►
From YouTube: Kubernetes SIG Node CI 20221116
Description
SIG Node CI weekly meeting. Agenda and notes: https://docs.google.com/document/d/1fb-ugvgdSVIkkuJ388_nhp2pBTy_4HEVg5848Xy7n5U/edit#heading=h.2v8vzknys4nk
GMT20221116-180416_Recording_1762x1286
A
A
A
A
or
1.5,
so
continuity
1.5
will
not
supported
any
longer
and
somehow
test
jobs
started
failing
because
of
that
I
mean
obviously,
we've
been
testing
our
compatibility
with
1.5
before
and
now.
We
cannot
because
it's
just
incompatible.
So
there
was
a
work
happening
to
switch
tests
for
1.5
to
kubernetes
1
to
25,
so
only
so
Canada
G
1.5
will
be
tested
with
125.
Only
and
it's
all
good
one
thing
that
we
discovered
during
this
switch
is
that
our
quite
fragile
to
the
situation
when
different
things
I
been
used.
A
So
in
this
example,
let
me
open
so
in
this
example.
What
was
used?
Let
me
scroll
up
so
yeah
secret
V2
will
so
when
we
also
this
job.
There
was
a
c
group
with
like
they
switched
to
latest
course
version
and
latest
cost
versions
using
C
group
E2,
while
security
support
wasn't
added
there
and
different
driver
was
used
in
kublet.
So
mismatch
of
driver,
runtime
and
qubit
led
to
very
strange
errors
that
were
really
hard
to
discover
and
yeah.
A
I
I
will
try
to
pull
off
his
David
tunnels
and
how
he
found
this
issue,
maybe
just
by
looking
at
configuration.
But
what
concerns
me
is
that
this
misconfiguration
between
runtime
and
kublet
is
not
easily
discoverable.
So
that's
a
first
issue.
A
Yeah
we
already
fixed
it
because
container
D
1.5
had
a
critical
issue
and
we
needed
to
fix
it
and
there's
another
topic
right
here.
So
containerd
is
using
kubernetes
tests
as
a
CI
for
their
jobs.
So
we
we
blocked
1.5
bug
fixes
because
our
entrance
wasn't
working.
It
was
a
problem.
A
Yeah
but
yeah
this
is.
We
fix
that
problem,
so
content
energy
is
now
unblocked,
but
we
still
have
this
problem
of
mismatch
is
not
easily
discoverable.
A
Another
topic
I
wanted
to
bring
up
is,
as
we
switch
into
latest
versions
of
operating
systems,
we're
switching
default
to
see,
group,
V2
and
I
think
what
will
happen
is
C.
Group
V2
will
be
primary
test
Target
from
126
on,
because
all
the
our
latest
Ubuntu
and,
of
course,
and
I,
don't
know
about
how
anything
but
Ubuntu
cost
will
be
switching
to
see
group
V2
as
a
default,
and
we
will
test
with
defaults
as
a
primary
target
and
non-default
C
group
E1
it'll
be
a
secondary
Target.
A
We
still
fully
support
C
group
E1,
but
we
need
to
make
sure
we
need
to
understand
that
it's
now
over
time.
It
will
be
harder
and
harder
to
test
with
C
group
you
want,
so
we
need
to
make
sure
that
customers
are
migrating
to
crfp2
and
whenever
we
do
changes,
we
keep
in
mind
that
there
is
a
secondary
targets
that
I
also
need
to
test
for.
A
And
switching
to
C
group
V2
by
the
way,
also
making
this
mismatch
of
driver
between
Cobalt
and
runtime
problem
harder,
so
customers
will
be
switching
and
they
may
misconfigure
something
and
we'll
have
this
hard
to
discover
problem
on
their
hands
so
yeah,
something
that
I
want
to
follow
up
and
if
somebody's
interested
to
join
me.
Please
do
okay,
oh
yeah,
it's!
It
was
different.
A
And
yeah
it
is
broken.
We
currently
have
nothing.
B
B
A
Nothing
moved
I
mean
there
were
no
major
changes
recently.
So
everything
looks
good
and
CPU
looks
good,
so
yeah
I
think
nothing
to
report
on
okay.
Is
there
any
other
agenda
items
for
today
or
we
will
switch
to
Quick
triage
and
okay?
Let's
switch
to
triage.
A
A
A
A
About
disruption
conditions,
so
in
jobs,
you
will
now
be
able
to
specify
whether
you
in
the
job
controller,
you
can
specify
whether
you
need
to
retrieve
the
job
based
on
conditions.
So,
let's
say
if
your
job
was
killed,
you
can
decide
now
whether
you
want
to
restart
this
job
or
not
restarted
shop.
A
A
B
Yeah
and
we.
A
Decided
we
needed
to
test
some
disruption,
conditions
and
disruption.
Conditions
were
added
into
serial
Lane
and
serial
tests
are
now
failing
with
eviction
tests.
So
we
wanted
to
make
sure
that
we
test
in
this
cap,
so
we
created
this
disruption,
condition
separate
job
and
I.
Think
this
is
fixing
it
somehow
so
I
will
I
will
take
a
look
if
somebody
interested
in
this
as
well.
Please
assign
themselves.
B
A
Yeah
I
will
go
ahead
and
archive
it
from
this
board,
but
I
would
be
interested
to
look
at
this.
A
Okay,
maybe
over
here
foreign.
A
Yeah
eight
files
like
it
attaches
everything
right.
So
maybe
we
need
to
look
at
it
from
perspective
of
Sig
node,
so
signal
tones.
A
B
A
But
I
think
it's
for
context.
Yes,
why
is
in
the
context.
B
A
A
Okay,
we're
done
with
this
part.
A
Anybody's
interested
to
look
at
funeral.
A
If
I
I,
just
don't
know
whether
it's
a
limitation
that
was
intentional
or
it
just
a
small
bug
somewhere,
that
we
can
easily
fix
I
I,
just
maybe
if
your
metal
storage
cap
has
a
description.
B
A
A
B
B
So
I
try
to
talk
to
people
who
know
about
ephemeral,
pod,
so
I'll.
Take
it
and
I'll
try
to
talk
to
somebody
who
knows
and
I'll
just
try
to
give
some
status
on
the
ticket.
Okay,
that.
A
A
As
well,
okay,
let's
make
a
first
step
to
ask
a
Google
Talk.
B
B
A
A
No
I
think
I
think
they're,
saying
it's
failing
all
the
time
yeah,
so
a
response
body
is
zero,
and
so
one
or
two
of
two
issues
first,
is
we
incorrected
using
response
that
is
zero
size
and
second
possibility
is
of
basically
content
like
this
is
159?
It's
not
zero.
A
No,
it's
not
that,
and
it
may
be
that
some
connectivity
issue.
So
let's
say
I
I
saw
customers
doing
crazy
things
with
foreign
not
like
having
checking
the
incoming
requests
and
the
user
agent
is
different
or
if
request
is
coming
from
localhost
and
they
really
reply
differently
from
that.
C
A
B
A
A
A
C
A
Okay,
thank
you.
Everybody,
sorry
for
running
this
meeting
short
I
think
we
need
to
cancel
next
week
because
she
not
made
meeting
main
meeting
who's
canceled
and
many
people
will
not
be
around
because
of
holidays.
So.
C
A
Thanksgiving,
who,
who
to
whoever
celebrates
sorry
for
imposing
the
U.S
holidays
on
everybody
but
yeah.