►
From YouTube: Kubernetes SIG Cloud Provider 2018-08-08
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
B
B
Provider
methods
for
finding
out
instance,
information
about
nodes
and
if
that
instance,
look-up
reports
that
the
instance
does
not
exist,
the
node
controller
will
delete
the
kubernetes
api
no
table
jecht
because
the
instance
doesn't
exist.
Some
of
the
cloud
providers
will
report.
Stopped
instances
still
exist
so
in
GCE.
If
you
have
an
instance
and
you
stop
it
we'll
say:
oh
yeah,
it
still
exists,
don't
delete
it
and
then
later
it
can
start
back
up
and
you
know
start
reporting
healthy
again
and
get
more
clothes.
Kids
will
do
it.
B
Other
cloud
providers
like
AWS
and
OpenStack
say
stopped
instance
does
not
exist.
Which
means
is
that
any
information
on
the
node
object
gets
lost,
so
labels
paints
any
annotations
anything
else
that
has
been
said
on
that
object
gets
lost.
So
the
question
is,
which
is
the
correct
behavior
which
do
we
want
to
converge
to,
and
so
we
we
talked
about
this
last
week
in
cig
AWS,
but
I
thought
we
talked
about
it
here
as
well.
B
Just
since
this
is
kind
of
a
cross
cloud
question,
the
two
arguments
for
and
against
I
guess
deleting
and
not
deleting
deletion
of
the
node
object
is
losing
data,
and
we've
actually
seen
this
in
our
setups,
where
you
know
it's
got
labeled
and
we're
distributing
workloads,
and
then
the
deletion
lost
that
information.
So
when
the
node
started
back
up
andrey
registered
itself,
it
was
no
longer
represented
in
the
correct
apology.
B
B
Shutting
down
and
then
starting
back
up,
it
could
be
shut
down
in
order
to
have
significant
changes
made
to
it
and
when
it
starts
back
up
it
Brazil
doesn't
bear
much
resemblance
to
the
the
note
it
used
to
be,
and
so
deleting
it
kind
of
wipes
the
slate
clean
and
cleans
out
any.
You
know:
cruft
resources
or
random
things
that
are
associated
with
it,
that
you
don't
want
to
be
reflected
when
it
starts
back
up.
I
actually
think
that's
racy.
B
If,
if
you're
counting
on
the
note
controller
to
delete
the
note
object,
while
your
instances
shut
down
so
that
when
it
starts
back
up,
it
has
a
clean
slate.
That
assumes
that
the
node
controller
is
always
running
and
happens
to
hole.
Your
instance
state,
while
it
is
shut
down
so
I,
think,
depending
on
that
for
a
clean
slate
is,
is
racy,
and
if
we
have
problems,
we
need
to
fix
those
to
report
correctly
when
the
cubelet
comes
back
up.
So
I
I,
don't
think.
B
That's
a
strong
reason
to
continue
deleting
this
anyway
I
linked
to
kind
of
the
main
issue
and
then
there's
a
bunch
of
other
issues
and
then
the
last
link
is
to
a
pull
request
that
fixes
the
AWS
implementation
and
I
think
Angus
had
who's.
One
of
the
OpenStack
folks
had
a
similar
change
right.
He
was
like
a
one-line
change
to
remove
the
filter,
but
that
that
is
the
issue
and
I'll
open
up
the
floor
for
other
people
to
weigh
in
perspectives,
opinions,
other
yeah.
A
B
A
couple
things
there,
the
instance
exists
by
ID-
is
only
used
by
the
cloud
node
controller,
not
by
the
node
lifecycle
controller.
Okay,
even
within
the
controller
managers,
there's
inconsistency
about
which
which
methods
are
called
that
that'll,
probably
wash
out
once
we
collapse
onto
one
of
those
but
I,
think
it's
important
to
define
what
the
API
means,
because
if
we
want
the
kubernetes
controllers
to
respond
coherently
across
providers,
yeah
I
think
it's
important
for
it
to
be
consistent.
C
A
A
A
E
C
B
D
Also
had
a
conversation
about
trying
to
converge
on
a
formal
state
machine
across
cloud
providers.
At
one
point
we
started
that
conversation
about
whether
suspended
is
a
real
state
or
not,
and
whether
we
have
just
properties
of
nodes
or
whether
there
is
a
formal
state
machine
but
I,
don't
remember
where
we
ended
up
on
that
anybody,
oh
I,
think.
B
But
for
very
providers
but
just.
F
B
B
B
The
OpenStack
maintainer
czar
in
favor
of
converging
to
but
GCE
and
others
do,
which
is
leave
these
in
place.
The
AWS,
maintainer
'z
were
cautiously
in
favor,
but
weren't
sure
what
the
implications
could
be
for
things
that
were
depending
on
the
queen
state.
If
you
know
of
implications,
please
them
to
the
issues,
so
we
can
be
aware
of
them,
but
I
would
like
to
recommend
we
proceed
with
converging
on
not
deleting.
A
F
I
wanted
to
get
the
group
status
update
on
different
fronts
that
we've
been
making
progress.
First
of
all,
they
kept
proposals
out.
They
are
all
the
suggestions
and
comments
have
been
incorporated
to
it.
We
didn't
set
a
deadline.
You
know
for
this
group
to
review
it,
but
we
are
starting
to
think
about
that.
So
especially
the
representatives
of
the
other
providers
would
be
great.
If
you
haven't
done,
take
a
look,
it's
linking
the
agenda
and.
C
In
fact,
so
we
folded
in
Andrews
feedback,
Nick
from
Amazon
or
AWS,
was
nice
enough
to
actually
send
a
PR
on
top
of
my
PR,
which
we
folded
in
with
all
of
Amazon's
feedback,
and
in
fact
both
we
and
AWS
would
really
like
to
get
this
finalized.
This
week,
I'm
Nick
specifically
asked
me
to
bring
that
up
in
the
meeting
for
him.
A
F
F
F
So
you
know
this
is
all
leading,
hopefully
when
we
can
sort
out
all
these
dependencies.
Some
points
of
conflict
to
appear
to
move
all
the
code
out,
but
working
on
that
it's
not
easy.
I
have
some
updates
on
other
6cg
storage
is
also
doing
progress
on
the
CSI
interface,
but
they
are
kind
of
facing
similar
issues
that
we
are
facing.
You
know
dependencies
mostly
last
I
heard
they
were
stuck
on
features
so,
but
they
are
not
clearly
working
on
this
and
the
other
sig
is
signaled.
F
G
C
F
A
A
It's
just
pending
review
from
someone
on
cig
network
I,
think
Tim
Hawking
says
she's
gonna!
Do
it,
but
pretty
much
it'd
be
nice.
If
every
provider
can
give
a
thumbs
up
on
this,
because
it
is
gonna
change
the
way
we
pass
along
like
the
provider
name,
but
the
little
bouncer
names
for
for
each
of
the
Flyers
so
pretty
much.
G
Go
ahead,
thank
you
yeah.
So
the
next
two
agenda
items
are
are
related.
They
cover
a
cloud
provider
and
other
documentation.
I
I
want
to
start
by
introducing
Cody
Clark,
who
I
I
guess
was
it
used
to
be
an
intern
for
Google
and
now
is
the
full-time,
employee,
Google,
server
and
say
hi
to
Cody,
and
there
are.
There
are
a
couple
of
issues.
G
Earlier
today
we
had
a
meeting
with
cig
Doc's
about
the
current
look
at
the
current
status
of
some
upstream
documentation
and
and
in
plans
for
it,
and
so
the
first
thing
that
we
wanted
to
flag
for
you
was
updating
this
picking
the
right
solution
document.
If
you
go
to
kubernetes
io
dot,
you
know
slash
Docs
and
you
go
to
setup,
and
actually
the
very
first
thing
get
send
you
to
is.
G
You
know,
or
something
in
between
and
they're,
looking
at
cleaning
up
these
documents
and,
of
course
this
is
going
to
have
an
impact
on
on,
say
cloud
provider,
because
many
of
you
are
are
either
providing
services
or
you
know
you
know
either
either
all-in-one
solutions
or
you
know
all
both
solutions
on
your
clouds.
That
will
then
will
be
impacted
in
here,
for
this
page
in
particular,
and
I
think
Cody
can
maybe
expand
on
it.
A
little
bit
more
is
for
a
number
of
these
solutions.
G
I
G
Yeah-
and
this
is
and
and
and
and
one
thing
that
you
know
we
want
to
make
sure-
is
that
you
we
want
to
make
sure
that
the
cig
cloud
provider
has
input
into
this,
and
so
you
should
consider
cody
as
your
point
of
contact
within
cig
docs.
If
you
have
ideas
or
you
eat,
there
are
things
that
that
you
see.
You
can
also
reach
out
to
me
since
I've
been
kind
of
working
on
the
docs
efforts.
G
You
know,
as
liaison
also,
but
you
know
one
of
the
things
that
there
that
were
asking
we're
actually
asking
for
is
volunteers
from
within
the
community
and
particularly
the
cloud
provider
community.
Who
will
be
able
to
take
a
look
at
some
of
the
pull
requests
that
Cody
is
going
to
be
sending
up
and
making
comments
about
how
it's
going
to
impact
them
so
that
we're
getting
you
know
a
larger
community
feedback
from
the
people
who
were
actually
going
to
be
impacted
by
this.
Oh.
I
G
D
G
Yeah-
and
that
was
one
of
the
things
that
we
talked
about
too,
was
you
know
a
page
like
this
is
going
to
serve
a
number
of
different
users,
they're
going
to
be
people
who
are
looking
for
solutions
that
are
they
just
they
want
to
go?
They
want
to
select
the
credit
card
and
be
able
to
have
a
kubernetes,
endpoint
and-
and
you
know,
you
know
and
get
started
to
people
who
are
going
to
be
trying
to
set
up.
You
know
kubernetes
within
their
own
organizations.
G
This
higher
level,
you
know
getting
people
started
down
the
path
that
they
want
to
go
down,
but
then,
as
they
get
into
more
specific
solutions,
you
know
working
with
cloud
provider
to
have
that
documentation,
be
the
responsibility
of
the
maintainer
zuv
that
and
to
link
out
to
it
more,
but
still
provide
a
later
a
layer
of
consistency,
so
that
was
that
was
kind
of
the
way
we
were
talking
about
this
at
a
high
level
in
the
in
the
meeting
we
had
earlier
today.
You
know
in
terms
of
the
organization
that.
D
Sounds
great
thanks
for
the
update,
and
this
seems
like
a
great
effort
to
undertake
it's
a
it's
a
sprawling
mess
right
now.
So,
okay.
G
Similarly,
you
know
the
really.
The
next
step
that
we
need
to
work
out
is
is
providing
solid
good
documentation
on
how
to,
in
general
mode
an
external
cloud
provider.
I
know
that
that's
something
that
we've
been
working
on,
but
if
you
know,
if
we
can
kind
of
set
that
as
one
of
our
goals
for
one
of
the
major
deliverables
that
we
can,
that
we
can
provide
for
state
cloud
provider,
I
think
that
would
be
very
useful.
D
One
conversation
I'm
eager
to
get
started
on
related
to
that
is
where,
where
to
find
the
binaries
that
a
user
needs
to
run
kubernetes
in
various
environments
and
I
think
think
that
conversation
will
likely
take
quite
a
while
to
figure
out,
but
we're
no
longer
in
the
world
where
you
can
just
take
what
is
built
in
the
open-source
communities
project
and
run
it
on
any
one
of
the
places
where
it
will
run.
But
you
need
other
binding
that
are
provided
externally.
D
D
G
D
D
This
is
going
to
become
more
and
more
of
a
challenge,
so
the
conversation
may
start
here
and
bleed
into
other
areas
to
intended
to
displace
other
conversations,
if
they're
going
on
new
and
is
involved
in
other
SIG's
that
are
discussing
this
issue.
Please
loop
me
in
because
I'm
interested
in
thinking
about
it
I
will.
H
D
Yeah
I
know
that
the
ingress
folks
are
also
working
on
a
decoupled
approach
and
now
there's
an
external
add-on
URL
field,
that's
being
used
for
a
couple
of
things
so
add-ons
are
not
always
bundled
into
the
kubernetes
version.
So
there's
versioning
issues
and
coordinated
upgrades
and
it
becomes
we
don't
we
don't
have
no
problems.
We
just
have
different
problems
and
we
should
go
into
their
eyes
open,
so
yeah.
E
It
seems
tempting
to
think
that
maybe
you
could
bundle
all
of
these
different
things
like
storage
networking,
along
with
the
cloud
provider,
but
in
fact
I.
Don't
think
it's
necessarily
that
clean,
because
on
vSphere,
for
example,
you
can
host
and
people
do
host
other
distros
than
vmware
once
and
people
perhaps
need
to
mix
and
match
these
various
components.
A
J
K
Hi
Alton
Ishii
here
sorry
I,
was
having
some
problems
with
my
mic.
I
had
a
couple
of
questions
to
ask
I,
don't
know
if
you
guys
have
the
answer,
but
insig
AWS.
We
are
working
on
a
CSI
driver
for
EBS.
K
Today
we
host
this
in
in
someone
else's
folder
we're
thinking
of
moving
it
to
sake
AWS,
but
I
had
questions
around
the
volume
controller,
which
is
part
of
CCM
today
and
CSI
driver.
So
essentially
the
volume
controller
as
I
understand
it,
creates,
deletes,
attaches
and
detaches
volumes
to
nodes
and
in
the
future
we
are
going
to
add
CSI
driver
and
then
all
the
PDS
and
PVCs
will
be
attached
to
the
driver.
K
D
C
Definitely
talked
to
six
storage:
we've
talked
a
little.
The
six
storage
I
know
in
the
migration
period.
Part
of
the
concept
had
been
that
things
like
GCPD
and
a
AWS
ECB
would
probably
turn
into
wrappers
around
the
CSI
to
allow
backward
compatibility
and
migration
to
be
easy.
But
what
the
long-term
volume
controller
plan
is
yeah.
You
definitely
want
to
talk
to
Saad.
Okay,.
K
C
K
D
I
K
We'll
do
the
other
question
I
had
is
linked
to
what
Iago
said
about
where
you
want
to
store
the
CC
n
binary
I'm,
trying
to
figure
out
how
I
can
provide
Amazon
credits
to
different
SIG's.
So
they
can
utilize
those
credits
for
individual
testing
efforts
and
it
seems
to
be
all
over
the
place
and
you
might
not
have
the
right
answer,
but
for
end-to-end
testing
also
you
there
will
be
a
need
for
infrastructure.
K
H
Can
give
you
some
some
possible
that
there's
a
Jew
cuts
from
from
VMware
is
working
with
suggesting
to
figure
out
a
way
to
standardize
some
of
the
cloud
provider
testing
facilities,
there's
nothing
set
in
stone.
Yet
it's
just
a
discussion
happening
in
suggesting,
but
there's
definitely
something
happening
already.
So
you
know
what
I
reach
out
directly
can
give
you
the
contacts.
It's
not
in
this
call.
Yeah.