►
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).
B
B
D
C
E
E
C
It
allows
us
it
allows
us
so
there's
the
case.
You
will
always
run
the
case
yet
right
at
least
I'm
guessing
you'll
always
run
the
case
here.
It's
actually
up
to
you,
but
most
of
us
will
always
run
the
KCM
and
then
we'll
add
the
CCM
as
well
as
a
separate
binary,
and
so
the
case
cm
is
responsible
for
running
all
controllers
that
don't
need
a
cloud
provider
implementation.
C
B
Big
yeah,
so
I
think
what
a
good
thing
was,
because
we
kind
of
like
once
this
didn't
make
the
116
kept
deadlines.
Kind
of
put
this
in
the
back
burner,
so
I
think
a
good
thing
to
do
would
be
to
go
through
what
we
have
in
the
cap
and
see
what
details
were
missing
and
if
what
we
have
in
the
cap
is
still
what
we
want,
yeah.
B
B
So
you
would
say,
like
in
117,
I
want
like
service
and
ROC
controller,
to
run
in
the
cloud
control
manager.
Sorry,
you
would
say
in
117,
I
want
service
for
our
controller
to
run
in
the
KCM,
the
controller
manager,
but
then
in
118
I
want
the
service
rep
controller
running
in
the
cloud
controller
manager
right,
and
so,
as
you
flip
the
version
and
all
the
components.
The
components
just
know
how
to
coordinate
between
the
two
by
switching
where
the
controller
runs
and
so
I
think.
B
The
things
we
agreed
on
is
like
the
the
lock
that
you,
so
the
lock
is
specified
on
a
per
provider
basis.
So
the
library
is,
we
have
a
library,
but
then
the
provider
would
specify
in
some
go
file
or
something
that
says
you
know
like
there
would
be,
maybe,
like
maybe
there'd,
be
a
file
only
for
AWS,
and
then
you
can
decide
as
either.
C
If
I
can
interrupt
briefly,
this
is
really
important
for
two
reasons.
The
first
is,
it
may
be
true
that
amazon
wants
to
migrate,
the
contribute
to
their
kate
into
the
CCM
and
what
going
from
120
to
121,
but
asher
wants
to
migrate.
That's
do
that
same
migration,
but
they
want
to
do
it
as
they
go
from
1:21
to
1:22.
C
So
the
first
thing
you
get
is
the
ability
to
determine
when
you
want
to
do
the
migration
depending
on
your
local
realities.
The
other
thing
is
that
we
know
there
are
certain
controllers,
like
the
node
type
M
controller,
where
Google
runs
its
own
custom
version
of
the
node
IBM
controller
in
the
KCM
when
I
say
that,
basically,
what
I'm
saying
is
our
version
of
the
node
I
Penton
controller
uses
the
cloud
provider
and
no
one
else's
does
so.
C
A
B
I
know
no
worries,
I'm
glad
you
stopped
to
clarify
there.
One
thing
to
clarify,
though,
is
that
each
lock
or
each
yeah,
each
lock,
you
specify,
is
a
new
leader
election
lock
resource,
which
you
know
obviously
takes
performance
out
of
the
API
server
and
whatnot,
and
so
obviously
you
don't
want
to
do
like
one
lakh
per
controller,
which
adds
up
so,
if
you're
doing
like
the
full
migration
over
time,
you're
gonna
end
up
adding
five
extra
watches
four
locks,
so
you
want
to
do
it
within
like
two
or
three
locks
max
right,
yeah.
B
C
B
C
B
B
Well,
the
properties
of
the
migration
lock
are,
must
have
a
unique
name.
Set
up
controllers
in
the
lock
is
immutable.
So
if
you
do
a
release
specifying
a
lock,
you
can't
change
what
controllers
and
a
lot
in
that
version.
No
two
locks.
No
two
migration
Lawson
have
over
lock
controllers
the
controller
manager
where
the
lock
runs
can
change
across
versions.
B
So
what
this
means
is
when
you
specify
lock-
and
you
say
like
on
117
I'm
doing
these
and
I'm
doing
these
controllers
and
in
the
cue
controller
manager,
like
the
only
thing
you're
allowed
to
change
about
that,
lock,
is
what
component
you're
switching
to
and
ideally
you
don't
have
to
you're,
not
gonna
switch
back
to
the
controller
manager
right
and
then
for
minor
release.
It
should
should
run
exclusively
either
in
the
case
yet
or
the
CCM.
C
C
Sorry,
it's
a
subtlety
so
that
last
point
for
a
minor
release.
It
should
run
exclusively
in
one
type
of
controller
manager
which
I
completely
agree
with,
but
I
think
it's
also
worth
mentioning
that
all
versions
of
that
minor
released,
which
is
a
lock
right
and
the
reason
that
that's
true
is
then
my
other
point,
which
is
that
this
migration
mechanism
should
work
for
both
upgrades
and
downgrades.
C
C
F
F
F
C
A
Want
to
make
sure
I
understand
the
point
about
the
multiple
migration
locks,
so
you
have
the
set
of
controllers
that
are
staying
in
the
KCM,
those
used,
the
KCM
lock.
You
have
a
set
of
controllers
that
are
state
that
are
potentially
in
the
CCM,
those
whose
CCM
lock
and
the
migration
lock
only
applies
to
the
ones
that
are
transitioning.
So
what
is
the
case
where
you
need
to
migration
locks
again?
Can
you
just
explain
that
I
don't
think
we.
C
So
there
are
a
couple:
let
me
take
a
real-world
example
that
Google
has
I,
run
and
and
IBM
I
know
does
as
well.
I
run
for
controller
managers
today,
okay,
so
one
of
the
things
I
want
to
do
is
not
only
I
want
to
move.
I
want
to
migrate
things
from
the
KCM
to
the
CCM
and
make
sure
that
works,
and
then
I
want
to
start
pulling
stuff
into
the
CCM.
From
some
of
my
other
controllers.
C
C
Okay,
there
are
some,
let
let
us
say
there
was
some
root
level
controllers
that
run
in
the
KC
app
right,
for
instance,
there
is,
there
is
a
controller
whose
job
is
to
determine
what
rights
you
know,
what
rights
and
privileges
a
controller
hats
I
do
not
believe
that
that
should
be
running
in
the
same
controller
manager.
That's
all
all
of
the
lesser
privileged
controllers,
because
if
you
can
somehow
gain
access
through
some
sort
of
privilege,
escalation
to
that
controller,
you
can
then
give
yourself
any
rights.
C
So
that's
really
more
what
this
is
about,
but
also
the
other
use
cases.
I
mentioned
that
we
do
special
things
for
node
IPAM
controller,
so
that
we
have
finer
grained
control
of
what
IP
spaces
are.
Our
customer
nodes
will
have.
I
am
entirely
willing
to
believe
that
Amazon
is
going
to
want
the
same
food
feature
at
some
point
in
the
future.
C
A
C
A
C
I,
let
me
let
me
finish
and
I
think
we're
saying
the
same
thing,
but
I'm
gonna
be
since
it's
being
recorded
and
other
people
may
come.
I
want
to
be
very
clear
here:
one
KCM
lock.
We
need
one
CCM
lock
and
then
we
need
a
lock
for
every
set
of
controllers
that
are
being
migrated
between
versions
a
and
a
plus
one.
A
F
A
C
B
That
would
be
based
on
what,
like
the
the
leader
election
lock
that
use
you
as
a
cloud
provider
specify
so
like
if
you,
if
the
first
lock
on
one
time,
team,
you're
saying
that
I
want
like
all
the
node
controllers,
then
that's
the
order
that
you're
gonna
migrate
but
I
think
I.
Think
80%
of
the
providers
are
just
gonna.
B
Do
like
I'm
gonna,
do
one
lock
that
migrates
all
of
them
and
like
that's
like
that's
the
end
of
the
story
for
that,
but
we
wanted
to
leave
the
option
open
in
case
either
provider
like
pretty
much
everything
Walter
said,
but
but
also
adding
the
case
or
a
provider
might
think
migrating.
All
the
controllers
at
once
is
too
risky,
so
they
want
to
do
like
one
controller
or
two
controllers
at
a
time,
so
they
can
kind
of
vet
out
the
process
and
and
have
it
test
it
without
kind
of
blowing
everything
up,
yeah.
C
E
C
E
B
I
mean
like
one
other
thing
that
could
be.
A
concern
is
like,
if
you
have
different
versions
of
the
controllers
across
the
two
binaries,
and
that
could
cause
some
issues.
But
that's
I
kind
of
went
back
to
yesterday's
combo
like
that's.
Why
we
have
the
legacy
provider
proposed
so
that
whatever
you
build
into
your
CCM
is
the
same
provider
or
if
you
deviate
at
least.
You
should
be
aware
that
you
know
your
if
you're
gonna
change
the
controller
version
that
you're
they're
compatible
between
you
two
so.
C
So
I
agree
with
that,
but
I'm
gonna
make
the
odd
argument
that
that
is
no
different
than
if
the
controller
didn't
move
like.
If
I
move
from
one
16
to
one
17
and
the
controller
changes,
I
mean
the
fact
that
it
migrated,
isn't
and
changed.
You
know
or
the
fact
that
it
changed
and
migrated
it's
the
fact
that
it
changed
and
whether
it
migrated
or
not
is
is
in
the
sort.
A
C
C
C
F
B
C
B
So
it
so
it
wouldn't
be.
Any
config
file
like
so
essentially
what's
happening
is
like
the
the
config
for
the
leader.
Election
happens
inside
of
both
acacia
min
the
CCM,
and
the
providers
are
building
it
into
both
binaries
and
so
like
the
migration
is
forced,
depending
on
the
changes
you
make
into
the
binaries
on
that
version.
C
Question
on
that
Andrew,
it's
easy
enough
for
for
any
cloud,
any
cloud
provider
to
run
a
build,
their
own
custom,
CCM
I'm
a
little
concerned
about
people
having
to
build
their
own
custom
KCM.
Are
we
sure
this
isn't
something
that
would
be
better
done
as
like
config
baby
at
a
yeah
mole
or
whatever?
What.
C
So
if
we
do
it
as
a
dot
go
file
right
then
I
actually
have
been.
Then
whoever's.
Shipping
I
mean
it
for
me,
it's
no
big
deal.
We
do
that
anyway,
but
I
would
imagine
there
are
a
lot
of
cloud
providers
out
there
who
just
take
the
KCM
binary
that
is
built
by
you
know
the
the
the
really
oh,
the
CN
CF
release
team
and
just
used
the
KCM
binary
as
is,
and
so,
if
we're
gonna
say,
oh
you've
got
to
add
this
go
file
and
then
recompile,
but.
B
It's
only
applicable
to
their
entry
cloud
providers
right
so
like
it
like
we're.
Only
adding
migration
locks
in
the
case,
like
the
migration
lock
for
case
name,
is
only
enabled
if
you're
already
in
an
entry
cloud
provider
the
cloud
provider
that
needs
to
be
migrated
out
of
tree
right.
C
C
B
F
C
B
C
C
Yeah
and
I
suddenly
discover
that
my
customers,
one
to
be
able
or
admitting
all
my
customers
I,
want
to
be
able
to
allow
some
central
Amazon
system
to
control
BDD
IP
addresses
of
the
cluster
mm-hmm
and
so
I'm
gonna
switch
the
the
NotI
Pam
controller
to
start
making
cloud
provider
calls
which
means
I
have
to
migrate.
The
no
Taipan
controller
out
of
the
case
Yemen
into
the
CCF.
B
B
If
you
at
that
point,
where,
like
nothing
in
the
case,
is
making
the
cloud
calls,
but
you
have
stuff
in
the
KCC
and
then
the
solution
would
be
just
actually
just
add
it
directly
in
the
CCM
and
in
the
same
version
you
add
in
the
caesium.
You
didn't
turn
it
off
in
the
case.
Yet.
C
Well,
just
to
be
clear,
I
mean
I'm
looking
fairly
far
down
the
road
and
maybe
just
not
a
problem.
We
want
to
worry
about
now,
but
the
idea
that
and
there's
a
lot
of
nice
things
to
it.
The
idea
that
that
Andrews
pushing
is
the
B
we
need
to
configure
the
locks
and
we
need
to
configure
the
locks
on
a
per
cloud
provider
basis
and
since
we're
doing
your
basic,
oh
yeah.
C
B
C
F
B
F
B
The
like,
somewhere
in
CMD,
/keep
controller
manager,
like
you're,
still
checking
you're,
calling
that
library
and
you're
checking
your
component
name
and
what
version
you
are
and
then
you
decide.
And
then
you
pull
in
from
that.
Like
what
controllers
that
we
classify
as
cloud
controllers
and
we're
out
of
whatever
lock
you
want.
B
C
Yeah
I
mean
I
think
my
biggest
concern.
That
is,
we
can
say
that
it's
not
the
cloud
provider,
implementation
or
interface,
but
for
all
intents
and
purposes
we're
just
creating
a
second
cloud
provider,
interface
or
implementation.
That
is
specifically
for
migration
right
and
then
we're
gonna
come
video.
Okay,
who's
allowed
to
make
changes
like
is
VMware.
Its
Rancher
is
at
all
allowed
to
make
changes
in
there.
F
C
Fact,
I
kind
of
what
I'm
gonna
play
off
of
what
you
see
in
just
that.
In
fact,
if
we
do
this,
we
could
even
just
at
some
point
build
a
implementation
of
that
interface,
which
is
the
thing
that
reads
the
config
file
and
returns
based
on
that.
So
I
don't
think
we're,
even
particularly
closing
ourselves
out
so
I
mean
we
can
start
here,
and
if
we
have
time
we
can
even
we
can
even
build
the
config
file
version.
So.
A
B
B
F
C
C
B
A
A
B
A
B
C
The
other
thing
I
would
like
on
that
one
I
mean
let's
talk,
goal
and
and
fall
back
I
think
the
goal
would
be
to
get
something
approved
for
alpha,
but
I
had
a
minimum
I
want
to
make
sure
unless
it's
already
been
done,
I
want
to
make
sure
that
we
at
least
have
the
cap
proposal.
Where
cap
proposal
means
that
we
have
an
agreed-upon
problem
that
needs
to
be
fixed.
C
C
Cool
and
if
you
need,
if
you
need
help
reaching
out
to
the
node
team
on
that
one,
please
let
me
know
I'm
actually
dealing
with
a
node
team,
quite
a
bit
right
now
and
may
be
able
to
get
a
little
bit
of
traction.
Getting
a
node
review
and
I'm
guessing
right
now
give
them
certain
things
that
you're
gonna
have
a
hard
time
get
anyone
know
anyone
else
to
get
you
that.