►
From YouTube: Sig-Testing Weekly Meeting for 20230110
Description
Sig-Testing Weekly Meeting for 20230110
A
A
If
I
remember
correctly,
I
think
we've
mentioned
it
in
a
previous
meeting
as
well,
but
we
have,
if
nothing
else,
slots
for
maintainer
tracks
so
stuff,
like
the
Sig
testing
introduction,
which
is
also
encouraged
as
a
panel
thing
to
be
representative
of
folks
across
a
bunch
of
different
companies
and
not
all
sticking
to
the
same
one.
So
basically
I'll
put
out
a
call
as
well
in
the
channel.
A
A
B
I'm
gonna
working
in
VMware
and
so
basically
I'm
working
on
the
Sona
boy
project
and
while
working
on
the
suno
quiet
project,
we
found
some
issues
reported
by
some
some
of
the
users.
Basically,
is
that
like
when
they
use
the
e2b
framework
right
and
if
so,
the
e2e
test
actually
spawns
ports
in
separate
namespaces,
and
if
those
pods
use
private
images
from
the
private
registry
they
go
into
the.
A
C
Oh
yeah,
so
that's
from
me
right
so
I'm
thinking
of
deploying
some
slightly
complicated
applications
to
the
kubernetes
Clusters,
the
ones
under
kubernetes
that
I
O,
not
the
ones
that
are
not
the
proud
control
plane.
But
these
apps
are
like
bundleless,
Helm
charts,
for
example.
So
they're
not
really
kubernetes
manifests
that
sit
in
the
repository.
So
I
was
thinking.
A
A
D
I
mean
we've
definitely
had
problems
with
computer
Cube
cuddle.
We,
the
I,
think
the
biggest
problem
we
have
is
printing
for
everything
else.
I
think
we're
very
happy
with
having
manifest
checked
in
that
exactly
represent
what
we're
doing.
But
if,
if
we
prune
something
then
we
have
to,
we
tend
to
have
to
do
that
manually.
D
The
last
time
we
removed
a
prow
component
was
taught
okay
a
couple
years
ago,
so
that
hasn't
been
a
big
issue
in
some
other
parts
of
the
project
in
kids,
him
for
a
there's
a
little
bit
more
turn,
but
even
there
it
hasn't
been
a
super
big
deal
to
have
someone
apparently
delete
something.
It
isn't
ideal,
though
I
think
you
can
I
think
if
you
have
a
namespace,
it's
possible
do
pruning.
C
E
D
Yeah
I
I
actually
have.
We
haven't
done
this
approach.
But
if
you
like
in
the
case
that
you're
applying
a
directory
and
you
you
use
a
label,
it's
possible
to
have
Cube
cuddle,
do
pruning,
but
yeah
I,
don't
think.
We've
used
Argo
CD
anywhere.
We've
used
everything
from
bash
and
rules
Kates
and
terraform,
and
G
Cloud,
CLI
python,
but
I,
don't
I,
don't
think
any
I've
seen
any
parts
of
the
kubernetes
project
using
arcocd.
C
C
B
C
C
C
D
D
There
was
an
older
one
that
it
replaced.
It
was
inside
of
Google.
That
was
done.
It
was
like
something
like
kubernetes.
Public
was
a
gcp
project
with
a
tools,
I
might
be
confusing
names,
but
there
was
like
a
tools
or
utility
cluster
and
AAA
I.
Think
it's
named
that
because
it
was
intended
to
be
shortlisted.
Okay,
I
think
Tim
was
like
We'll
spin
up
something,
and
you
can
put
things
here
and
we'll
figure
it
out
later
and
then
it
didn't
go
away.
So.
F
A
E
F
A
A
G
I
do
have
Apple
request
on
KK,
which
basically
picks
all
those
prequels
I
mentioned,
plus
a
few
other
ones,
which
adds
more
coverage
to
pretty
much
a
lot
of
the
things
in
kubernetes,
and
that
should
be
passing
already
with
everything.
So
that's
one
of
the
goals
that
I
have
for
the
for
this
release,
at
least.
F
F
G
D
We
kind
of
on
the
we
kind
of
on
the
clustered
e
to
B
test.
We've
helped
Shepherd
that,
but
we're
really
only
Exposed
on
the
framework
there,
but
in
practice
people
who
own
the
edoe
framework
own
the
tests
as
well
it's
more
convenient
but
with
the
unit
tests,
because
prow
doesn't
have
granular
ownership.
Yet,
even
if
we
wanted
to
set
that
up,
you
would
effectively
have
to
give
Sig
testing
approval
over
like
all
of
the
all
of
the
source
code.
D
D
D
We
should
get
this
CI
signal
release
blocking
as
soon
as
it's
green
or
well
informing,
and
then,
if
you
see
release
blocking
signal,
frequently
broken
by
something
that
should
have
been
caught,
that's
our
signal
that
we
need
to
start
running
it
on
every
PR.
It's
a
lot
more
expensive
to
run
it
on
every
PR.
D
If
it's
flaky,
you
know
it
blocks
people's
PR,
so
we
actually
periodically
want
to
reevaluate
and
look
at
moving
things
out.
But
when
we
see
something
that
is
in
release
blocking
signal,
that
is,
it
needs
fixing
on
a
regular
basis,
because
it's
because
things
are
not
getting
caught
and
it's
difficult
to
track
them
down
then
so
like,
for
example,
the
some
of
the
scale
testing
is
the
kind
of
thing
that
isn't
actually
run
on
every
PR.
We
can
it.
G
D
Yeah
that
would
be
similar
to
I
mean
the
e2e
test.
For
example,
I
mean
one
of
the
ways
that
those
that
the
important
jobs
that
are
kept
healthy
is
that
sometimes
that
unit,
an
e
to
be
test
is
deemed
like
not
important,
kicked
out.
We're
like
this
is
fundamentally
a
flaky
test,
because
the
design
or
something
it
gets
tagged
flaky
or
whatever.
We
don't
do
it
a
lot.
D
These
days,
I
think
we
have
a
pretty
stable
set,
but
in
the
past
we
used
to
pretty
often
like
skip
some
tests
out
of
CI
Suites
I
hadn't
seen
a
unit
tests
before,
but
I
think
that's
fair
for
getting
the
signal
in
place
because
also
to
meet
release
blocking
status.
You
do
need
to
show
that
the
job
has
been
stable
for
a
while.
So
that
makes
sense
to
me.
G
D
D
We
never
got
this
written
up
for
prison
properly,
but
Aaron
in
particular
worked
with
release
on
the
release
blocking
one.
So
we,
you
know,
we
push
things
there.
First
I
mean
it's
also
the
most
value
that,
if
you
get
it
into
release
blocking
status,
then
we
will
not
ship
a
release
without
it
being
fixed
and
CS
signal
will
also
be
helping
like
keep
an
eye
on
and
file
issues
for
failing
tests
for
things
that
meet
that
bar
and
that
some
of
those
things
need
to
be
copied
over
to
produce
a
bit.
F
D
D
Okay,
well
I'll
mention
this
year,
we'll
be
looking
for
more
chairs.
In
particular,
Michelle
is
doing
almost
all
of
the
chair
work
right
now,
Aaron
and
Steve
I'm,
not
sure
exactly
about
their
status.
I'll
reach
out
of
them
that
haven't
been
around
and
for
a
while,
I
admit,
I
kind
of
intentionally.
Let
Michelle
focus
on
it
to
you,
know,
cement
that
position,
but
I
you
know,
I
have
now.
D
I've
have
steering,
and
some
other
things
and
I'm
definitely
not
doing
enough
of
my
part
here
so
I'm
looking
to
not
be
part
of
the
problem
and
to
rotate
out
of
some
of
these
suggesting
positions.
We
have
added
some
chairs,
but
we've
only
or
some
tech
leads.
We've
only
had
one
chair
I
would
like
to
find
at
least
one
more
chair.
D
So
if
any
of
you
are
interested
we're
going
to
be
trying
to
figure
something
out,
we
talked
to
contributor
experience
at
the
end
of
last
year
about
how
they
run
chair
mentoring,
and
we
haven't
gotten
that
running
yet,
but
we'll
be
bringing
that
back
up
and
I
wanted
to
go
ahead
and
mention
here
that
you
know
we'll
be
looking
for.
Someone
to
help
run
the
chair
role,
doing
the
meetings
and
the
annual
report
and
some
things
like
that.
D
No
I
don't
know
that
better,
but
I
know
it's
something.
That's
been
talking
about
we're,
there's
a
there's
a
so
stick.
Testing
is
well
ahead
of
the
curve
there,
but
there's
a
bit
of
a
general
push
to
have
six
split
chair
and
Tech
lead
roles
so
that
more
people
can
be
recognized.
The
workload
can
be
spent
around
a
better
continuity
Etc.
Historically,
the
chair
role
in
six
has
kind
of
been
a
combination
of
chair
and
Tech
lead.
D
Sig
testing
does
have
separate
chair
and
technically
rules,
and
you
can
actually
be
in
both,
but
they
are
separate
listings
and
they
have
different
requirements.
Tech
leads
are
mostly
expected
to
be
helping
approve,
spinning
up
and
shutting
down,
sub-projects
and
caps,
but
in
practice
even
caps
like
while
they
technically
requires
to
sign
off
it's
pretty
common
to
delegate
some
of
the
review
to
other
people
that
know
about
these
things.