►
From YouTube: 20180425 sig cluster lifecycle kubeadm office hours
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
Hello
today
is
April
25th
2018.
This
is
the
coop
ADM
office
hours
and
we
don't.
We
have
a
some
thorny
stuff
on
the
agenda,
so
Lee
had
written
it
down.
We
have
one
cherry-pick
that
were
pending
on
to
go
backwards.
You
want
to
start
to
kick
off
the
conversation.
I'm
sure
we're
gonna
kind
of
sprawl
from
there.
A
I'm,
okay,
with
the
copying
of
the
the
problem,
is
the
mutation
of
the
config
doesn't
have
a
backwards
mutator
right,
but
it
goes
it's
kind
of
broader
than
that
in
that
I
don't
think.
Downgrades
in
general
are
kind
of
a
supportable
thing
once
we
start
to
getting
to
high
availability,
we're
just
having
a
conversation
about
this
and
it's
a
second
ago
without
a
full
backup
and
restore
utility.
A
So
there's
there's
a
bunch
of
don't
think
it's
a
supportable
thing
that
we
should
be
doing.
Instead,
we
should
have
a
we
should
we
should
punt
to
other
tooling
or
other
processes
and
policies
about
like
before
you
upgrade
make
sure
you
do
XYZ.
That's
my
personal
take
on
it,
but
I
do
recognize
that
that
will
alienate
some
folks
because
it
is,
it
makes
it
much
more
cumbersome.
B
Yeah
one
thing
is:
there
was
an
issue
that
came
up
when
Joe
Beda
did
TGI
K
about
the
backup,
tear
disappearing,
so
I
just
linked
a
related
issue
in
an
additional
bullet
points
about
keeping
the
old
manifests
and
potentially
like
the
X
key
back,
but
that
around
also
I
guess.
We
could
probably
put
in
a
comment
here
about
storing
the
old
config
and
storing
at
CD.
B
A
Well,
beyond
that,
the
the
mutation
of
the
API
objects
itself
occurs
on
API
server
startup
on
the
first
access
of
any
resource
object
inside
of
kubernetes.
That's
the
way
the
API
server
API
machinery
works.
So
if
I
did,
if
I
did
a
list
on
an
API
server
start
because
the
controller
restarted
on
an
upgrade
right,
so
it
was
listing
all
the
resources.
If
there
is
any
mutation
of
any
object,
they
would
be
mutated
forwards
and
there
is
no
backwards.
Compatibility,
mutate
backwards
right.
B
A
C
B
I
mean
what
I'm
getting
at
is
that
we
already
have
an
existing
rollback
mechanism.
That
is
best
effort,
and
so
it
seems
you
know
like
it
doesn't
seem
infeasible.
You
know
to
basically
just
not
delete
those
directories
and
then
like
provide
a
simple
interface
to
actually
rerun
the
rollback.
You
know
I.
C
A
B
I
mean
I,
don't
think
we
need
to
solve,
for
that
like,
but
sure
would
be
nice
to
like
in
the
event
of
a
failure
scenario.
Even
if
somebody
is
running
an
H,
a
cluster
with
every
Bell
and
whistle
enabled
to
like
provide
them,
the
breadcrumbs
that
are
necessary
so
that
they're
unhappy
path
isn't
so
miserable.
Well,.
A
A
B
B
A
I
think
before
it
starts
right
in
the
very
initial
phases
of
the
upgrade
plan,
I'd
have
to
take
a
look
at
what
phase
is
the
precursor
in
the
very
beginning
of
upgrade?
It
is
to
squirrel
the
data
away
right
locally
right
like
to
have
an
even
to
have
it
well-defined.
It
seems
like
a
reasonable
thing.
Yeah.
B
C
A
A
B
A
B
Yeah
I
mean
if
I
guess
like
the
important
thing
is
that
we
documents
or
like
make
known:
that's
the
one
nine
binary
like
if
you
maybe
run
the
static,
pod
control,
plane,
thing
or
I,
don't
know
if
we
just
add,
like
a
format
like
a
print
statement
to
the
upgrade
the
upgrade
plan
where
we
we
say
hey,
you
know
like
downgrading
from
one
sense,
a
one.
Nine
is
no
longer
supported.
So
please
back
up
your
manifests,
I
mean
like
apps,
you.
A
Can
you
could
even
do
that
in
the
release?
Note
so
like
the
release
know
for
your
cherry-pick,
should
basically
it
we
should
have
been
on
Liz's
cherry
pick
originally,
but
it's
too
late
now.
But
if
you
put
it
release
note
on
your
cherry
pick,
it
will
show
up
in
the
general
release
notes
for
this
thing
and
it
will
be,
it
will
be
there
for
all
eternity.
That'll
be
the
first
stage
and
then
I
think
the
second
stage
will
be
like.
We
need
an
issue
in
a
doc
fix
with
both
of
these
things.
A
B
A
C
B
I
mean
well,
we
had
a
lot
of
that
stuff
this
for
this
cherry-pick
cycle.
We
did
yeah,
Jason
and
I
ran
into
all
kinds
of
very
strange
bugs
with
the
upgrade
where
it's
like
wow.
How
did
this
ever
work,
but
a
it
same
thing
as
well?
The
sed
rollback
actually
was
non-functional
it
would
it
would.
It
was
guaranteed
to
to
error
out
all
right
anytime.
The
NCD
upgrade
failed.
So
that's
that's
now
fixed,
but
so.
A
How
does
this
sound
as
a
path
forward?
Add
a
release,
note
that
outlines
the
constraints
that
we
have
to
your
cherry-pick
we'll
get
that
thing
emerged
as
soon
as
possible
and
then
open
an
issue
with
regards
to
our
you
know,
eliminating
downgrade
and
issue
to
also
document
the
process
for
upgrades
much
more
rigorously.
A
A
A
A
C
B
I
mean
I
I,
put
up
a
PR
this
morning
with
the
110
conformance
results
for
the
CN
CF
conformance
repo
Tam
I
know
that
you've
been
lightly
involved
with
some
of
the
son
of
Bowie
stuff.
So
I
don't
know
if
you
have
any
concerns
there,
but
yeah
I
put
up
that
and
maybe.
B
A
C
Have
my
very
initial
pass
at
some
IDI
tests
that
I
wrote
that's
up
for
merge
right
now
and
I
have
a
emerge
request
in
queue
test
to
provide
a
runner
for
it,
I'm
getting
some
pushback
about
people
who
don't
think
that
it's
quite
in
a
candidate
for
inclusion
and
cube
test.
So
if
I
can
I
can
link
that
PR
and
people
can
either
back
me
up
or
turn
against
me.
Sorry.
A
Thing
I
did
want
to
discuss
because
we
don't
have
all
the
automation
in
place
and
we
want
to
get
it
fixed
in
the
111
cycle.
One
thing
I
wanted
to
open
up
an
issue
about
to
discuss
is
to
have
a
well-defined
test
plan
that
you
know
any
person
within
the
cig
could
execute
and
give
feedback
to,
because
without
that
test
plan
we
are
look
like
you're
stumbling
across
things,
lis,
that
we
should
have
written
down
and
be
able
to
execute
pretty
easily.
B
A
And
it's
not
that's
not
just
any
binary
from
a
previously
released.
It's
a
cross
built
weirdo,
binary,
that's
not
being
maintained,
so
I
need
to
actually
get
it
to
the
I
was
going
to
the
fix
or
hack
I
was
gonna,
get
to
try
and
fix.
That
problem
was
to
just
use
the
full
release
binary
and
not
the
cross,
build
product
stuff
that
was
initially
added
and
I
know.
I.
B
I
mean
maybe
in
the
the
end-to-end
tests
that
we
have
currently
I.
Guess
that's
what
what's
being
depended
on
I
I,
don't
honestly
know
how
to
run
those
locally.
So
that's.
A
Part
of
I
think
the
reason
for
Liz's
work
is
to
be
able
to
have
a
single
binary
that
we
can
execute
locally,
that
anybody
should
be
able
to
run
as
part
of
the
their
local
build
environment
to
validate
it,
and
if
we
have
a
document
and
to
go
along
with
it
for
the
test
plan,
I
think
that's
even
better.
Oh.
B
C
I
think
that's
the
the
major
thing
for
the
test.
I
would
love
to
get
to
the
point
where
we
could
use
either
Q
tests
or
communities
anywhere
or
something
to
stand
up
the
actual
tests
and
run
a
more
power
edie
thing
right
now.
It
basically
just
verifies
that
after
you've
done
cube,
ADM
in
it
and
you
maybe
a
join
and
stuff
that
all
of
the
artifacts,
we
expect
to
be
they're
still.
E
C
B
B
I
have
brought
up
this
repo
before,
which
is
the
vagrant
box,
that
I
use
and
kind
of
the
setup
and
scripts
like
just
very
simple
stuff
that,
like
it's,
somebody
from
0
to
60,
so
I
mean
I.
Don't
know
if
that's
like
useful
for
this
environment
or
if
we
should
do
anything
else
to
prepare.
But
this
might
be
a
chance
to
like
get
more
more
contributor
cycles
and
more
people
introduced
to
the
tooling
that
we
have
I.
A
Think
on
the
main
repository
for
development,
you
can
absolutely
add
some
developer
notes,
like
the
main
actual
cube
ad
Enrico
and
add
happy
to
merge
anything
at
this
point
that
it's
people
up
to
speed
faster
and
we
can
refine
that
over
time.
You
know
I'm
sure,
as
people
look
at
the
repository,
the
Covidien
repo
for
it
from
the
development
perspective
that
that
the
main
documentation
should
not
have
development
Doc's
just
in
general,
but
the
the
comedian
repo
can
have
whatever
wants.
A
B
I
think
I
think
we
already
have
plenty
of
resources
for
kind
of
like
the
ladder
of
development
needs
so
like
what
I've
been
primarily
concerned
about
is
that
first
rung
is
like
how
do
you
get
somebody
from
like
what
is
this
go
file
to
like
actually
compiling
the
cuvette
idea
in
binary
and
like
running
it
and
testing
it,
and
then,
like
there's,
there's
all
the
other
stuff
that
we
could
link
to
as
well.
B
A
B
A
A
For
that
session,
three
parts
one
is
like
here
is:
the
here
is
the
initial
scope,
and
history
here
is:
come
some
of
the
things
around
HH
self-hosting
and
upgrades:
here's,
the
here's,
how
you
do
configuration
for
your
prizes
where
meets
and
some
of
the
constraints
from
them.
So
it's
a
lot
of
content
in
a
pretty
short
period
of
time.
Okay,.
B
Cool,
so
that
sounds
like
pretty
information
dense
agenda,
which
means
that
you
know
it
doesn't
sound
like
it'll,
be
like
a
workshop
style
thing.
No.
A
A
B
A
C
B
A
D
I
was
thinking
about
the
idea
of
having
a
separate
like
command
for
Haq.
Idiom
I
was
wondering
if
we'd
consider
that,
because
it
seems
like
a
lot
of
these,
various
problems
are
like.
Are
we
doing
an
HOA
upgrade
or
NHA
downgrade,
which
is
not
supporting,
but,
like
some
things
are
supported
for
some
environments
and
something
and
not
supported
for
so
environments?
So
may
it
might
be
unclean
way.
Just
have
a
separate
command
entirely.
B
That
to
me
sounds
like
encoding,
similar
information
that
would
be
previously
proposed
as
like
having
explicit
configuration
structs
for
the
different
types
of
at
CD.
The
different
types
of
upgrades
I
think
that
way
that
also
encoding
that
information
in
separate
phases
and
a
separate
command
entry
points
is
reasonable.
I,
don't
think
that
that
is
a
bad
idea
and
we
should
think
about
it.
A
Yeah
I
think
I
think
we
need
to
enumerate
the
Delta
to
about
you
know
if
we
talk
about
upgrades
that
one
thing
that
I
kind
of
introspect
too
much
and
having
the
discrepancy
of
a
versus
B
actually
listed
down,
would
be
super
helpful
because
we
we've
identified
key
gaps
that
exist
there.
But
if
we
actually
had
like
these
are
all
the
things
that
you
do
sort
of
like
here's.
Your
upgrade
plan
here
would
be
your
up
clip
and
if
you
did
AJ
and
here's
how
it's
different.
E
A
A
That's
pretty
funny.
I
have
a
kid
who
loves
her
out.
That
series
is
its,
but
there
was
there
was
a
line
from
there
anyways.
If
you
want
to
call
it
something
else,
we
can
call
it
that,
but
that's
probably
gonna
be
the
big
one
that
we
have
to
rally
room
for
111,
because
the
upgrade
configuration
nightmare
was
something
that
was
unexpected
yeah.
B
F
B
B
D
Yeah
about
the
audit
params
changing
in
one
of
them,
I'm
not
sure
about.
F
The
top
of
my
head
so
I,
just
it's
deprecated
and
it's
in
the
area
and
this
pull
request
and
so
I
wasn't
sure
one.
What
might
PR
be
against
master
or
something
else,
and
then
is
it
a
one-to-one
mapping
when
we
do
a
compile
of
Kubb
admin?
Should
it
map
precisely
to
the
version
of
kubernetes
we're
deploying.
B
Yeah,
so
to
your
last
question,
which
is
what
version
of
kuba
ADM
or
what
versions
of
kubernetes
does
kebaya
TM
support
the
plane?
As
of
this
release,
we
are
only
supporting
the
current
minor
version,
so
could
be
idiom
110.
Why
can
upgrade
to
or
install
a
new
110,
X
or
110
anything
cluster?
Does
that
make
sense.
A
Cli
parameter
should
not
be
deprecated
there's
we
have
to
go
through
the
formal
process
and
policy
for
doing
that,
and
even
Peter
as
we've
been
going
back
and
forth,
and
a
couple
of
changes
that
he's
proposed
for
that
we've
long
had
for
deprecating
or
moving
command-line
arts.
So
we
need
to
follow
the
formal
policy
there
and
that's
actually
documented
we're
in
the
community
repos.
F
F
Then
the
other
thing
this
is
being
able
to
debug
when
we're
using
a
config
file.
There
was
you
can't
provide
v4
verbosity
when
you
provide
a
config
file
and
we're
trying
to
debug
issues
can
be
a
pain.
D
A
Should
be
a
separate
PR,
that's
not
yeah
yeah,
for
for
PRS
against
upstream,
has
made
all
conflation
1
PR
for
atomic
feature
right,
never
for
upstream,
never
conflate
PRS,
just
because
it's
it's,
it
becomes
very
thorny,
so
sure,
but
that
that
is
how
they
explore.
It
is,
for
the
time
being,
we're
gonna
have
to
yeah
yeah.
B
We
would
totally
welcome
ownership
of
any
of
those
changes
to
whatever
level
you're
looking
to
help
out
and
if
you
need
help
ramping
up
on,
you
know
any
of
compilation.
For
that,
and
just
let
us
know
Thanks
are
you.
Do
you
know
where
the
so
a
lot
of
the
cuvette
ATM
development
stuff
happens
in
the
sig
cluster
lifecycle
slack
channel,
if
you're
not
already
in
there.