►
From YouTube: Kubernetes SIG Storage 20180201
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 01 February 2018
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.hfej0v9dycbb
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Saad Ali (Google)
Chat Log:
[none]
A
A
There's
anything
you
want
to
discuss
today
feel
free
to
add
it
to
the
agenda.
The
link
is
included
in
the
calendar.
Invite
first
up
we're
gonna,
go
through
the
planning
spreadsheet
and
get
a
status
update
of
the
things
that
this
sake
is
working
on.
For
this
quarter
then
we'll
go
through
the
design
reviews
looks
like
Tomas.
A
Has
something
he'd
like
to
discuss
feel
free
to
add
to
this
as
we
go
so
go
into
the
spreadsheet
right
off
the
bat
we
have
generic
storage
topology
work,
which
Michelle
is
leading
Michelle
where
you
she
was
having
trouble
joining
I,
don't
know
if
she
was
able
to
get
into
this
room.
Michelle.
Are
you
there.
A
B
B
A
A
All
right
next
up
is
the
CSI
volume
plug-in
implementation.
We
hold
meetings
three
times
a
week
and
put
the
status
update
in
a
separate
doc,
which
is
linked
here.
Please
feel
free
to
take
a
look
at
that.
For
the
most
part
we
are
on
track.
There
are
a
couple
of
big
items
that
were
tracking
closely
like
moving
adding
block
device,
support
to
CSI
the
topology
work,
which
will
have
effect
on
CSI
and
a
couple
other
things,
but
I
think
largely
were
on
track
to
go
to
beta
this
quarter.
A
What
we
decided
to
do
was
break
the
CSI
project
up
and
move
the
core
API
to
beta
and
not
necessarily
have
not
necessarily
blocked
all
of
CSI
on
additional
features
like
block
volume,
support
and
things
like
that.
That
way,
we
can
get
the
core
API
to
beta
and
folks
can
start
developing
against
the
core
API
and,
as
new
functionality
is
added,
they
can
add,
add
support
for
that
as
well.
A
A
D
A
A
F
F
F
F
A
G
Yeah,
so
we
started
to
design
so
I
didn't
20.
If
you
go
to
me,
how
do
you
add
a
snapshot?
Functions
in
also
the
the
snapshot
controller
and
the
driver
change
is
needed
to
make
this
work
so
deciding
you
know
which
piece
of
controller
should
be
industry
and
which
piece
should
be
out
which
me
so
yeah.
We
are
working
on
that.
Okay,.
G
A
There
was
a
meeting
last
week
to
summarize
what
the
contents
of
that
meeting
was
was
to
decide
whether
we
should
go
in
tree
or
not,
and
if
we
do
what
the
timeline
for
that
should
be,
and
what
other
issues
exist
around
the
snapshot
API.
The
conclusion
of
that
was
as
a
sig
we
decided.
We
would
like
to
move
the
snapshot
api's
entry.
It
gives
us
benefits
and
things
like
being
able
to
build
higher
level
primitives
on
top
of
it.
If
it's
part
of
the
core
API,
that
said,
we
decided
not
to
pursue
pushing
entry.
A
A
A
H
A
A
J
K
K
J
So
thanks
note
as
Adam
and
Jamie
today,
he
said
waiting
for
the
Tangut
nodes,
with
status
basically
approved
by
no
team,
and
that's
really
someone
to
kind
of
work
on
that.
That's
a
clean
array,
so
the
other
PR
opens
is
also
works.
It
just
I
kind
of
propose
to
use
this
way,
but
if
anyone
has
comments,
discuss.
F
K
A
So
I
think
last,
where
we
left
off
was
discussion
about
what
exactly
this
would
look
like.
There
is
a
proposal
to
pull
mount
devices
out
of
attach,
but
the
open
question
was
around
access
modes
and
whether
that
should
be
enforced
for
these
types
of
volumes
and
I
will
follow
up
on
that
discussion
and
see
what
the
current
status
of
that
is
would
lock
with
Lynn.
A
L
K
Sod
I'm
interested
in
this
book
because
I've
been
cracking
several
bugs
that
are
related
to
this
that
affect
EBS
and
I,
realize
it
is
not
just
GC
GC
problem
and
yeah.
So
I
try
to
open
a
proposal
looking,
and
should
we
consider
this
as
a
bug
fix
or
like
it
could
be
a
design
problem
as
well,
because,
let's.
A
A
M
Quite
a
simple
proposal:
what
is
it
about?
There
is
a
security
context,
object
that
allows
user
to
specify
various
security
related
attributes
of
a
port
in
it
includes,
say,
FS
Brook,
which
basically,
it
tells
qiblah
to
mount
all
the
or
change
ownership
of
all
the
files
in
the
mound's
certain
settlement
or
group.
That
means
when
the
period
is
mounting
volumes
for
hot,
with
a
physical
set,
it
starts
with
records
ability
changing
ownership
of
all
the
files
on
that
volume
up.
M
A
N
H
N
Like
there's
other
permissions
or
as
Linux
changes
that
cause
it's
actually
a
kind
of
a
design
pattern,
we've
been
thinking
about
for
these
deep
recursive
actions.
You
know
so
instead
of
you
know,
waiting
two
or
three
minutes
and
then
letting
it
fail.
We'll
do
the
top
level
and
then
asynchronously
change
the
rest
of
this
stuff,
reverse
and.
N
A
M
K
K
This
has
to
be
a
follow
up
design
to
this
as
well.
That
applies
the
the
like
small,
like
desert,
like
the
non-recursive
SELinux
relabeling
right
now
we
are
doing
by
default,
recursive,
facial
and
living,
and
that
also
times
cause
the
power
to
time
on.
The
talker
gives
both
options,
but
that's
a
different
thing.
O
A
N
Know
we
told
it
it
could.
I
mean
this:
it's
probably
not
the
best
way
to
do.
We
should
probably
have
a
hook
to
notify
the
application
that
the
permission
changes
complete
and
then
let
it
you
know,
go
instead
of
just
assuming
that
the
top
level
is
good
enough.
Mm-Hmm
they're,
just
you,
know,
kind
of
a
super
kind
of
a
feeling
that
the
top
level
may
get
a
majority
of
the
cases.
N
A
M
N
D
N
A
A
A
D
O
O
A
N
H
A
F
So
Luis
and
I
are
meeting
tomorrow
to
start
scheduling
those
out.
There
was
a
significant
from
yawn
that
maybe
we
could
reach
out
to
the
communities,
because
some
of
the
topics
may
be
overlapped
with
general
kubernetes,
but
Louisa
and
I
were
going
to
try
to
make
it
very
storage
focused.
If
there
are
some
that
overlap,
maybe
we
can
collaborate
but
I
I'm,
happy
with
the
list
and
I
haven't
gotten
any
suggestions.
Otherwise,
and
a
lot
of
people
have
volunteered.
So
if
you
volunteered,
please
look
forward
to
an
invite
from
Luis
and
I
to
do
that.