►
From YouTube: Kubernetes SIG Storage 20190912
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 12 September 2019
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.4curlgbn0zv5
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Saad Ali (Google)
Chat Log:
09:30:27 From Saad Ali : https://kubernetes-csi.github.io/docs/drivers.html
09:30:48 From Saad Ali : Create a PR at https://github.com/kubernetes-csi/docs/pulls
09:40:14 From Jan : san diego
09:40:19 From Ben Swartzlander : I'm okay with either
09:40:23 From Deep Debroy : San Diego
09:40:33 From Xing Yang : I prefer KubeCon San Diego
09:40:37 From xiangqian : East coast
09:40:40 From Ripul Patel : San Diego
09:40:55 From Ardalan Kangarlou : Either
09:45:25 From Ben Swartzlander : Monday
09:45:29 From Xing Yang : Monday
09:45:46 From Deep Debroy : Monday
09:46:01 From Ripul Patel : Monday
09:46:09 From Ardalan Kangarlou : Monday
None
A
All
right
today
is
September
12,
2009
teen.
This
is
the
meeting
of
the
kubernetes
storage
special
interest
group.
As
a
reminder,
this
meeting
is
recorded
public
and
posted
on
youtube.
Today
we
have
a
lot
of
things
on
the
agenda:
we're
going
to
start
off
by
doing
a
quick
review
of
items
that
the
cig
worked
on
for
q3
for
the
1.16
kubernetes
release
and
we're
gonna
get
an
end
of
quarter
status
on
what
features
made
it.
What
features
didn't
make
it
and
then
next
time
in
two
weeks,
we're
gonna
start
our
q4
planning.
A
So
if
there
are
items
that
you
want,
this
thing
to
work
on
keep
an
eye
out
for
an
email
I'm
going
to
send
once
I
create
a
new
spreadsheet
or
a
new
sheet
inside
the
spreadsheet
for
next
quarter,
and
we
can
now
get
started
reviewing
that
next
time.
So,
let's
jump
into
our
planning
sheet
and
start
going
through
the
items
and
get
an
end
of
quarter
status.
A
A
A
D
A
D
A
D
B
A
A
D
A
E
F
B
D
E
A
A
Okay,
next
item
is
volume
expansion
for
stateful
sets.
This
was
a
sick,
apps
project
that
we
were
helping
Shepherd
specifically
a
month
a
month.
Any
updates
on
this
no
come
on
snot
here
today.
So
last
status
was
Kemp
has
been
lgt.
Emden
approved
needs
a
PR,
reviewer,
relaxing
PVC
validation,
so
I
think
the
status
here
was
started.
A
G
A
A
Next
item
is
CSI
spec,
alpha-beta,
GA
designation.
We
had
a
number
of
items
go
in
to
the
CSI
spec
for
what
is
effectively
going
to
be
CSI
1.2.
None
of
those
items
required
an
alpha-beta
GA
designation
because
they're
not
new
features,
so
we're
going
to
punt
this
to
1.17.
If
there
are
new
features
that
pop
up,
we
will
apply
alpha
beta
G,
alpha
beta
designation
to
those.
A
B
So
the
entry
portion,
which
was
pretty
simple,
got
done
and
we
got
some
good
end-to-end
tests,
Vivat
using
the
GCP
file
store
driver
by
peter
Hanyu,
so
that
piece
was
working,
but
that
was
not
continued
rised.
So
for
the
continuous
piece
the
proxy
is
definitely
in
progress,
but
not
anywhere
near
completion
for
116,
okay,.
A
G
A
E
A
A
F
A
A
Next
item
is
three
items:
CSI
drivers,
one
for
an
F
S
one
for
I
scuzzy,
one
for
fiber
channel
for
NFS
and
I
scuzzy.
The
drivers
already
have
their
own
repo.
The
task
was
to
add
image,
building,
testing,
C,
ICD,
documentation,
etc
and
then
for
fibre
channel.
It
was
to
move
it
into
a
separate
repo,
so
we
never
got
a
owner.
So
we're
gonna
move
these
three
items
to
next
quarter.
A
A
A
Highlights
for
this
quarter
that
got
completed
volume,
cloning
moving
to
beta
the
ephemeral,
inline
William
support
for
CSI,
moving
to
beta
the
attached,
sliminess
sorry
go
ahead
and
then
online
offline
volume
resizing
for
CSI
to
beta.
So
a
lot
of
the
CSI
features
are
moving
over
to
beta,
so
good
progress.
There,
a
lot
of
work
that's
pending,
and
this
is
normal.
We
have
lots
of
very
large
items,
so
they
tend
to
span
multiple
quarters,
hopefully
the
yellow
items.
We
can
make
some
progress
on
next
quarter
and
move
some
of
the
Reds
to
yellow.
A
A
H
So
yeah,
my
name
is
Josh
bucket
jet
stuck
here
with
James
amico,
a
concert
manager.
Yes,
if
you're
not
too
familiar
it's
an
operator
in
communities
to
not
to
make
killer
certificates
and
we're
doing
quite,
they
have
like
recent
need
to
kind
of
increase
the
kind
of
integration
points
with
kind
of
other
parts
of
kubernetes.
H
So
this
is
involve
kind
of
creating
a
new
resource
kind
of
intermediary
between
like
a
certificate,
difficult
quest,
yeah-
and
this
is
another
to
kind
of
enable
people
to
do
kind
of
out
of
tree
issuers
rather
than
just
kind
of
entry
stuff.
But
it's
also
enabled
us
to
kind
of
play
around
with
things
like
a
CSI
driver
which
I've
been
looking
into
so
yeah
I'll.
Give
a
little
demonstration
here,
so
I've
got
something
a
jirt
installed
here
with
the
CSI
driver
and
I
can
play
an
example.
H
Application
here
and
I'm,
making
use
of
the
new
kind
of
inline
volume
mount
as
well,
which
is
pretty
cool,
so
what's
happening
here.
Is
they
kind
of
send
it
off
this?
These
volume
attributes
to
the
kind
of
set
manager
CSI
driver,
the
CSI
driver,
was
going
to
kind
of
create
this
certificate
request,
resource
certain
manager,
the
operators
gonna
come
and
resolve
that
certificate.
H
The
CSI
driver
grabs
a
certificate
and
then
writes
it
to
file
which
is
pretty
cool.
It
means
that
we
can
kind
of
generate
private
keys
locally
on
the
node
store
them
locally
on
the
node.
They
never
leave
yeah,
it's
all
kind
of
local
file,
it's
quite
cool,
and
you
also
get
this
kind
of
nice
property
where
of
course,
because
it's
a
volume
pods.
H
Cool
yeah,
yeah
kind
of
going
forward.
We
want
to
think
about
things
like
renewal,
so
since
we're
kind
of
not
relying
on
certain
managers,
kind
of
certificate,
resources,
kind
of
one
layer
below
its
CSI
driver's
responsibility
to
kind
of
handle
things
like
renewal,
which
should
be
quite
simple
to
do,
I
guess
it
kind
of
is
a
concern
of
just
having
long-running
go
routines
kind
of
that
understand
what
a
duration
of
a
certificate
is,
and
it
should
be
quite
simple
to
kind
of
for
CSI
job.
G
H
D
A
A
H
A
Cool
I
mean
from
the
CSI
and
storage
perspective.
This
is
super
cool
and,
like
Michelle
said,
this
is
exactly
the
kind
of
use
cases
we
had
in
mind.
So
it's
really
cool
to
see
this
light
up,
I'd
like
to
see
it
on
the
drivers
list
as
something
that
can
be
advertised
to
folks
who
are
looking
at
CSI
drivers
and
Michelle.
Maybe
we
should
segment
our
drivers
page
into
persistent
and
ephemeral
drivers
to
make
it
easier
to
kind
of
parse
through
now
that
we're
getting
more
and
more
ephemeral
drivers,
yeah.
A
A
Step
would
be
presented
to
sig
sig
on
sig
security,
whatever
their
name
is
because
they
probably
have
a
much
better
understanding
of
the
the
use
cases
for
this
and
then,
if
they
are
onboard,
then
yeah
I
think
a
blog
post
would
be
a
good
next
step.
It
would
demonstrate
multiple
things,
security,
as
well
as
these
storage
interfaces,
also
cool.
A
B
Yes,
the
question
basically
was
when
we
do
unknown
onstage
volume
in
in
CSI:
there's
no
present
requirement
or
recommendations
around
having
to
dismount
the
volume.
So
it
seems
it's
not
you
CSI,
for
some
of
the
entry
cases
like
whether
operate,
whether
a
plug-in
needs
to
dismount
a
volume
or
not
is
kind
of
unspecified
and
you're
trying
to
get
some
guidance
around
that
like.
B
C
D
D
C
D
If,
if
it
was
different,
if
it
was
like,
kubernetes
was
responsible
for
the
file
system
and
I
would
say
it
should
be
kubernetes
responsibility
to
on
flood
to
flush
before
calling
unpublished
on
the
driver.
But
in
this
case,
since
the
driver
is
the
one
that
creates
the
file
system
and
the
driver
is
the
one
that
actually
knows
if
it
needs
to
melt
or
amount,
then
it
should
be
the
drivers,
responsibility.
C
B
Make
sense
so
I
think
that
counterpoint
that
was
raised
by
I
believe
James
and
others
was
what
happens
to
app.
Like
should
apps
start
to
rely
on
this
functionality
and
whoa,
and
if
they
rely
too
much,
then
they
might
not
be
able
to
deal
with.
You
know
a
node
crashing
and
angry
and
bubbly
scenario.
So.
F
A
A
Thank
you.
Deep
next
item
is
face
to
face
meeting.
We've
had
a
number
of
kind
of
major
questions
and
issues
resolved
or
come
up
around
the
future
of
snapshots
and
data
source
and
cloning
and
data
populate
errs
and
in
the
past,
when
we've
had
a
lot
of
big
design
questions.
What
has
served
us
well
is
a
face
to
face
meeting
the
last
one
was
in
2018,
so
we're
a
little
bit
overdue
for
it,
and
so
I'd
like
to
propose
doing
another
face
to
face
meeting.
We
have
a
couple
of
options.
A
One
is
at
Q
Khan
in
San,
Diego
in
which
was
happening
between
November,
19
and
21.
So
we
could
set
up
something
there.
Another
option
was
you
potentially
do
something
before
that,
maybe
in
October
and
maybe
on
the
East
Coast
I
was
talking
to
been
offline
and
he
was
talking
about
potentially
being
able
to
host
that
yeah.
C
Yeah
we
have.
We
have
space
here
here,
Netta
and
you
know
RTP
North
Carolina
that
I'm
could,
depending
on
the
date
and
the
number
of
people,
that
the
main
thing
I
need
to
figure
out
is
how
many
people
would
come
and
then
need
to
look
at
dates
to
see
when
I
could
get
space.
But
we
definitely
have
availability
and
can
probably
sort
out
the
budget
for
food
kinds
of
things
as
well.
A
A
C
A
A
So
we
could
use
that
as
a
backup
option,
we'll
start
with
the
CNC
F.
If
there's
anybody
on
the
call
who
thinks
their
company
might
be
able
to
sponsor
something
like
a
hotel
room
for
probably
a
day,
then
that
would
be
great
feel
free
to
reach
out
to
me
or
Ben
offline
and
and
we'll
take
it
from
there.