►
From YouTube: Sig-Testing Biweekly Meeting for 20230530
Description
Sig-Testing Biweekly Meeting for 20230530
A
All
right,
hi
folks,
welcome
to
successing,
yeah
I,
think
we're
filling
in
names
of
attendees
and
whatnot,
but
just
to
start
off.
Welcome
and
if
there's
anybody
here
who
is
new
and
wants
to
unmute
and
give
a
brief
introduction
of
like
what
they're
here
for
and
anything
you
want
to
accomplish
or
like
just
want
to
give
details
about
yourself.
Please
feel
free
to.
B
C
E
F
You
hello,
everyone,
I'm,
Mattel,
I
work
at
the
data
dog
I'm
joining
for
the
first
time
of
this
sick
meetings,
because
we've
been
using
the
end-to-end
framework
for
for
a
while.
We
like
it
a
lot,
but
we
I
raised
the
niche
to
the
agenda
and
I'd
like
to
I.
Don't
know
how
know
how
to
address
or
like
how
to
start
discussing
a
possible
fix
with
the
community
nice
to
meet
you
all.
E
B
A
All
right,
it
looks
like
it's
good,
so
yeah
thanks
everybody
and
again,
if
there's
anything,
to
follow
up
with
just
in
case
we
either
like
don't
get
to
this
meeting,
or
so
we
have
a
more
permanent
record.
Please
feel
free
to
hang
on
the
slack
Channel
as
well,
especially
for
like
following
up
for
any
tips
on
good
first
issues
or
help
on
it
or
stuff
like
that.
A
Yeah
all
right
interrupted
the
agenda
with
a
quick
welcome
discussion
and
then
we'll
get
to
the
stuff
that
everybody
put
on
the
agenda,
but
just
first
off
I
wanted
to
say
everybody.
Please
welcome
the
newest,
State
Testing,
chair,
Brady
prep
he's
doing
an
excellent
job
so
far,
acting
as
basically
a
chair
already
so
now
we're
making
it
official
and
there
should
be
some
pierras
coming
down
the
pipeline
to
get
all
of
that
settled
this
week.
A
D
D
Essentially,
there's
this
issue
wherein
we
want
to
there
was
this
request
to
have
warnings,
errors,
annotations
coming
from
proud
jobs
to
be
added
to
com,
pull
requests
as
comments
right
and
see.
The
talk
that
I've
written
is
more
in
terms
of
the
grow
length,
strict
CI
lender
that
Patrick
has
been
working
on,
which
and
there's
a
brow
job
for
it.
It's
all
linked
in
that
talk,
but
the
job
is
not
functional.
D
As
of
now
and
to
make
that
I
mean
when
I
say
functional,
it's
not
being
used,
then
to
make
that
to
make
it
working,
and
you
know
to
make
it
to
readily
use
that
this
is
another
feature
sort
of
being
proposed,
wherein
it'd
be
kind
of
good
to
have
proud
jobs.
Basically,
posting
errors,
warnings
annotations
back
to
critical
requests
as
comments.
D
But
then
there
are
some
design
things
to
be
considered
over
here.
How
we
want
to
store
that
artifact?
How
do
you
want
to
store
those
comments?
One
way
is
to
store
this
comments
or
store,
whatever
needs
to
go
as
a
comment
as
a
file,
preferably
maybe
markdown
or
something
of
that
sort,
and
in
this
particular
case,
with
contents
of
the
Galaxy
Island.
D
That's
one
of
the
Solutions,
so
I'm
kind
of
hoping
that
you
know
I
can
get
some
review
on
this
and
also
looking
for
some
sort
of
an
approve,
so
Ben
posted
on
that
issue.
That,
instead
of
going
through
like
a
full-fledged
cap,
we
can
go
through
like
a
lightweight
design
review
like
a
design
dock
which
can
then
get
approved
by
proud
tools.
D
And
then
maybe
we
can
start
with
the
implementation.
So
this
is
my
first
attempt
at
like
breaking
it
up.
This
testing
and
one
I'm
hoping
to
get
some
eyes
on
this
and
then
kind
of
planning
on
sending
it
out
on
this
testing
mailing
list
and,
following
up
over
there,
would
be
cool
if
folks
get
review.
This
are
there
any
questions,
concerns
comments.
A
B
A
E
Yep
Mike
or
El
Miko's,
fine,
so
I
put
together
a
little
document
here,
just
to
describe
some
of
the
background.
I
see
Patrick
and
someone
else
has
already
made
some
comments.
But
basically
there
are
a
bunch
of
tests
that
we
were
running
with
the
cloud
controller
managers
and
now
that
we're
getting
closer
and
closer
to
having
all
the
the
cloud
controllers
out
of
the
kubernetes
tree
and
run
as
external
controllers.
E
One
of
the
things
that
I've
been
looking
at
and
I
think
Sig
cloud
provider
is
interested
in
is
creating
like
a
generic
set
of
tasks
that
we
could
have
live
in
KK
and
then
we
could
have
specific
Cloud
providers
kind
of
participate
in
those
tests
or
not,
and
this
is
like
an
issue-
that's
been
around
for
several
years
now.
I
have
a
link
in
the
document
to
an
issue
that
Timothy
Sinclair
opened
up
back
in
2018
about
making
the
test
more
generic
and
getting
rid
of
the
cloud
provider
specific
code.
E
So
the
first,
the
first
question
I
had
was:
does
the
current
CI
testing
workflow
allow
for
like
this
conditional
builds,
which
would
change
depending
on
which
infrastructure
they're
being
run
on
and
I,
do
see
that
that
Patrick
has
responded,
that
he
doesn't
think
it's
currently
being
done,
but
doesn't
think
it
would
be
impossible.
I
haven't
read
through
all
these
comments,
because
I
just
saw
them
now,
but
I'm
kind
of
curious.
If
anybody
has
thoughts
about
that
or
if
maybe
I
should
not
go
down
that
path
or
something.
G
G
E
Yeah,
okay,
cool
well
I
got
some
more
questions,
then,
to
follow
up
with
here,
but
and
like
I
guess
like
just
to
kind
of
dive
into
that
one,
a
little
bit
more
though
like
currently
in
our
tests,
we
have
skips
depending
on
which
provider
it's
being
run
on.
It
is
the
CI
system,
that's
run
by
the
kubernetes
community.
Does
that?
How
do
we
choose
which
platform
it
runs
on
between
like
AWS
azure
gke?
G
I
I
figure
well,
I,
don't
know
specifically
and
I
don't
want
to
give
you
like
a
solid
answer.
It's
just
going
to
be
defined
in
whatever
we
have
the
prowl
job
set
up
to
be
so,
if
someone's
running
against
gke
clusters,
which
I
guess
we're
doing
now,
are
running
those
on
and
provisioning
those,
then
it
may
be
locked
to
that.
I
actually
don't
know
much
about
how
the
provisioners
work
as
far
as
like
when
we
spit
out
a
cluster,
then
to
then
be
tested
against.
That
would
be
a
good
question.
A
A
Somebody
is
doing
like
a
wild
setup
that,
like
dynamically,
chooses
things
and
then
we'll
route
to
a
different
one,
but
as
far
as
like
the
project,
config
and
whatnot
I'm,
guessing
it's
defined
on
the
cluster
that
the
config
runs
on
and
then,
if
there's
anything
else
that
it
deals
with,
then
that's
dealt
within.
The
parameters
of
the
like
job
might
not
be
in
the
project
with
big,
depending
on
what
it's
doing
like
Bosco's
resources,
for
instance,
I
think
are
going
to
be
handled
by
like
these.
Whatever
the
thing
is
running
roscos,
but.
G
And
for
what
it's
worth,
let
me
go.
That's
how
I
do
the
testing
that
we're
doing
for
OSD
and
red
hat
is
that
when
a
morning
tests
against
Ro,
then
I'm
provisioning,
an
RO
cluster
and
I
am
specifically
passing
labels
to
select
those
tests,
so
I
assume
that
that
degree
of
of
control
is
there
right
now
it
may
just
not
be
as
standardized
like.
E
G
I'd
love
to
hear
some
some
of
the
others,
thoughts
who
who
may
not
have
that
much
experience
with
it.
But
I
really
do
like
the
idea
of
having
a
contract,
I
guess
per
se,
with
the
interface
that
these
providers
need
to
implement
across,
because
that's
pretty
much
what
it
is
right.
Enforcing
it
at
a
type
level
would
be
very
beneficial
but,
like
you're
saying
we
do
get
in
like
a
a
tick
in
the
egg
situation
where
you
either
have
to
import
it
this
way
or
imported.
G
This
way,
then
we're
back
to
importing
Cloud
providers,
or
we
have
to
do
some,
some
strange
building,
which
the
second
option
didn't
sound
as
bad
is,
as
importing
I
I
think
pushing
those
dependencies
as
far
as
importing
Cloud
providers
kind
of
puts
us
back
into
the
same
same
boat.
In
my
opinion,
yeah,
like
that's
all
I
got
and
sorry
I
I
feel
like
you.
You
may
get
a
lot
more
responses,
be
going
through
slack
from
talking
to
Ben
or
talking
to
Paul
or
Patrick
Moore.
E
No
I
mean
I
I,
appreciate
the
discussion
at
this
point,
I'm
trying
to
gather
kind
of
a
better
view
of
things,
because
you
know
I'm.
What
I
would
like
to
do
is
draft
off
of
previous
decisions
that
we've
made
that
we
like
in
the
community,
so
really
I'm,
just
trying
to
get
everyone's
feeling.
For
this
I
mean
it
sounds
like
making.
A
A
E
Yeah
I
mean
I
I.
Think
it's
going
to
be
a
longer
term.
Discussion
too,
like
and
I
and
I.
You
know
I
think
I'm
working
towards
building
a
cap
around
this,
so
that
we
can
have
a
better
understanding
of
how
exactly
it's
built,
but
to
push
that
along
what
I'm
going
to
try
to
do
is
build
a
proof
of
concept
that
we
could
start
like.
You
know
batting
around
and
kind
of
arguing
about
or
whatever,
so
that
so
like
right,
like
really
all
these
thoughts
are
great.
E
D
It's
going
back
to
your
question
on
conditional,
build
on
like
specific
infrastructure
platform.
I,
wonder
if
looking
at
how
cluster
API
does
it
for,
like
different
providers,
may
also
help
over
here,
I'm,
not
sure,
but
just
a
thought
because
take
extensively
use
boss,
cars
for
like
AWS,
Azure,
gcp
things
like
that.
So
that
would
be,
and
as
far
as
I
know,
there
is
a
different
e2e
test
framework
there,
which
is
built
on
top
of
Ginkgo
in
for
cluster
API,
but
from
an
infrastructure
cloud
provider
perspective.
E
E
I
think
the
different
one
of
the
big
differences
there
is
that
there
most
of
their
testing
is
done
through
the
interfit,
the
API
that
they've
created,
whereas
the
tests
that
we're
going
to
be
doing
here,
you
know
like,
for
example,
the
node
lifecycle
test.
We
need
to
have
an
interface
that
each
provider
can
Implement
to,
like
kind
of
you,
know,
destroy
nodes
or
create
nodes,
and
similarly,
with
like
the
service
load,
balancer
test
right,
like
there
I've
identified
a
couple
service
tests
that,
like
probably
could
be
used
generically.
E
But
then
we
have
other
tests
where,
like,
for
example,
on
gke,
you
need
to
do
specific,
annotation
and
clean
up
after
creating
a
load
like
a
static
load
balancer,
and
there
might
be
situations
where
we
want
to
check
to
say.
Okay,
if
you've
asked
for
a
service
load,
balancer
type
and
the
cloud
provider
creates,
you
know,
elastic
load,
balancers,
get
an
elastic
load,
balancer
actually
get
created
in
the
in
the
the
infrastructure
or
whatever.
So
like
that,
you
know.
That's
it's
something!
E
That's
like
kind
of
a
difficult
problem,
but
I
think
lies
just
between
what
cluster
API
does
and
like
what
we're
trying
to
do
in
the
KK
test,
because
you
know
in
cluster
API
you
can
say
like
well,
you
know
make
me
a
bunch
of
machines
through
the
machine
deployment
and
then
you
can
just
look
to
see.
Did
those
machines
get
created
and-
and
you
know
the
cloud
provider
handles
everything
about
doing
that,
but
that
that
is
another
good
suggestion.
E
E
Would
it
make
sense
to
move
some
of
these
like
more
specific
tests
into
that
location,
so
that
we
could
have?
We
could
have
the
individual
Cloud
providers
kind
of
participating
in
that
repository
for
a
common
set
of
tests?
That
would
be
a
little
bit
more
specific
than
the
tests
that
run
out
of
KK
or
is
there
or
is
there
maybe
a
better
way
to
do
this?
Should
we
just
say
that
each
cloud
provider
should
have
a
suite
of
their
own
tests
or
whatever?
E
E
It
will
happen
because,
like
for
example,
there's
you
know
we're
seeing
a
problem
with
a
test
that
we
had
to
remove
recently
because
AWS,
you
know,
wanted
a
very
specific
set
of
annotations
to
make
some
sort
of
behavior
happen
and
that
and
that
behavior
was
different
than
the
other
clouds.
We're
testing
so
I
have
a
feeling.
This
is
one
of
those
things.
That's
just
a
I
think
it's
going
to
be
a
difficult
question
to
answer.
A
A
Responsible
for
making
sure
that
the
Ada
tests
for
like
specific
Cloud
providers,
someone
bought,
get
run
so
I
think
that
will
also
influence
whether
like
specific
features,
especially
like,
depending
on
how
specific
they
are
not
also
need
to
get,
should
be
tested
in
like
a
main
thing
or
a
side
repo
or
like
not
at
all.
And
then
just
in,
like
whatever
the
cloud
providers
have
separately.
G
B
G
I
I
think
you've
probably
done
way
more
research
on
this
than
I
have,
but
everything
you're
saying
it
really
sounds
like
there
has
to
be
parallels
to
something
like
what
Sig
storage
did
with
CSI.
Like
you
pointed
out
like
there
has
to
be
parallels
there
as
far
as
kind
of
how
they
handled
This
and
like
we
talked
about
earlier,
it
may
not
have
been
the
standard
like
best
practice,
but
at
least
getting
some
of
those
ideas
of
how
they
did.
It
would
be
really
good
and
I.
E
Okay,
cool
yeah,
I
think
like
like
Raja
said
before,
like
you
know,
looking
at
the
cluster
API
stuff
I
think
the
CSI
tests
are
very
similar
to
that,
like
the
real
big
difference
between
the
CSI
tests
and
what
we.
What
we
want
to
do
with
the
CCM
test
is
that
the
CSI
test
can
all
be
manipulated
through
kubernetes
objects
right
so
like
whenever
they
want
to
test.
E
You
know,
they're,
just
creating
persistent
volumes,
making
sure
they
can
write
to
them,
etc,
etc,
whereas
with
the
CCM
tests
like
from
what
I've
seen
so
far
like,
although
there
is
a
subset
of
functionality
that
can
be
exercised
just
by
creating
Cube
objects,
like
you
say,
did
it
make
a
service
load?
Balancer?
Okay,
great
you
know,
did
the
node
get
marked
with
some
sort
of
zonal
information
or
Regional
information?
E
What's
the
best
way
to
go
forward
with
this
I
really
appreciate
the
discussion
here,
though,
I've
got
some
notes
on
my
document
and
I
think
this
gives
me
at
least
some
ideas
of
where
to
go
next
and
kind
of
what
to
try
out
so
I'm.
You
know
happy
for
any
more
comments,
but
I'm
also
I'm
cool.
If
we
want
to
move
on
to
the
next
topic.
A
F
You
so
we
are
using
the
end
to
end
the
framework
to
run
like
some
sort
of
well
end-to-end
test
before
giving
the
Clusters
to
our
users
and
considering
that
we
have
a
different
amount
of
controllers
and
things
we
wanted
to
execute
all
those
tests
in
parallel
as
fast
as
possible.
So
the
idea
is
that
we
create
a
namespace.
We
do
our
own
test
inside
the
namespace
and
we
delete
it
and
we
started
using
a
couple
of
LPL
functions
from
the
framework
before
each
test
after
each
test.
F
But
we
found
like
some
random
issue
and
error
and
so
running
them
with
the
race
flag
in
go
test
was
bringing
up
that
there
is
a
race
condition
and
I
think,
because
in
the
framework
there
is
a
test
and
object.
That
is
the
test
environment,
object,
and
that
has
a
context.
And
so,
when
we
run
the
multiple
test
in
parallel,
I
believe
that
we
are
trying
to
write
back
from
parallel
things
to
the
same
context.
F
And
then
we
overwrite
the
of
the
value
I
raised
the
an
issue
where,
where
I
was
able
at
the
beginning,
I
thought
it
was
like
our
testing
and
then
was
able
to
get
to
the
bare
minimum
Repro
with
the
framework
itself,
I
also
reached
the
PR,
but
either
a
fixed,
the
rest
condition,
but
then
we
are
losing
the
ability
to
use
the
context
to
pass
a
value
between
the
steps
of
the
framework
or
I
keep
the
values.
The
nicest
circle
of
the
value,
but
I
still
have
the
rest
condition.
F
So
at
this
point
I,
like
don't
know
what
is
the
best
approach?
If
other
people
have
ideas,
everyone
here
came
with
the
document.
I
don't
have
any
documents
that
prepared
with
the
suggestions,
so
I
I,
basically
I'm,
like
I'm,
happy
to
to
work
and
like
to
push
this
forward.
I
would
like
some
guidance
about
what
is
the
best
way
to
do
with
the
you
know,
building
concepts
with
the
community
and
to
propose
a
thing
that
makes
people
happy
to
use
it
rather
than
all
these
I
think
we
should
do
this.
G
Yeah
I
can
take
this
one,
so
I
will
say
we
are
quite
small
as
far
as
contributors
in
intent
framework
there's
a
handful
of
us.
We
appreciate
anybody
who
can
jump
in
and
help,
and
obviously
this
fix
is
great.
So
thank
you
for
that.
It
does
look
like
you're
on
the
right
track.
As
far
as
passing
contacts
through
there
may
still
be
a
copy
in
here
that
you're
doing
that
may
be
giving
us
an
issue
that
copy
may
be
causing
the
problem
as
far
as
the
race
condition.
G
I
want
to
investigate
a
bit
more
to
try
to
understand
the
issue,
but
this
does
look
like
the
right
direction
and,
if
you're
ever
not
getting
like
traction
on
one
of
these
PRS,
just
please
feel
free
to
Ping
and
slack.
You
can
tag,
you
can
tag
Vlad.
You
can
tag
some
of
the
others
that
you
see
as
far
as
contributors
and
we'll
try
to
take
a
look
at
it.
Yeah
we're
still
trying
to
to
build
a.
F
It's
not
the
problem.
I
have
some
more
experience
to
go.
Colleagues
that
they
are
a
little
I
would
say
skeptical
around
using
the
context
to
pass
the
things
around,
so
they
told
me
well,
we
should
probably
don't
do
that
and
they
said:
hey
I
can
go
to
the
community
and
say
people
suggest
me
to
notebook,
but
it's
a
pretty
big
change
compared
to
what
the
framework
is
so
again
before
pushing
a
bunch
of
code
removing
the
contest
someone
says
well,
why
you're
doing
that
I'm
a
talking.
G
F
H
Hey
folks,
Rahul
here
I've
been
working
with
kubernetes
for
over
a
year
and
a
half
now
and
like
over
the
past
few
weeks,
I've
been
trying
to
use
Cube
test
to
to
run
conformance
testing,
end-to-end
testing
and
all
I've
been
facing
a
few
issues.
So
I
thought,
like
maybe
I'll,
get
those
things
like
sorted
and
I
also
came
to
understand
that
kubernetes
always
open
for
contributors.
A
Yeah
glad
to
have
you
I
think
in
terms
of
probably
starting
to
sound
like
a
broken
record
a
bit,
but
like
definitely
a
follow-up
on
the
slack
for
any
like
specific
questions
or
things
like
that.
I
don't
know
currently
who's
like
an
expert
for
keep
test
yourself.
For
instance,
I
suspect
it's
not
that
many
people,
but
at
least
this
will
actually
have
a
good
chance
of
being
able
to
point
people
towards
it.
If
you
do
have
anything
specific.
A
E
H
H
E
H
Right
yeah,
yeah,
I
I
did
take.
That
course,
like
so
I
got,
I
tried
running
things
through
tube
test
two
and
then
I
realized,
hey,
there's
an
abstraction
to
this
then
I
started
working
on
Queue
it's
on
a
boy,
but
I
didn't
get
much
response
from
the
said
channel.
So
I
thought
I'll
I'll
flag
it
here,
so
that
if
folks
here
have
some
context,
they'll
be
able
to
help
me
out,
but
yeah
yeah
thanks
thanks
for
the
well.
E
I
just
Rahul
I
just
know,
for
example,
like
we
have
a
tool
at
red
hat
that
we
use
sonobui
for
to
run
the
cube
and
then
tests
so
I
know
like
sonobari,
is
a
great
rapper
for
running
the
end-to-end
test,
but
I'm
not
sure
that
there's
much,
you
know
I'm
not
sure
that
there's
much
relationship
downward
from
the
sonobui
test
into
the
end-to-end
testing
I
think
it's
more.
The
other
way
around
where
the
end-to-end
tests
are
base
layer
and
then
Sona
Brewery
can
ingest
them
to
run
them
in
specific
ways.
E
So,
if
you
have
a
cluster,
you
can
use
sonobui
to
automate
the
running
of
the
end-to-end
test
so
that
you
could
then
later
see.
If
it,
you
know,
passes
conformance
or
something
like
that,
and
that's
the
way
we're
using
it
inside
red
hat.
But
from
what
I've
seen
from
our
usage
of
it
I
think
the
relationship
is
more
of
sonobari
is
Downstream
from
the
end-to-end
test,
Suites
that
come
out
of
KK,
but
you
know
that's
just
my
impression.
As
a
as
a
user
episode.