
use,
but
I
think
we
should
be
able
to
generate
the
rest
of
it,
because
we
can
use
the
well
I.
Guess
we
don't
know
yeah
we
can.
We
can
have
a
convention
where
we
can
generate
it,
because
we,
the
updated
pic
plugin,
is
very
pushing
prada
animal
and
the
plugins
we're
not.
The
only
issue
would
be
if
you
named.
C
F
F
F
Like
actually
existing,
if
you
pick
somewhere,
but
it
would
exist
in
the
config,
you
are
on
yes
like
we
needed
to
actually
exist
in
the
list
of
jobs
Liancourt
smoothly,
but
I'm
wondering
if
we
want
to
like
have
this
happen
at
config
load
time
or
something
where,
like
these
jobs,
are
automatically
applied
to
rebus
that
ethnicity,
polish,
I
would
I
would
like
like
cuz.
This
is
something
that
we
want
it
to
be
automatic,
but
we
also
would
like
yeah,
I
guess
if
it
shows
up
on
deck
that
could
be
okay.
We
probably
would.
E
B
E
E
Obviously
that
we
have
to
pay
for
this,
both
didn't
like
so
the
identity
that
would
be
to
have
some
component.
That
gets
all
events
for
anything
that
could
change
the
config.
So
that
is
the
conflict
for
the
and
the
because
I
self-service
manages
the
testing
for
repo
and
that's
events
for
changes
to
the
base
of
a
pull
request
and
events
to
for
changes
to
the
head.
E
So
we
would
sort
of
have
to
build
something
that
gets
notified
when
one
of
those
three
changes
and
then
checks
if
it
has
to
retire
something.
What's
the
sorry
I
forget,
what's
the
effect
of
the
reconciler,
what
are
you
doing
that
add
to
to
remove
github
set
of
contacts
that
don't
have
a
job
anymore
and
that's
yeah?
Basically,.
C
C
E
F
E
Because
the
thickness
you
need
to
get
notified
for
changes
to
the
base
and
to
the
head
of
the
pool
request,
because
those
could
change
the
config,
probably
the
cells,
because
they
can
be
extended
to
do
that.
I,
don't
know,
I
just
had
a
quick
look
at
its
code,
but
the
main
thing
that
would
then
need
to
be
changed
is
that
the
cells
become
Tyler,
gets
all
push
events
in
order
to
be
notified
when
either
of
those
two
changes,
that's
clearly
not
happening.
E
Why
would
I
need
to
get
push
events
because
imagine,
for
example,
I,
do
a
pull
request
onto
a
master
branch
and
someone
else
so
and
then
a
drop
gets
created,
it
fades
and
the
reach
and
someone
else.
We
move
that
job
from
the
master
branch.
Now
this
context
need
to
get
retired
and
some
of
the
status
we
consular.
Whatever
does
this
needs
to
know
that
something
changed
it
has
to
check
if
it
needs
to
do
something,
and
that
would
be
this
push
event
also
another
way.
F
C
E
F
So,
but
that's
only
the
it's
only
a
case
if
the
prowl
jobs
are
actually
being
her
if
the
PR
is
actually
changing.
The
crowd.
Jobs
right,
which
is
a
very
small
fraction
of
the
ours,
will
and
not
very
not
blocking
the
statuses
on
those
would
be
kind
of
like
a
big
deal
like
we,
for
example,
that'd
be
a
hard
blocker
for
kubernetes.
E
F
D
F
C
F
F
F
Is
not
on
a
per
branch
level
right,
our
config
defines
all
the
jobs
for
all
the
branches
potentially
like
you.
Could
you
could
split
that
you
know
if
you
wanted
to,
but
it
wouldn't
work
right
like
we
have
one
config,
that
yellow
file
and
the
jobs
for
all
of
our
branches
are
fine.
There
does
it
make
sense
first,
like
maybe
we
should
just
allow
that
having
branched
and
big
is
like
how
what
does
that
even
beat
that
like,
and
we
have
multiple
sources
of
truth
in
it?
We're
gonna.
C
F
E
C
C
F
It
just
makes
config
validation,
make
no
sense,
because
if
we're
now,
we
no
longer
have
a
single
source
of
truth.
You
know
our
config
defines
jobs
across
all
branches.
If
we
now
have
branches
that
each
can
define
define
config
across
all
branches
like
that,
just
doesn't
really
make
sense.
There's
gonna
be
overlap,
so.
E
F
So
imagine
that
on
the
master
branch
we
have
some
jobs
configured
for
release
115
and
then
on
the
release.
115
branch,
you
change
those
jobs
like
which
jobs
actually
run
on
release
115
now
is
it
the
branch
jobs
that
are
defined
on
the
release,
115
branch
or
like
the
ones
that
are
defined
for
the
release,
150
branch
and
master,
but.
F
F
E
F
E
F
E
E
C
F
This
really
a
better
UX
like
this
seems
like
we're
trying
to
make
UX
worse.
In
my
opinion,
you
know
having
to
define
like
having
to
switch
branches
and
define
your
jobs
and
different
branches
like
in
the
most
cases.
All
of
your
jobs
are
four
different
branches
are
pretty
similar
with
slight
tweaks
and
I.
Don't
know
why
you'd
want
to
have
to
define
those
like
you
know,
actually
have
those
in
separate
files
on
separate
branches
and
everything,
because.
E
F
I
think
we,
but
then
it
just
makes
it
really
hard
to
do
any
sweeping
changes
trip
across
all
of
your
jobs,
because
you'd
have
to
actually
make
changes
to
all
those
branches.
Yes,
it's
really
annoying.
That's
why
we
are
talking
about
the
whole
thing
well,
this
would
make
it
at
that
annoying
right.
That's
what
I'm
saying
if
you
then
have
to
like
make
changes
to
all
your
jobs.
You'd
have
to
make
a
PR
at
every
single
branch
is
what
she
doesn't
do
if.
E
A
C
F
C
C
F
Security
patches
and
stuff
for
that
you
know
like,
even
if
we're
not
changing
what
versions
images
are
running
on
like
or
you
know
like,
the
major
version
is
frequently
you
like
I,
don't
know,
I,
guess
I'm,
just
imagining
that
this
is
not
gonna,
be
what
would
be
usable
for.
You
know
like
our
use
case,
because
I
think,
like
we
very
heavily
make
a
little
more
sweeping
changes
across
a
lot
of
branches
like
for
you.
F
Every
time
we
do
a
config
felt,
for
example,
like
every
I
mean
that,
because
you
guys
are
going
to
cute
kids,
for
instance,
it's
not
even
cute
kids,
but
just
like
all
of
our
test
images
are
published
daily.
Now,
if
there's
any
changes
to
them,
do
you
actually
validate
that
those
old
released
jobs
work?
Can
you
do
that
they're
on
test
rent,
so
we
get
emails
as
they
start
failing
yeah.
C
I
mean
I,
think
I
think
it's
just
the
difference
in
like
approach,
because
I
think
having
a
job
that
works,
freezing
it
in
that
state
and
then
not
touching.
It
is
way
more
effective
as
a
strategy
when
you
have
released
branches
that
are
three
years
old,
like
they're
like
the
amount
of
overhead
and
like
just
combinatorial
and
nonsense.
That
happens
too
much
yeah.
C
That's
not
I
mean
that's
Cupid
any
straight
like
we
support
all
of
those
releases
and
they
need
to
run
exactly
how
they
run.
Rightly,
it's
a
huge
amount
of
effort
and
that
Cody's
in
so
I
think
in
like
even
just
putting
it
out
into
separate
rainbows
like
that,
basically
kills
the
ability
to
do
searches,
regardless
of
whether
it's
in
register.
C
F
F
B
C
F
F
Okay,
yeah
all
I'll
try
to
talk
to
some
gates,
people
and
see
if
that
the
hole
is
splitting
it
across
branches
would
be
a
hard
longer
cover
Nettie's
or
not,
because
none
of
our
jobs
are
like
that
now
and
I'm,
not
sure
if
that's
out
of
necessity,
I
kind
of
my
understanding
was
that
was
kind
of
out
of
necessity,
but
we'll
also
try
and
see.
If
that's
the
case,.
C
F
Question
about
the
generating
that
check
config!
Do
we
want
to
have
we
want
to
like
have
some
configuration
options
where
you
specify
the
job
that
is
then
run
everywhere,
or
do
we
want
to
like
totally
generate
it?
Just
have
that
be
all
automatic
like
do
we
want
to
make
it
so
that
somebody
would
say
like
here's,
the
here's,
the
check,
config
job
I
want
to
run,
so
they
would
like
to
tweak
it
because.
F
C
F
Since
this
is
a,
this
would
be
like
a
convenient
way
for
you
to
like
enforce
your
requirements
everywhere
across
all
the
prada
animals
that
you're
proud
uses.
You're
right,
you
can
put
a
pre
sim
in
each
individual
repo,
but
since
this
is
party
like
spread
everywhere,
it
could
be
nice
to
make
this
like
a
way
to
do
that
as
well
or
to
extend
this
functionalities
that
you
could
apply.
You
know,
maybe
multiple
jobs
could
be
applied
universally.
This
mechanism,
maybe.
F
Yeah
I'm
not
sure
if
we
need
to
do
any
of
that
now,
but
it's
probably
worth
just
like
thinking
about
it.
That's
you
know
generalizable,
because
it
seems
like
a
pretty
useful
thing
to
be
able
to
say,
like
all
of
these
jobs
that
are
gonna
use,
proud.
You
know
like
we
want
to
make
sure
that
their
formatting
their
files
correctly
or
whatever
yeah,
okay.
B
E
F
C
A
D
C
C
C
E
C
E
It's
a
very
good
point
just
and
I
agree
that
we
should
probably
do
it.
Do
not
do
it
like
that,
but
it
has
something
that
that
gets
all
this
events
and
I
just
wanted
to
say.
The
reason
why
I
initially
came
up
with
that
it's
to
build
I
can
probably
put
it
an
hour.
The
thing
that
gets
events
from
all
these
free
things
take
a
lot
more
time,
but.
E
Yeah,
it's
just
a
bit
more
complicated
because
it
can't
be
different
for
every
request:
yeah
yeah.
Okay,
so
do
it
the
whole
thing
up.
I
guess!
The
next
thing
I
would
like
to
do
is
to
create
a
pool,
because
that
does
the
changes
in
the
config
package
and
like
actually
it's
this
or
resubmit,
getting
function,
and
so
on,
doesn't
yet
implement
in
week
of
conflict,
but
has
is
just
a
stock
for
right
now
and
to
then
plug
into
other
plugins
and
whatever
else
we
have
that
accesses
the
priests
admin
console.
C
C
E
Actually,
that's
what
I
told
us
took
us
two
months
to
using
poet
and
we
realized
Benitez,
so
I
have
to
yeah
okay,
so
from
my
side
for
now
that
would
be
or
understand
more
question.
I
think
you
know
cool,
then,
thanks
to
both
really
for
for
your
time
on
this
I
really
appreciated.
Yeah
have
a
nice
thing.