►
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
A
So
I
created
a
package
called
migration
and
component
base,
and
it
pretty
much
has
a
type
called
migrated
controllers,
and
in
this
type
you
specify
a
unique
name.
The
list
of
controllers
that
you
are
migrating,
the
component
name,
which
contains
like
the
binary
name
of
the
controller
manager
you're
running
in
and
then
it
has
leader
election
fix.
A
A
A
What
I
added
was
a
migration
config
method
to
the
main
interface
and
so
implement
this
method,
and
if
it
returns
a
list
of
migrating
controllers,
which
is
destruct
that
we
just
looked
at
and
then
pretty
much
the
both
the
cloud
controller
manager
and
the
cue
controller
manager
will
start
with
the
Cloud
Control
Manager.
It
calls
that
methods
from
the
cloud
provider
here
right
and
then,
if
it's
enabled
it
goes
through
each
migration
config.
It
checks
the
existing
list
of
controllers
and
then
it
asks
if
it
if
it
had.
A
If
so,
it
goes
through
all
the
controllers,
if
it
sees
that
the
migrated
controller
type
owns
that
controller,
and
if
the
component
name
matches
the
current
component
name,
it
then
append
it
to
a
list
of
migrating
controllers
and
then
later
it.
It
removes
that
controller
from
the
primary
list
of
controllers.
And
then
here
someone
asks
something
yeah.
So
here
it
it
runs
the
main
controller
controllers
from
the
primary
controllers
list
and
then,
if
it
sees
that
the
migration
is
enabled
it
goes
through
each
config
again
and
then
it'll
start.
A
C
A
Good
to
know
yeah
I,
don't
think
the
types
or
the
interface
names
or
anything.
This
is
final,
very
open
to
like
I,
literally
just
slapped
us
together
yesterday,
I'm
open
to
changing
those
I
think
like
what
I
want
consensus
on
is
like
the
general
approach
of
like
where
were
we're,
we're
allowing
the
migration
control
the
migration
config
to
be
in
to
be
satisfied
and
then,
where
we're
calling
it
and
how
we're
wiring
it
together
to
form
like
how
we're
doing
the
leader
election
from
each
of
the
binaries.
A
So
for
AWS
you
would
just
specify
this
is
just
one
way
to
do
it,
but
you
can
specify
variable
here
that
says
you
know
this
this
this
particular
one
is
all
controllers,
so
we
put
in
the
list
of
all
controllers,
we
know
or
the
cloud
specific
ones.
We
may
want
to
have
a
constant
or
something
for
this.
So
so
you
don't
accidentally
put
in
the
wrong
controller
name,
but
this
is
the
wrong
idea
and
then
you'd
say
I
want
this
controller
to
run
and
controller
manageress.
So
this
is.
A
A
Did
you
see
one
or
the
azure
one
here?
You
know
ashramite
wanna
just
migrate,
only
the
network
controllers
and
one
release,
and
so
you
would
do
that
and
then
in
another
release
you
would
create
a
new
variable.
That's
you
know
as
you're
migrating
node
controllers
and
you
do
the
same
thing
for
the
GC
curse.
A
You
might
want
to
do
you
can
specify
both
together,
so
you
might
want
one.
That's
the
network
controllers,
another!
That's
just
the
node
controllers.
There
should
be
no
controllers
by
the
way,
and
so,
like
allowing
the
provider
to
say
like
what
versions.
A
set
of
controllers
should
be
migrating
and
then
what
component
they
should
be
running
on.
A
That's
the
best
pretty
much
the
PR
I
feel
like
so
I
feel
like
the
logic
here,
it's
a
little
hard
to
follow
it's
kind
of
a
big
blob
of
for
loops
and
if
we
can
simplify
this
but
I
feel
like
this
is
a
good
way.
This
is
like
the
logic
and
its
most
simplest
form
of
what
we
would
be
doing
with
this.
With
this
change,
and
so
yeah
questions
concerns
comments.
A
No
reason
like
in
the
slack
Channel
I
put
in
that
I
didn't
do
that,
just
because
it
was
hard
to
prototype
it,
but
like
pretty
much
like
this
call
here
can
read
that,
like
a
ya
know,
file
or
whatever
I,
don't
care
too
much.
How
that's
being
read,
I
well,
I.
Do
care
it'll,
like
I,
think
either
way
works
as
long
as
we
are.
We
are
in
agreeance
with
this
logic
here
and
this
logic
here.
B
A
Right
so
like,
if
we
did,
this
look,
I
think
this
is
fine
as
long
as
there's
a
feature
gate
right,
and
so,
if
the
feature
gets
off,
then
you
run
all
your
controllers
everywhere
in
the
case.
And
then,
if
you,
if
your
feature
gate,
is
on
you're,
assuming
that
they,
the
user,
understands
that
they
need
to
run
the
cloud
controller
manager
somewhere
in
the
cluster.
D
But
it's
still
is
there
anything
that,
like
hard
codes,
what
like
so,
would
they
still
be
able
to
just
enable
that
feature
gate
in
the
next
version?
Is
that
how
they
would
control
it
if
they
want
to
take
more
time,
for
example,
with
their
migration?
Do
they
just
keep
that
feature
gate
disabled
and
then
enable
it
whenever
they
want
to.
A
So
I
think
what
would
happen
is
if
we're
going
with
this
design,
which
is
not
I,
know
right,
but
just
for
as
a
thought
exercise.
If
we
went
with
this
and
then
we
added
a
feature
gate,
what
it
would
mean
is
you,
as
the
implementer
of
the
provider,
define
like
what
version
the
migration
happens
and
the
user
is
responsible
for
enabling
the
future
gate
on
that
version
switch.
D
A
A
B
Yeah
I
think
that's
my
biggest
objection
and
in
fact,
if
we
put
in
one
layer
of
indirection
I,
don't
even
care
if
we
do
pull
it
from
as
long
as
it's
not
obvious
where
you
have
in
here,
which
is
in
the
controller
manager
if
the
controller
manager
is
sort
of
given
so
instead
of
it
pulling.
If
someone
pushes
they
to
here,
then
I
don't
care.
If
it's
coming
from
a
file
or
coming
from
the
cloud
provider.
And
then
we
then
it
it's.
B
A
B
Think
that's
fair.
My
thought
was
that
if
we
provide
them
with
the
the
default
Yama
file
and
for
the
cloud
providers,
the
cloud
providers
are
just
going
to
produce
their
own
when
they
want
to
do
the
migration
and
then,
if
we
just
produce
a
Yama
file
and
make
it
part
of
the
standard
images
that
we
know
of-
and
you
know,
push
out
the
message:
hey
you
need
to
do
it.
Then
it's
really
no
different
than
doing
it
through
the
through.
Through
the
cloud
provider
interface
right
I
mean
it's
the
same.
B
B
We
provide
them
with
hey
here's,
here's
here's
the
suggested,
llamo
file,
you
can
just
pick
up
and
use
it
or
you
can
override
it
and
I
think
that's
a
level
of
work.
They
would
have
to
do
even
with
the
solution
you
have
here,
because
they're
gonna
have
to
put
that
code.
Put
those
code
changes
into
their
cloud
provider,
implementation.
A
A
B
We
could
in
fact
say
that
there
is
a
default
migration
implementation
and
anything
you
do
is
overriding
the
migration,
the
default
migration
implementation
and
the
default
mating
migration
implementation,
let's
say
just
says
you
will
be
migrating.
These
three
controllers
from
the
KCM
to
the
CCM
across
the
120
to
121
relates,
but.
B
The
default
behavior
to
do
the
migration
here
and
you
can
provide
if,
if
the
file
isn't
the
way
you
want
to
do
it,
you
can
either
have
to
use
the
override
that
says:
I
want
to
pick
it
up
from
file
or
you
can
implement
your
own
migration
implementation.
That
says
where
you
want
it
and
how
you
want
to
do
it
right.
D
Think
we
need
to
figure
out
what
we're
okay
with
in
the
user
story
before
we
can
be
certain
about
what
we're,
okay
with
in
the
actual
interface
on
the
code,
because
if
we're
okay
with
users,
you
know
forcing
users
to
upgrade
over
one
version
which
I
have
doubts
about
about
my
mother
and
and
what
we
can
do
in
the
code.
But
if
it's
gonna
have
to
be
a
longer
window,
you
know,
if
there's
I
don't
know,
does
that
make
sense
or.
B
Yeah
now
makes
sense
to
me,
I
mean
might,
but
my
thought
is
I'm
hearing
two
things
and
I'm
trying
to
find
the
comfortable,
in-between
I.
Think
there
is
this
story
of
the
right
thing
should
happen.
If
I
do
nothing
and
that's
for
me
where
defaults
fall
in
and
I,
don't
want
to
force
people
to
a
particular
release
and
again
I
think
that's
where
I'm
hoping
that's
where
the
default
were,
you
you
can
say,
look
if
you
do
naturally
nothing!
B
Yes,
you're
gonna
get
migrated
from
this
release
to
this
release,
but
there
is
a
well-defined,
fairly
easy
mechanism
for
you
to
determine
where
and
what
you're
going
to
migrate
right,
and
so
it's
more
difficult
than
to
do
nothing
option
because
you're
not
doing
nothing,
but
it's
sort
of
I'm
hoping
it
provides
that
nice,
where
you
know
the
this
standard,
behavior
is
easy
or
free
and
the
the
exceptional
is
all
work
and
aren't
too
difficult.
So.
A
Here's
what
I
keep
here's,
what
I
think
it's
gonna
happen
if
you
do
that,
we're
gonna
they're
gonna,
we're
gonna
reach
that
version
where
we,
if
users
didn't
do
anything,
they're
gonna,
get
migrated.
They're
gonna,
get
migrated
without
the
CCM
and
everything's
gonna
break
and
then
they're
gonna,
look
up
docs
for
that
and
they're
gonna
find
oh,
do
this
thing
to
disable
the
migration
and
everyone's
gonna
disable
the
migration
and
then
we're
stuck
in
a
situation
where
everyone
has
disabled
migration.
Everyone
is
still
intrigued
and
then
we're
now
trying
to
get
people
onto
the
CCM.
B
B
This
code
is
going
away
in
122
121
we're
going
to
migrant
you
from
KCM
to
CCM
and
that's
what
should
happen
and
by
the
way
you
can
delay.
But
if
you
try
and
delay
past
122,
it's
not
going
to
work
and
there
is
no
amount
of
not
migrating,
which
is
going
to
fix
the
fact
that
you're
not
running
a
CCM
and
the
you
know
the
relevant
controllers
no
longer
work
in
the
case
gim
because
they're
not
even
built
into
the
case
yeah
yeah.
C
E
A
D
B
My
thought
is,
we
have,
we
could
even
have
layered
defaults,
so
my
thought
is
the
basic.
The
baby
slept
default
is
the
default
that
would
be
picked
for
something
where
the
cloud
provide
here
has
done
nothing
to
help
you
right,
and
so
we
just
pick
a
release,
saying
120
to
121
we're
gonna
move
you
across
and
hopefully
that
all
works.
Now
it
is
possible
and
that's
the
base
default.
It
is
also
then
possible.
B
If
you
look
at
the
way
the
cloud
providers
inject
themselves
to
have
the
cloud
provider
come
in
and
say:
I
would
like
to
go.
You
know,
even
if
it's
not
your
what's
a
KS
or
eks,
you
can
say
like
just
for
the
default
open
source
Amazon
whatever,
when
the
Amazon
Cloud
provider
in
it
runs
it'll
inject
a
separate
default
which
will
provide
Amazon,
specific
defaults,
and
then,
on
top
of
that,
you
could
always
then
say
by
the
way.
B
D
Mean
I
guess
the
the
different
users
that
I'm
thinking
about
so
there's
like
the
the
user
that
did
everything
by
hand
and
maybe
used
cube
ATM
or
we
know
whatever
they.
They
have
a
lot
of
control
of
their
cluster.
Then
there's
like
the
you
know
for
AWS.
We
have
like
cops
users
who
are
using
some
kind
of
a
provisioning
system,
that's
not
a
kiss
and
then
there's
yes
and
so
I
see
like
cops
and
eks.
D
You
know
both
the
the
framework
provides
the
config
and
chooses
the
version
and
all
the
user
does
is
say:
I
want
to
upgrade
to
this
version
and
then
the
framework
does
the
work
and
picks
the
version
of
the
cloud
provider.
You
know
it's
TCM
that
is
going
to
run
and
creates
the
migration
configure
whatever,
but
for
the
the
user.
That's
doing
everything.
Themself
I'm
just
wondering
like
what
are
all
the
things
that
they
have
to
do
so
in
this
case
they
would
just
create
a
override
migration.
B
So
my
hope-
and
obviously
things
have
to
be
done
to
make
this
work-
is
that
lets
say
it's
120
to
121.
Someone
has
gone
in
and
made
that
CCM
the
default
in
121.
We
know
that
and
so
we
provide
a
default
migration
that
says:
okay,
CCM
is
now
the
default
in
121
and
if
you
upgrade
from
120
to
121
the
migration
locks
in
120,
say
KCM
should
should
get
a
migration
lock
in
121
and
CCM
should
grab
the
migration
lock.
&Amp;
KCM
shouldn't
run
those
and
from
there
the
right
things
should
happen.
Yeah.
B
Presumably
not,
although
that
may,
we
could
have
that
partly
configured
by
this
migration
lock,
but
even
there
I
mean,
as
I
said
at
some
point,
whether
it's
122
or
123.
Those
controllers
won't
even
be
in
built
into
the
case,
see
em
anymore
right
now.
The
other
thing
is
at
some
point,
probably
that
same
release
the
cloud
provider
isn't
going
to
be
built
into
the
KCM
anymore.
B
Well,
I
think
the
good
news
is:
we've
got
a
channel
slack
channel
for
this,
and
we've
got
a
PR
and
I
think
those
are
both
good
places
to
really
capture
a
lot
of
this
conversation
and
the
goal
is
to
get
something
into
the
cap.
So
we've
got
a
couple
of
months
to
really
iron
out
all
the
details.
So
I
don't
think
any
of
this
is
pressing
well.