►
From YouTube: Kubernetes SIG testing 20171010
A
B
Had
a
PSA
that
won
8-1
is
slated
to
go
out
later
this
week.
There
are
a
couple
of
testing
items
that
we
know
about
that
we're
trying
to
get
fixed
in
the
intent
tests.
So
if
you
know
of
any
other
items
that
need
addressing,
please
try
to
get
them
in
in
the
next
couple
of
days
and
please
either
CC
the
this
sig
or
also
CC
I,
believe
his
name
is
his
handle
is
at
j1
second
JPB
etz
and
that's
all
I
have
for
psays.
A
C
B
So
that
way,
the
the
upgrade
process
would
just
be
like
a
rolling
upgrade
right
with
some
type
of
strategy,
and
then
you
could
monitor
the
results
of
that
as
part
of
a
test
right,
because
we
don't
do
that.
The
process
is
kind
of
jiggery-pokery
for,
however,
which
way
somebody
is
deployed.
These
things
right.
C
B
For
the
most
part,
yes,
I
think
that's
the
biggest
problem.
Most
of
the
sku
version
problems
we
come
across
are
not
necessarily
cloud
provider
specific,
although
I'm
sure
there's
probably
a
set
of
problems
that
exist
there
most
of
the
ones
we
have
that
I've
seen
at
least
have
been
just
base
level
functionality,
API
promotions,
things
like
that
which,
when
they
go
from
version,
8
a
version
B,
they
they
break
everything
because
they
switched
API
group
serve.
This
they've
done
some
other
shifting
in
the
api's.
C
C
New
I
guess
I'll
follow
up
and
see
if
I
can
get
my
cube
on
cube
solution.
Working
I
know
that
if,
if
it's
something
we
can
get
working
fast
enough,
it
may
be
of
general
interest
to
test
infra
beyond
upgrade
tests,
because
it
would
just
reduce
the
cost
of
deployment.
You
wouldn't
need
nodes.
You
just
need
an
existing
kubernetes
cluster
and
then
you
could
just
apply
with
standard.
You
know
cube
infrastructure
eating
your
own
dog
food
as
it
were
anyway,
so
yeah
I'll
follow
up
and
maybe
have
something
back
next
week.
D
C
C
Maybe
it's
a
back
story
like
what's
what's
the
is
this
existing
state
of
a
sonication
for
problem
I.
A
A
A
D
C
A
Yeah
and
now
I
understand
yeah
I,
don't
know
if
anyone's
really
given
a
whole
lot.
This
like
right
now,
I
guess
you
know
the
the
logs
would
be
I
guess.
If
we
wanted
to
be
have
the
logs
rather
than
blogs
and
everything
being
public,
they
would
be
more
privately
available
and
then
yeah
I
guess
potentially
have
all
the
infrastructure.
It
seems
like
a
lot
of
infrastructure
that
have
come
through,
though,
because
like
test
grid
and
ruminator
and
prowl
I'll
use
those
logs.
A
D
Could
you
could
quite
easily
so
if
you
had
some
way
of
flagging
a
proud
job
as
sensitive,
not
ruin
logs
a
bit
Edward
guberniya
does
know
what
your
github
username
is
and
that's
reasonably
secure.
So
we
can
have
a
whitelist
on
people
that
trivial
to
see
those
it's
still
kind
of
hacky,
but
at
least
you
wouldn't
have
to
redeploy
everything,
and
presumably
you
would
upload
it
to
a
separate
bucket
where
goober
nadir
would
have
read
access
and
it
wouldn't
be
color.
Creed,
mm-hmm,.
D
D
Just
like
we
currently
have
operators
where
we
can
specify
that
this
runs
on
Jenkins
or
in
this
runs
on
crannies
because
specify
the
blogs
are
hangings
or
I
could
even
the
job
itself.
Is
it,
but
how
much?
If
it's
an
embargoed
CV?
How
much
could
you
show
like
the
fact
that
there's
a
that
we're
running
the
secret
jobs?
Do
we
have
to
hide
the
jobs
altogether?
A
So
Steve
has
a
few
more
comments
here
on
slack,
so
Kenna
blogs
exposed
to
this
test
and
can't
have
logs
of
those
tests
expose
and
crown
goo
brands
with
you
today
we
don't
and
that's
every
deployment
for
everything
we
do
either.
So
when
I
have
an
office,
a
first
class
idea,
often
these
components
just
get
out
or
whatever
to
get
access
to
the
logs
an
overview
before
me
again
how
she
got
some
ideas.
Suppose
I
just
want
to
ask
generally
if
community
felt
like
this
is
something
we
wanted
to
do,
separate,
buckets,
etc.
A
D
Well,
governor
has
the
OAuth
flow
already
built
in
so
that
would
be
the
one
really
easy
place
to
put
most
of
this,
and
it
also
has
the
recent
runs,
which
looks
something
like
what
you
get
from
test
grade
or
whatever.
So
you
could
it's
you
it
so
we
could
get
something
hacky
set
up
pretty
quickly,
but
anything
beyond
that
we'll
have
to
discuss
otherwise
right.
A
A
If
the
general
agreement
is
not
yuck,
we
can
do
some
design
ideas,
he
writes
so
I
think
the
general
idea
or
the
general
agreement
seems
like
this
seems
fine
I,
don't
think
I,
don't
hear
any
major
objections
to
this.
So
I
think
it
seems
reasonable.
I
think
my
sort
of
at
least
and
working
on
some
of
the
grout
configurations,
I,
feel
like
sort
of
the
community
security
repository
has
sort
of
been
neglected,
abandoned
not
abandoned.
There
like
it
seems
like
sort
of
the
test
in
front.
A
We
haven't
really
done
any
work
on
making
this
actually
work
or
making
sure
it
still
works.
I
know:
we've
all
done
some
work
in
the
past
to
get
it
all
set
up,
but
it
may
have
been
working
at
some
point,
but
I
think
it
would
be
good
yeah,
for
somebody
don't
make
sure
that
all
this
stuff
works,
because
right
now,
it's
kind
of
there
on
the
side
and
sort
of
its
kind
of
maintaining.
But
nobody
actually
knows
we
can.
D
D
C
A
F
F
D
D
And
they
seem
to
be
working
fine,
but
that's
not
it's
minimally
tested
Joe
is
working
on
the
front
end
for
that,
so
you
can
actually
see
some
state
for
the
system
and
once
that's
available,
we
can
roll
it
out
to
more
rebuilds
and
perhaps
even
think
about
it
replacements
to
make
you
a
fit
on
the
truck.
Let's.
E
F
All
of
that
done
by
the
end
of
the
quarter
for
sure,
hopefully
sooner
but
yeah,
there
should
be
no
more
student
used
by
named
a
quarter.
Okay,.