►
From YouTube: Kubernetes SIG Storage 20180426
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 26 April 2018
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.evgsvfys7vqx
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Saad Ali (Google)
Chat Log:
N/A
A
All
right
today
is
April
26
2018.
This
is
the
meeting
of
the
kubernetes
storage
special
interest
group.
As
a
reminder,
this
meeting
is
public
recorded
and
published
on
YouTube
on
the
agenda.
Today,
we're
going
to
go
over
the
planning
spreadsheet
for
1.11
for
q2,
then
we're
gonna.
Take
a
look
at
some
of
the
PRS
that
folks
have
posted
in
the
agenda,
talked
about
the
face-to-face
and
onboarding
and
testing.
A
Not
here
this
time,
so
we'll
get
an
update
next
time,
kubernetes
CSI,
topology
implementation,
it's
something
that
Chang
is
working
on
at
the
moment.
I'm,
not
sure
if
Chang
is
on
the
line
either,
but
I
can
provide
an
update
here.
Chang
has
been
working
on
a
design
that
is
ready
to
send
out
and
he'll
be
sharing
that
shortly.
A
Csi
topology
design
the
PR
is
still
pending.
There
are
some
open
questions
that
I
need
to
address,
and
I
intend
to
do
that,
probably
after
a
Q
Khan,
which
is
this
coming
week,
gcpd
topology
integration,
I,
don't
believe.
Michelle
has
gotten
a
chance
to
start
on
that
same
thing
with
AWS.
We
don't
have
anybody
assigned
to
this
if
anyone's
interested
in
working
on
something
in
the
storage
say.
This
is
an
open
item.
A
B
B
Step
is
we
need
someone
to
accept
or
review
the
accept.
The
design
proposal
and
len
is,
has
not
updated
the
design
proposal
with
the
latest
change
yet
because
he
is
writing
the
code
to
verify
that
bit
that
we
should
only
poll
for
resizable
PVC.
So
he
will
do
that
by
tomorrow
or
something
and
then
okay.
A
C
C
D
B
A
B
A
Cool
anyone's
interested
in
that
it's
a
pretty
fun
PR
with
a
lot
of
interesting
design,
questions
on
the
CSI
side,
volume
snapshot
and
restored
design.
So
this
is
the
restored
design.
Shing.
Do
you
want
to
talk
about
that?
So
I
broke
these
items
into
three
different
things,
because
they're
essentially
three
different
tasks?
One
is
coming
up
with
a
design
for
snapshot.
Restore
second
is
moving
the
volume
snapshot,
controller
entry
and
third
is
adding
volume,
snapshot
and
restore
capability
to
CSI,
so
are
either
Shing
or
Jing
on
the
line
today,.
A
E
Gene
might
be
troubling
so
delicious,
oh
my,
so
that
was
the
CSI
basically
right
now,
I
think
folks,
at
the
CSI
side,
are
pretty
much
okay
with
the
with
aspect.
It's
just
so
there
are
some
small
things
we
need
to
sort
out
before
that
can
be
merged
and
another
thing
is
we
run
into
your
github,
but
because
there
are
so
many
comments
and
some
folks
cannot
open
the
arch
review
so
couldn't
get
it
much.
Wasn't
that
and
then
so
I
have
to
actually
open
and
I
submitted
another
PR
and
they
reference
the
app.
E
F
E
Yeah,
so
this
one
basically
has
dependency
on
the
CSI
and
everything
else
right
so,
but
I
have
started
coding.
Basically
because
this
one,
you
know
as
many
pieces,
you
need
a
well.
We
do
need
to.
First
you,
the
api's
entry
and
and
also
we
need
the
snapshot
controller
piece
in
trace
of
that
actually
has
a
lot
of
generating
a
code
involved.
E
A
E
I
yeah
I
started
to
looking
at
this.
This
is
so
like
the
entry
piece
right,
so
this
is
together
and
then
and
I
started
and
basically
that's.
The
different
repo
will
have
a
external
snapshot.
Err
which
is
under
crouppen
et
si
si
area
missile,
but
that
will
be
dependent
on
the
CSS
back
that
has
to
emerges.
E
G
E
E
A
Proposal
is
a
very
good
reason
why
this
design
should
be
pulled
entry.
I
think
a
lot
of
it
depends
on
it
being
entry
mm-hmm,
once
that's
approved
and
finalized,
then
I
think
the
U
and
Jenkins
set
up
some
time
with
cig
architecture
and
present
to
them
the
reasons
why
we
think
it's
compelling
for
this
to
move
entry
and
if
they
approve,
then
we
can
move
forward.
A
E
A
A
E
A
B
So
we've
presented
this
in
front
of
signaled
and
and
and
sad
you
and
Michelle
and
Tim
had
a
discussion
as
well.
I
think
the
current
consensus,
that
is
the
video
need
the
CLI
parameter
I,
have
to
updated
the
proposal
with
that
we
don't
need
the
the
values
to
come
from
cloud
provider.
It
has
to
come
from
volume.
Plugins
are
outdated.
The
proposal
with
that
there's
an
opera
there's
a
way
to
work
with
wall
and
plugins
to
and
the
only
thing
I
think
concerned.
B
The
only
like,
like
problematic
bit,
is
figuring
out
whether
it
has
to
go
in
capacity
or
like
the
status
start
attach
limits,
as
proposed
in
the
proposal.
The
sick
note
just
like
yeah
capacity,
has
its
own
problems.
I
am
currently
leaning
towards
using
status
limit.
It
seems
simpler
and
scheduler
already
has
logic
for
dealing
with
volume.
So
it's
not
like
unheard
of,
but
we
need
to
build
a
consensus.
Basically.
B
B
A
A
A
H
A
C
A
Thanks
and
then
the
PV
helped
monitoring
how
month?
Could
you
open
an
issue
for
that?
Actually,
we
don't
need
an
issue
since
it's
a
design,
I'll
fix
this
coloring
I,
don't
know
why
it's
just
showing
up
as
red
and
then
is
Vlad
on
the
line
at
all.
Yeah
Vlad
looks
like
there's
a
couple
of
issues
for
CSI
that
you're
working
on.
Could
you
make
sure
that
there
is
issues
open
for
them?
Yeah.
A
G
A
A
A
G
A
A
A
Okay,
I
can
give
an
update
on
their
behalf,
so
they're.
Taking
a
look
at
this
design.
It's
going
to
be
a
massive
massive
design.
We
need
to
make
sure
that
we,
when
we
do
this,
it's
absolutely
seamless
to
the
end
user.
They
should
be
able
to
use
the
entry
volume
plugins
like
GCE,
PD
or
AWS
EPS,
and
not
know
that
it's
actually
being
fulfilled
by
a
CSI
driver
under
the
covers
instead
of
the
entry
code,
when
we
swap
it
out.
This
requires
planning
how
the
swap
out
is
actually
going
to
work.
A
A
Other
challenges
are
figuring
out
how
to
do
upgrade
and
downgrade
so
that,
if
you
do
a
rollback
from
the
version
that
this
is
automatically
enabled
on
to
one
that
it's
not
that
we
have
a
good
way
to
switch
back
and
forth.
I
have
some
mechanism
to
opt
in
as
this
is
gradually
rolled
out
and
be
able
to
switch
back
and
forth
and
then
deciding
what
set
of
volume
plugins.
We
want
to
tackle
for
this
migration.
A
Initially,
the
initial
feeling
on
this
is
that
for
the
alpha
and
beta
implementations,
we
are
gonna
focus
on
the
cloud
provider
volume
plugins.
First.
The
reason
for
that
is
that
there's
a
big
push
within
kubernetes
to
move
the
cloud
provider
code
out
of
the
kubernetes
core
and
storage
is
a
big
user
of
that
cloud
provider
API
the
sooner
we
can
get
those
plugins
moved
off
to
CSI
the
sooner
we
unblock
this
removal
of
cloud
provider
code
from
the
core
kubernetes
codebase.
A
A
There
are
party
plugins,
like
port
works,
etc,
basically
have
a
way
to
proxy
them
to
their
CSI
driver
equivalents
and
then,
finally,
that
will
leave
basically
just
ephemeral
volumes
and
local
storage.
Ephemeral
volumes,
I
think,
are
fine,
where
they
are
because
they're
very
specific
to
kubernetes
things
like
secrets
downward
api,
and
there
are
performance
implications
of
being
in
tree
which,
for
for
these
ephemeral
volumes
is
important,
because
every
single
volume
basically
has
a
secret
mounted
to
it.
Sorry,
every
single
pod.
So,
for
those
reasons,
I
think
leaving
the
ephemeral
volumes.
Entry
will
be
fine.
A
A
A
A
No,
okay:
this
is
a
lower
priority.
So
don't
worry!
If
you
don't
get
to
it,
it's
kinda,
nice
to
have
I
know
you
guys
have
a
lot
on
your
plate,
so
that
is
perfectly
okay
move
GCU
cloud
provider,
disk
api
to
auto
generated
code,
I
believe
Chang
sent
out
a
PR
for
this
I.
Remember
seeing
it
at
some
point
Michelle
do
you
have
any
idea.
A
J
I
have
direction
from
the
steering
committee.
It's
just
a
matter
of
going
and
breaking
out
our
projects,
either
in
they'll.
Look
into
that
ya
know
and
then
also
into
suit
the
stored.
Sig
I,
don't
know
what
it's
called
manifest
or
whatever
I
know.
There's
a
couple
I.
You
know
broader,
like
six
storage
items
that
we
have
to
have
like
that
are
scoped
and
like
a
mission
statements
and
other
things
that
kind
of
tied
into
this
I'll
probably
pick
those
up
along
the
way
cool.
B
J
A
J
A
A
K
A
A
L
This
started
that
as
an
issue
filed
by
Red
Baron,
it
also
happened
to
affect
me,
so
this
is
there's
a
potential
for
data
corruption
with
certain
types
of
I
scuzzy
backends
that
have
multiple
multiple
luns
on
the
same
target
iqn
because
of
the
way
cubelet
does
attaches
and
detaches.
So
I
have
a
PR
for
this.
That's
already
up,
and
this
is
my
first
kubernetes
PR,
so
I
probably
did
something
wrong,
but
please
let
me
know
my
feelings
won't
be
hurt.
D
H
A
A
A
A
A
A
H
H
A
K
A
A
A
A
J
A
All
right
cool
and
if
anybody
else
in
the
sig
is
interested
in
understanding
this
project,
maybe
Karen
you
guys,
could
do
a
presentation
at
a
future
state
meeting
cool.
Thank
you
very
much.
Next
on
the
agenda
is
a
discussion
about
the
face
to
face.
It
is
on
May
15
and
16.
Steve
Wang
and
VMware
have
graciously
offered
to
host
this
Steve.
You
want
to
talk
about
that
sure.
M
I
think
the
details
are
in
the
meeting
notes
and
then
there's
a
link
to
the
document
with
a
proposed
agenda
for
those
who
haven't
been
to
it
before
we'd
normally
start
with,
let's
say
an
ambitious
list
of
agenda
items
and
end
up
with
too
much
to
discuss
compared
to
the
two
days
available.
So
we
start
these
meetings
with
a
vote,
but
please
get
your
agenda
items
into
that
document
now,
because
others
can
see
them
and
have
a
little
time
to
think
about
them
and
maybe
combine
items
and
things
like
that.
M
There
have
been
people
who
retro
actively
edited
their
name
in
after
the
deadline
I'll
be
watching
for
that
and
what
is
the
deadline?
It's
tomorrow
tomorrow?
Okay,
so
if
you're
a
didn't
say
what
time
so,
I
think
I'm
willing
to
be
pretty
flexible,
although
I
think
there's
limited
numbers
of
seats.
So
if
you
wait
until
midnight,
it
might
be
too
late
anyway,
then
it
just
gives
time
for
planning.
The
other
thing
I'd
like
to
point
out
is:
we've
historically
held
a
dinner
on
the
Tuesday
night.
M
It's
been
great
for
networking
meeting
and
getting
to
know
I'd
like
to
do
that
again,
but
you
know
young
we're
kind
of
blew
the
budget
just
renting
the
venue.
It
turned
out
that
the
on-campus
rooms
were
all
full
for
a
group
of
this
size
and
we
didn't
have
enough
to
also
host
the
dinner,
but
if
some
other
organization
would
like
to
do
it,
that
would
be
great.
M
The
reason
I
say
that
is,
you
know
we
could
always
go
there
and
everybody
buy
their
own
dinner,
but
having
arranged
these
dinners
before
once
you
get
over
a
group
of
about
12.
These
venues
want
a
contract
and
if
you
call
some
restaurant
on
the
phone
asking
them
the
table
for
30,
what
payment
in
advance-
or
at
least
some
document
from
a
company
that
they
think
is
they
could
hold
accountable
if
you
don't
show
up
yeah.
M
So
if
anybody
has
an
org
that
wants
to
sponsor
that
great,
if
not,
if
anybody
even
has
ideas,
I'm
I'm
gonna
be
traveling
from
LA
and
I'm,
not
that
familiar
with
the
mountain
view
area.
But
if
somebody
knows
of
a
big
pizza,
parlor
setting
or
something
like
that,
that
can
accommodate
mixed
dietary
demands
and
has
huge
tables.
That
would
be
great
and
please
edit
your
suggestions
in
there.
If
you
think
of
anything,
yeah.
M
One
other
thing:
some
people
have
sent
me
emails
for
access
to
that
document
for
RSVP'ing
or
editing
the
agenda
and
the
state.
The
community
standards
for
the
kubernetes
project.
That
and
I
recently
learned
this
and
I'm.
Appreciating
the
huge
amount
of
work
sod
is
doing
managing
this
group
because
I
became
chair
of
another
sig
a
few
weeks
ago,
but
they
actually
advised
that
you
constrain
editing
to
group
members
and
there
there
have
been
unfortunate
in
sin
in
the
past
and
other
open-source
projects
where
people
for
whatever
reason
graffiti
tag
in
essence,
documents.
M
So
rather
than
having
an
admin.
Add
your
name
one.
At
a
time.
If
you
simply
join
the
list
in
Google
Plus
and
if
you're
logged
in
under
whatever
ID
used
to
join,
you
should
have
no
trouble
editing
it,
but
sending
the
request
to
add
your
secondary
email.
Probably
isn't
the
most
efficient
way
to
scale
a
saying
so
just
pointing
that
out.
Yep
completely.
M
A
A
And
if
you're
interested
in
helping
sponsor
the
dinner,
please
reach
out
to
two
to
both
me
and
and
to
Steve
and
we'll
figure
that
out.
If
we're
unable
to
get
folks
I'll
see
if
we
can
get
Google
or
CNCs
to
sponsor
and
of
course,
last
the
worst
case,
we
can
just
all
pay
on
our
own,
which
shouldn't
be
an
issue
either
all
right.
Next
up,
Aaron
Boyd
has
been
organizing
a
cig
onboarding
series,
which
is
a
bunch
of
live
meetings
with
folks
from
this
or
veterans
who've
been
around
for
a
while
presenting
on
various
topics.
A
I
That
was
yeah
rescheduled,
sad
she's.
G
A
Keep
an
eye
out
for
those
I
think
there's
a
link
to
the
document.
If
someone
can
post
that
here,
we're
all
if
this
is
being
tracked,
feel
free
to
attend
those
meetings
live
or
eventually
they
will
be
posted
on
our
YouTube
channel
and
it'll
be
great
content
for
how
to
on
board
if
you're,
a
new
contributor
and
you're
interested
in
contributing
to
the
storage,
sig
and
then
finally,
I
believe
Aaron
is
also
planning
to
restart
that
the
the
testing
subgroup
for
six
storage
at
the
end
of
May.