►
From YouTube: Kubernetes SIG Testing 2019-02-26
Description
A
A
A
B
So
we've
looks
like
a
yesterday.
We
updated
proud
to
a
commit
from
February
23rd,
so
we
are
not
perfectly
at
the
most
recent
date
it
looks
like,
but
we
are
essentially
resolved
from
where
we
were
before,
because
we
were
previously
in
January
I'm,
not
sure.
There's
anything
blocking
us
from
updating
to
today's
commit
I
can
try
and
verify
that.
A
What
were
the
so
the
the
known
issues
that
were
affecting
us
last
week?
We
had
CPU
usage
issues.
There
was
the
mill
pointer
problem
and
hook
there
was
the
github
right
loading
issue
in
plank
weather,
any
other,
like
any
other
individual
issues.
You.
B
Know
the
triggering
problem
was
the
no
pointer
thing
so
yeah
that
that's
covered.
We
also
saw
the
problems
with
Pollard's
inside
the
service
cluster,
communicating
with
each
other,
but
we're
assuming
that
that
was
due
to
cpu
load
yeah.
So
we're
seeing
a
lot
of
symptoms
and
it's
kind
of
hard
to
analyze
exactly
what
all
the
causes
were,
but
it
I
think
we
have
we've
at
least
identified
some
of
those
causes
and
address
them
and
it
seems
to
have
fixed
all
the
problems
that
we
were
saying.
A
B
If
we
want
to
add
the
client
back,
I
think
we
should
just
slowly
add
it
back
two
more
components
and
see
if
we
get
any
more
problems
again,
it's
kind
of
unclear
if
that
was
actually
the
original
problem,
the
first
case
or
not,
given
that
there
were
so
many
other
issues
going
on
so
I
think
that
all
we
can
do
is
try
and
slowly
add
things
back
in,
but
given
that
we're,
we've
averted
everything
and
we're
now
able
to
deploy
successfully
again.
As
long
as
we
do
things
incrementally,
we
should
stay
in
a
healthy
state.
B
B
Don't
think
we
thought
about
that
yet
previously,
I
guess
the
only
one
that
we
thought
about
before
was
sinker
just
because
we
know
that
that's
a
safe
one
to
mess
around
with,
because
it
is
like
essentially
no
user
impact
I.
Think
that
should
be
our
guideline
for
what
ordering
we're
choosing
them
so
things
that
have
minimal
user
impact
if
we
have
to
muck
around
with
them.
A
A
B
A
So,
first
let
the
debug
handler
just
to
make
sure
we
can
do
traces
and
profiles
at
one
time.
Then
we
can
slowly
add
revert
spec
in
one
at
a
time
in
order.
These
are
in
fact
which
looks
kind
of
like
a
parrot
in
plank
and
deck
then
hook
and
tied.
If
sinker
and
flying
aren't
working
very
well,
it
doesn't
really
matter
like
C
or
D.
Resources
will
pile
up
the
pods
monka
schedule,
but
it'll
eventually
become
consistent.