►
From YouTube: Kubernetes SIG Storage 20180315
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 15 March 2018
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.9gqraxwdaha5
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Saad Ali (Google)
Chat Log:
09:22:52 From hayley : The survey for times for the f2f has yet to be put up, right?
A
A
This
is
just
to
get
things
kick-started
and
after
that
we'll
discuss
some
of
the
1.10
release
blocking
bugs
that
we
have
for
the
storage
sake
and
then
finally,
Paris
is
going
to
give
a
presentation
from
country
Bex
about
all
the
things
that
are
going
on
there
all
right.
So
let's
go
to
the
planning
spreadsheet
and
get
status
updates
here.
A
So
I
believe
Michele
is
out
this
week,
I'm
so
either
me
or
Yann,
or
someone
else
will
have
to
give
an
update
on
her
behalf
for
the
generic
storage
topology
solution
work,
the
design
work
was
completed,
so
we
can
go
ahead
and
mark
that
green
next
quarter.
We
can
begin
implementing
this
volume,
scheduling,
dynamic,
provisioning
design
was
also
completed
and
is
ready
for
implementation.
A
Work
green
on
all
of
these
CSI
volume
plug-in
implementation.
The
core
API
has
moved
to
beta.
We
are
finishing
up
some
of
the
external
components
in
the
CSI
group.
At
the
moment.
You
can
look
at
this
link
here
to
see
the
detailed
status,
but
we
are
on
target
to
be
ready
for
1.10,
so
CSI
is
going
to
beta
this
quarter
as
well.
B
B
A
D
D
D
B
D
A
F
Yes,
Todd
I
wanted
to
mention
one
thing.
Just
so
people
know
this.
These
end
end
tests
for
local
block
did
catch
an
issue
that
went
in
between
adding
local
block
to
the
plug-in
and
the
tests
creation,
which
was
only
about
ten
days
later.
So
is
this
for
everybody.
You
know
the
end
end
tests
are
really
valuable.
We
can
end
up
with
regressions
if
we
don't
have
them.
A
Absolutely
we
can't
consider
a
feature
done
until
testing
is
completed,
so
let's
not
skimp
on
that
and
make
sure
we
deliver.
High
quality
features.
We're
going
to
point
that
out
at
the
end
of
this
meeting,
when
we
go
over
some
of
the
bugs
for
1.10
I
think
SIG's
towards
a
little
bit
more
in
the
blocking
path.
This
release,
then
we
would
like
to
be-
and
testing
has
a
lot
to
do
with
that.
Next
up
is
to
do
local,
ephemeral,
storage
to
beta
this
got
merged
and
the
API
has
been
moved
to
beta
snapshots.
A
C
A
A
There
is
one
issue
around
this
that
you
should
be
aware
of,
if
you're
upgrading
to
1.10,
basically
the
way
that
this
feature
works
is,
if
you
create
a,
if
you
have
a
pvp
VC,
that's
bound,
the
pv
is
marked
with
a
finalizar,
it's
preventing
its
deletion
without
the
PVC,
deleting
it
first
and
similarly,
if
you
have
a
pod,
that's
referencing
a
PVC.
The
PVC
can't
be
deleted
until
the
pod
is
deleted.
There
is
a
controller,
a
new
controller
in
1.10
that
is
enabled
by
default.
That
manages
this
in
1.9.
A
For
now.
The
release
team
has
suggested
documenting
this
as
a
No,
an
issue
which
is
the
plan
moving
forward,
we're
hoping
to
also
add
in
logic
in
1.9,
in
a
1.9
X
future
1.9
X
release
that
will
basically
be
a
dumbed
down
version
of
this
controller
that
will
just
remove
any
of
these
finalized
errs
that
it
sees
to
prevent
these
objects
from
getting
stuck
in
a
state
that
they
cannot
be
deleted.
This
is
an
issue
that
was
caught
caught
by
the
upgrade
downgrade
tests
and
just
something
to
be
aware
of.
A
A
G
A
There
were
some
race
conditions
discovered
late
in
the
release
that
Jing
and
Yan
have
been
working
on
for
the
last
couple
of
weeks.
At
this
point,
it
is
considered
the
bugs
around
this
or
considered
a
blocker
for
the
release,
but
it
looks
like
we're
getting
closer
to
merge.
I
think
on
the
latest
fix
PR
for
this
yawn
left
a
couple
of
comments
and
the
release
managers
were
asking
if
those
are
blocking
or
not
blocking.
So
if
you
have
a
chance
yawn,
can
you
take
a
look
at
that
sure?
A
A
G
F
A
G
So
design
got
got
started.
There
are
some
feedbacks
I'll
open
it
to
github.
I
think
the
only
remaining
bit
was
like
I
have
to
add
CSI
into
consideration,
therefore,
and
I
think
I
figured
out
a
way,
so
I
just
move
it
to
get
up
soon
like
this
next
week.
Sometime
and
then
we
can
but
I'm
hoping
that
we'll
have
implementation
down
in
1.11.
A
And
then
we
scratch
these
off
for
1.10,
and
that
concludes
our
1.10
tasks.
I've
created
a
new
tab
for
1.11
feel
free
to
start
moving
over
tasks
that
you
would
like
to
work
on
for
the
next
quarter
and
or
that
you
think
we
should
work
on,
but
you're
not
necessarily
signing
up
for
it,
but
you
want
someone
to
work
on
just
things
that
this
cig
should
work
on,
go
ahead
and
start
pre-populating
that
list
and
then
we'll
review
it
at
the
next
meeting.
A
That's
pretty
much
it
on
the
planning
side.
Next
up
is
the
bugs
to
discuss
I.
Think
we
already
reviewed
the
two
big
ones
that
are
currently
blocking.
One
is
the
upgrade
test
for
a
PvP
VC.
The
folks
working
on
this
are
rude
FS.
So
there's
these
tasks,
which
are
documenting
the
issue,
the
woman.
Do
you
want
to
give
a
status
update
on
on
this.
A
Cool
okay
and
then
there
is
a
second
effort
in
to
try
to
fix
this
in
a
more
automated
way
and
Pavel
is
going
to
be
working
on
that
I.
Believe
football
is
no
longer
at
Red
Hat,
but
he
has
volunteered
to
to
continue
to
work
on
this
on
the
other
issue
that
we're
tracking
is
the
orphan
directories
and
mounts
after
downtime.
This
was
introduced
as
part
of
the
the
volume
reconstruction
logic
that
was
added
this
quarter,
and
at
this
point
there
are
two
PRS
that
are
out.
A
Okay,
it
looks
like
she
wasn't
able
to
join
us
today,
but
Paris
wants
to
give
us
an
update
from
contrib
ex
on
the
things
that
have
been
going
on
in
that
steak.
She
and
other
folks
from
the
contra
Beck's
sake
are
going
to
the
other
SIG's
to
try
to
share
information
and
educate
folks
and
make
sure
everybody
knows
what's
going
on
with
a
higher
level
with
the
project
we'll
go
ahead
and
move
that
to
the
next
meeting,
and
hopefully
she
can
make
it
and
the
final
agenda
item
added
here.
A
J
A
Okay,
yeah,
maybe
by
let's
see
so
if
we're
gonna
do
April.
That
is
right
around
the
corner,
so
we
would
probably
it
would
be
too
early.
So
I
think,
based
on
what
Steve
Wang
said
for
folks
who
are
actually
organizing
this,
it's
good
to
have
at
least
a
one-month
heads-up,
and
that
would
mean
if
we
spend
the
next
two
weeks
planning
for
which
location
we
want.
A
C
I
A
If
we're
gonna
end
up
doing
it
in
May,
there
is
the
option
during
the
first
week
of
May
to
do
it
in
queue.
Connie
you,
the
challenge
there,
of
course,
is
I'm,
not
sure
how
many
people
are
going
to
be
able
to
make
it
out
there.
So
maybe,
as
part
of
your
survey,
you
can
include
that
as
an
option
and
and
then
we
can
follow
up
on
that
in
our
next
meeting.
Sure
and.
C
A
Think
one
thing
that
we've
learned
on
the
kubernetes
CSI
side
is
that,
having
a
separate
reason,
each
component,
that
is,
that
ships
is
very,
very
nice
rather
than
clumping
them
all
into
a
single
repo.
We've
learned
the
same
lesson
from
external
storage,
so
one
way
or
another.
We
want
to
be
able
to
have
separate
repositories
for
each
one
of
the
features,
whether
they're
all
gonna
end
up
under
kubernetes,
kubernetes,
external
or
some
other
sig.
Specific
org
will
sort
that
out
and
thanks
Brad
for
help
and
drive
that.