►
From YouTube: 20191219 SIG Arch Community Meeting
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
A
A
A
Okay,
so
normally
we
do
so,
we've
started
doing
kind
of
standing
standing
agenda
for
this
meeting
AI
status.
There
were
no
realize
from
the
last
time
that
haven't
already
been
been
done.
We
cancel
last
time
at
the
time
before
them.
All
of
these
were
done.
So
there's
nothing
really
to
say
there
regarding
sub-project
readouts,
so
the
conformance
sub-project.
A
We
continue
to
make
progress.
The
ìiî
team
for
those
who
don't
know
CNC,
have
actually
pains
a
team
of
contractors
to
help
work
on
conformance
and
that's
led
by
hippy
in
occur
and
that's
AI
dot
co-op.
The
they've
been
working
on
a
number
of
things
and
identifying
waste
identify
existing
tests
that
hit
new
endpoints
that
we
haven't
been
hit.
They
haven't
been
here
before
so
they've
found
a
few
of
those
and.
A
I've
identified
and/or
started
creating
issues
around
tech,
new
tests
that
need
to
be
written
to
hit
these
end
points
things
like
that
as
well,
so
good
progress
there.
We
also
are
working
on
continue
to
work
on
defining
the
or
modifying
the
way
that
we
do
conformance
such
that
we
are
doing
listing
out
the
expectations
of
how
the
cluster
should
behave
with
the
API
should
behave
ahead
of
time
and
then
testing
them
separately,
and
that's
that's
progressing
slowly,
but
it
is
progressing
now
and
that
which
is
good.
A
Any
questions
on
that.
There's
really
not
that
much
to
report
right
now,
beyond
that
we
did
have
a
little
incident
where
I
I
have
to
say
I'm
a
call
by
approved
a
test
that
was
flaking
as
it
turns
out
that
that
test
wasn't
running
in
the
CI
due
to
some
other
issue,
and
so
we
didn't
see
the
flakes,
but
we
need
we
need
probably
more
diligence
on
ensuring
that
tests
aren't
flaking
before
we
push
them
forward
through
the
performance.
So.
B
Just
a
point
related
to
that
I
don't
know
if
anyone
saw
the
note
sent
to
kubernetes
dev
last
week.
Maybe
nobody
knew
that
our
CI
jobs
would
actually
silently
retry
flaky
tests.
So
if
you
had
a
test
that
was
flaky,
it
could
fail
and
get
retried
and
succeed
the
second
time
and
the
job
would
pass
and
it
would
report
green
status
and.
B
Would
never
even
know
that
it
was
leaky,
so
our
test
grids
were
lying
to
us.
Basically,
the
test
grids
that
you
would
look
at
and
say:
oh
let's
job,
but
this
particular
test
is
not
flaky
cool.
It
can
be
a
conformance
test
that
was,
that
was
a
lie
that
got
turned
off
for
master
last
Friday,
which
is
going
to
be
very
painful,
short-term,
but
very
good,
long-term
yeah.
B
Just
as
a
point
of
reference
since
the
beginning
of
the
week,
we
have
found
at
least
four
real
bugs
that
were
masked
because
of
this
two
of
them
pretty
significant.
There's
a
run,
see
race
condition,
cubelet,
pod
deletion,
race
condition
so
yeah.
So
if
you
are
wondering
why
you're
seeing
existing
conformists
tests
look
flakier
it's
because
they
were
actually
always
flaky
and
it
was
getting
masked
and.
B
Working
on
burning
those
down
and
any
that
are
actually
flaky,
causing
them
and
fixing
them,
and
if
we
can't
do
that
in
a
reasonable
timeframe,
like
order
of
a
few
days,
we
are
marking
them
flaky
and
taking
them
out
of
the
merged
jobs
and
opening
milestone,
critical
issues
for
the
owning
SIG's
to
say
you
have
to
resolve
this
basically
before
you
do
anything
else,
because
you've
got
code
that
isn't
being
tested.
So,
okay,
all
of
that
is
in
progress
and
it's
super
painful
short-term,
but
it
needs
to
be
done
and
yeah.
B
A
B
That
capability
is
now
there
we
updated
the
conform,
assess
that
test,
aggregation
to
a
117
level,
aggregated
back-end
server,
and
so
that
is
no
longer
using
beta
level
api's
to
do
authentication
authorization
checks,
so
the
last
API
that
is
not
GA,
that
is
used
by
the
things
bringing
up
clusters
for
conformance,
is
the
certificate
signing
request,
API,
and
so
there
is
a
motion
on
the
designs
around
that
to
identify
the
changes
that
need
to
be
made
or
additions
that
need
to
be
made
to
get
that
to
be
one.
So
what.
B
It
was
basically
an
API
that
was
designed
with
a
very
narrow
focus,
and
there
are
a
few
things
that
would
make
it
an
actual,
generally
useful
API.
B
B
We
need
a
bit
and
if
you
had
to
say
what
is
this
thing
even
for
and
then
sort
of
on
the
flip
side,
when
the
signer
issues
this
you
difficut
a
way
for
it
to
distribute
back
like
here,
are
the
intermediates
and
here's
the
route
like
so
okay,
some
signers
might
indicate
that
those
are
distributed
out-of-band,
but
some
might
need
to
be
able
to
say
here's
your
certificate
and
here's
the
trust
routes
that
you
need
to
distribute,
along
with
the
certificate.
Okay,
so.
A
B
A
Accessing
through
a
layer
so
in
proxy,
with
a
different
path,
acute
controller,
you
can
put
a
different
path
on
to
you
know
on
the
on
the
destination
that
URL
for
the
API
server,
and
we
wanted
to
show
that
we
can
do
that.
That's
a
conforming
behavior
that
you
can.
You
can
put
your
eight
guys
over
behind
a
layer,
seven
proxy
and
have
to
work
as
you'd
expect
at
different
paths,
but
in
which
case
we
might
want
to
use
ingress,
although
that's
doesn't
need
to
be
done
that
way.
So
you
know
anyway
beside.