►
From YouTube: Kubernetes SIG Storage 20170511
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 11 May 2017
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.x7a4cevohn2a
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Saad Ali (Google)
Chat Log:
09:15:24 From Jason Singer DuMars : Great work!
09:15:37 From venkat : Awesome work!
09:15:39 From venkat : Thanks
09:16:12 From fejta : Again, feel free to head over to #sig-testing on slack if you have any questions or want help. Thanks!
09:28:17 From eparis : Don't worry, both merrian-webster and the OED have redefined literally to mean figuratively.
09:47:50 From Jason Singer DuMars : thank you!!
09:50:25 From eparis : I'll let Erin know you said she is wonderful "right now"
A
All
right
good
morning
today
is
May
11
2017.
This
is
the
bi-weekly
meeting
of
the
storage
special
interest
group.
As
a
reminder,
this
meeting
is
public
and
reported
today
on
the
agenda.
First
item:
we
have
is
a
flaky
test
alerts,
so
we
have
eric
beta
with
us
him
and
his
team
has
been
working
on
something
that's
very
important
to
the
project.
Reducing
flakes
in
the
kubernetes
system
and
eric
has
been
trying
out
a
new
initiative
to
help
reduce
the
number
of
flakes
Eric.
Do
you
want
to
talk
about
that.
B
Yeah
hi
I
am
Eric
and
you
head
over
to
cig
testing.
If
you
want
to
contact
me
or
anybody
else,
I
was
pretty
active
over
there.
On
that
slack
channel
and
yeah.
We
have
a
new
initiative
where
we're
trying
to
drive
down
flakiness
and
increased
instability.
The
primary
output
of
that
I'm
sharing
is,
is
we've
been
sending
out
a
email
every
week,
so
on
Thursday.
B
If
you
head
over
to
the
test
infrared
photo
scroll
down.
You
know
there
are
a
lot
of
there's.
We
list
a
lot
of
the
tools
that
you
have
at
your
disposal.
The
primary
two
ones
are
that
are
going
to
be
useful
to
you,
our
test
grid.
Many
of
you
hopefully
have
seen
test
grid.
It's
not
please
check
it
out.
Something
that
we
are
trying
to
do
is
create
a
dashboard
for
each
cig.
B
B
If
you
compare,
if
you
compare
these
two
Suites
I'll
notice,
that
the
the
GCE
suite
with
the
IDI,
the
crew
brunet
ET
de
GCE
suite,
runs
a
bunch
of
tests
relevant
to
the
CLI
team
of
others,
or
guess
this
downward
volume
I
gave
you
that
is
relevant
to
you,
guys
it's
less
relevant
to
the
cig
CLI
team,
especially
they
created.
If
you
of
this
job,
which
only
shows
them
kubernetes
things
that
are
relevant
to
their
sick
and
then
they
did
that
for
a
bunch
of
different
jobs.
B
And
so
now
they
can
sort
of
hollow
summary,
page
and
quickly
find
out
is,
are
any
of
the
CLI
tests
causing
problems
in
any
of
the
Suites
that
I
care
about,
and
the
nice
thing
is.
If
you
want
to,
you
know
change
any
of
this.
You
can
just
go
into
the
test
grid
and
then
there's
a
config
yamo
which
specifies
all
the
different
dashboards
that
exist
and
yeah.
So
if
you
want
to
add
a
tab
or
change
a
tab
or
do
whatever
you
can
do,
that
test,
create
it
yourself,
sorry
yeah.
It's.
B
So
if
you're
in
test
grid
there's
a
bunch
of
things,
you
can
do
such
as
you
know,
change
how
wide
the
columns
are,
or
you
can
you
know,
have
it
sort
by
failures
so
that
tests
that
fail
the
most
recently
show
up
at
the
top,
and
then
you
can
also
filter
down
there.
There's
like
a
graph
option
as
well.
B
B
There's
also
so
that's
sort
of
a
like.
If
you,
if
there's
a
specific
test
that
you
want
to
monitor
test
grades,
a
great
way
to
do
that,
Ryan
on
the
team
has
been
working
pretty
hard
on
the
triage
dashboard
which
allows
us
or
look
for
tests
across
jobs
services
and
also
to
correlate
failures
between
tests.
So
like
we
can
do
the
same
thing
here.
You
can
filter
on.
You
know
tests
that
volume,
and
so
it
looks
like
the
thing
that
is
most
impacting
is
there's.
B
B
Looks
like
the
downward
api
volume
tests
are
having
some
timeouts
so
potentially
that
this
would
be
so
it's
also
how
the
failure
emails
are
coming
out
is
arm
every
tuesday.
Essentially,
we
just
scan
this
for
the
topic
issues
across
all
jobs
and
make
sure
that
issues
already
exist
for
those
to
click
on
hub
search
and
once
a
day.
We
will
create
issues
for
like
the
top
clusters,
and
so
you
can
see
that
here
and.
B
Yeah,
usually
right
and
or
at
least
are
the
things
that
are,
you
know
most
impacting
flakiness
and
we
were
pretty
sensitive
flakiness.
You
know
like.
I
know
people
get
frustrated
when
if
you've
ever
had
a
PR
that
was,
you
know,
approved
and
ready
to
get
committed,
but
then
it
flakes
tests
that
could
be
irritating
because
you
have
to
come
back
if
notice
that
it
did.
B
You
know,
request
the
bot
to
do
the
testing
again,
and
so
we
have
metrics,
unpublished,
scene
metrics
about
you
know
like
our
weekly
consistencies,
our
weekly
consistency,
and
you
know
so
I
guess
this
week.
I
guess
we're
doing
thus
far
we're
doing
better
than
last
week.
So
last
week,
around
13%
of
PRS
both
passed
and
failed
on
the
same
commits,
so
they
were
experiencing
some
amount
of
flakiness
this
week
gets
down
to
7%
and
I.
B
Think
that's
largely
because
of
last
week
the
unit
tests
they
had
a
couple
of
jobs
that
were
really
flaky,
and
so,
instead
of
only
flaking
13
times,
this
is
something
like
60,
so
yeah.
So
having
you
know,
even
though
one
or
two
flaky
tests
can
cause
a
lot
of
problems
but
yeah,
the
main
thing
is,
you
know
we're
filing
those
three
issues
each
week
we're
trying
to
focus
on.
B
A
So
the
big
ask
for
the
storage
sig
is
to
please
pay
attention
to
these
emails
when
they
come
out
and
if
there
is
a
storage
test
highlighted
in
one
of
these
emails,
we
need
to
jump
on
it
and
if
one
of
these
tests
or
these
issues
is
assigned
to
you,
please
do
your
best
to
prioritize.
It
I
help
deep
lake.
It.
B
A
E
I
added
that
yesterday,
so
on
Tuesday,
we
met
with
the
sig
PM
group
and
I.
Think
one
of
the
things
they're
trying
to
do
is
get
more
visibility
into
what
all
of
the
groups
are
doing,
so
making
sure
that
we
don't
introduce
changes
that
kind
of
are
conflicting
to
the
overall
kubernetes
goals
and
so,
as
a
result
of
that
I
think
they
wanted.
E
So
one
of
the
things
that
they're
trying
to
do
is
is,
like
I
said,
get
us
to
demo
things
that
we're
working
on
and
I.
Think
I'm,
June,
13th
I
can't
remember
the
exact
date
but
I
think
June
13th
is
when
they're
slated
to
to
kind
of
do
something
for
sig
storage,
so
we've
got
a
15
minute
time
slot
there
and
we
have
two
features
going
to
beta.
That
I
think
that
we
should
probably
demo.
One
of
them
is
the
mount
mount
storage
option
for
the
mount
mount
options
and
the
cloud
provider
metrics.
E
A
Thank
you
cool.
Could
you
Max
and
the
invites
to
to
both
of
them
sure
I'll
do
that
all
right?
Thank
you
all
right.
So
next
up,
let's
go
over
the
planning
spreadsheet
and
get
a
status
update
from
everyone.
This
is
the
boring
part
of
the
meeting.
No
one
wants
to
do
but
kind
of
have
to
share
my
screen.
A
G
F
So
kind
of
welcome
to
a
design
to
how
to
fix
that
and
I
posted
a
comment
like
though
it
seems
that
the
prop
mounts
scanning
is
out
of
out
of
one
of
the
options,
but
we're
going
to
use.
So
the
only
solution
that
we
have
is
the
architectural
sticks
around
using
the
full
of
go
routines
and
make
sure
that,
if
there's
a
leak
that
leak
is
contained,
so
I
will
have
a
proposal
by
this
weekend
like
and
but
it
should
be
done
in
time.
I
think
ok,.
F
I
think
right
now
be
a
yawn
poster
change
that
six
for
the
likely
mount
point
cause.
So
we
might
be
back
in
the
red
actually
there,
because
yeah
I'm
correct
me
wrong,
but
yeah
your
cheat
for
the
that
delete
should
not
cross
file
system
boundaries.
That
basically
makes
is
likely
mount
point
checks
check
before
deleting
the
photo
volumes
and
that
will
again
to
the
stat
cause.
A
So,
let's
figure
out
if
this
is,
if
the
workaround
is
still
in
place,
if
it's
still
an
issue,
if
it
is,
we
can
prioritize
it
higher.
We've
got
a
few
weeks
left
until
code
freeze.
If
we
decide
that
this
is,
we
have
a
reasonable
workaround
in
place.
We
can
because
it
sounds
like
the
ticks
are
going
to
be
a
larger
architectural
redesign
and
I
would
rather
us
focus
on
designing
that
or
putting
out
a
proposal
for
the
redesign
in
this
I'll
stone
and
then
working
on
it.
A
A
Okay,
cool
we'll
look
out
for
an
update
next
time.
Stabilizing
aw
got
support
for
EBS
we're
going
to
skip
over
that
there's
not
much
going
on
there.
Some
testing
that
was
being
added
out
of
tree
volume.
Plugins
I've
been
driving
on
this
with
the
other
folks
that
are
working
on
CSI.
We
have
a
very
solid
draft
that
we
are
going
to
publish
very
soon,
hopefully
at
the
end
of
this
week
or
early
next
week.
A
A
Look
like
not
with
us
today,
but
I
can
probably
give
a
quick
update
on
her
behalf.
She
was
working
on
a
Alfa
prototype
for
this
last
time
we
spoke.
We
still
have
a
few
weeks
left
for
four
until
code.
Freeze,
so
I
think
we're
looking
we're
looking
good.
There
I
believe
she
was
getting
the
API
changes
reviewed
for
this,
so
I'm
going
to
say
we're.
Okay,
there,
local,
persistent
storage,
support,
Michelle,
I'm,
not
sure
she's.
Here.
A
G
I'm
up
I
just
was
muted
yeah,
so
we're
still
making
progress
on
the
testing.
I
had
I
got
a
new
laptop
and
then
there
was
a
network
issue
with
the
laptop
so
that
the
ladies
of
crime,
but
I'm
the
last
tonight
back
up
at
full
speed.
John
is
on
paternity
leave
for
a
week
or
so
or
two
I'm,
not
sure
exactly
how
long.
G
A
F
A
H
A
I
Can
we
can
we
just
get
a
prototype,
so
this
has
worked
in
other
things
too,
of
jism
getting
a
prototype
of
code.
You
know
you
don't
even
have
to
check
it
in,
but
I
don't
know
who's
doing
the
development
but
I'm
having
someone
just
work
in
a
github
or
even
a
shared
github,
where
you
know,
people
just
start
throwing
code
together
and
they
can
go.
Look
at
it.
Yeah.
J
A
H
H
I
I
mean
I
also,
it
sounds
like
it's
in
good
state
for
alpha
and
also
it
also
feels
like
both
the
process
of
how
this
worked
crossed
continent
cross
companies-
it's
been,
you
know,
sort
of
a
fundamental
teacher.
I
almost
feel
like
it's
worth,
presenting
to
the
sig,
not
just
how
snapshots
work,
but
how
we,
the
state,
works
together
to
do
it.
One
thing
that
I
think
the
sig
may
take
for
granted
is
I,
know,
I,
say
it
a
lot
and
I'm
really
sorry,
because
I
work
with
so
many
FIGS.
I
This
is
a
very
healthy,
fig
and
I
know.
It
took
a
low
barrier
amount
of
time,
snap
shots
where
it
is,
but
you
know
it's
always
been.
Making
steady
progress
and
I
feel,
like
many
folks
have
been
included
and
I
know
that
other
things
are
literally
wrestling
with
how
to
do
this,
not
literally
laughing
figuratively
wrestling
with
how
to
do
this
and
I.
Think
presenting
it
to
the
community
forum
would
be
a
really
great
idea
if
someone
could
volunteer
to
do
that
either
like
jingler
Tomas
or
oh.
F
I
spoke
with
surgical
that
and
another
proposal
that
I've
added,
which
is
kind
of
a
small,
a
part
of
that
that
was
all
that
is
like
desizing
PVC
release,
resizing
support
suzy
and
it's
basically
one
part
of
it
like
that
was
all
talks
about
changing
stories,
classes
and
whatnot.
But
so
it's
like
I
think
for
now
we'll
just
focus
on
three
sides
and
then
we'll
go
from
there,
but
I'm
changing
everything.
So
that
is
the
plot.
Okay,.
F
H
A
M
A
A
So
I'm
going
to
go
ahead
and
mark
this
started
and
I
think
those
were
the
two
big
things
that
they
were
looking
for.
If
we
could
get
to
those
this
milestone,
that
would
be
great.
If
not,
we
can
push
it
out
to
the
next
milestone.
I
think
there
were
fairly
large
changes
and
require
changes
through
the
core,
which
you
always
need
to
do
very
carefully,
so
take
a
first
pass
at
it
and
pull
me
or
Jing
or
someone
else
in.
If
you
need
another
set
of
eyes.
A
I
A
That
makes
sense
we're
going
to
go
ahead
and
mark
this
has
not
started
and
will
attract
it
again
for
the
phone
milestone
next
up,
CC
capacity,
usage
stats.
This
was
a
pet
project.
You're
doing
I,
don't
believe
she
had
a
chance
to
work
on
this.
Yet
she's
been
focused
mostly
on
the
local
storage
stuff.
We're
going
to
market
had
not
started,
and
if
she
gets
the
chance
to
work
on
it,
this
milestone
will
be
great,
otherwise,
no
big
deal
provisioning
configuration
in
PVC
yan.
C
A
C
C
N
C
A
A
A
Improving
the
chroot
hack-
this
is
something
that
Jing
on
our
side
was
going
to
focus
on
I'm,
not
sure
what
the
status
of
this
is
currently
in
will
sync
up
with
Jing.
Once
you
get
back
fixing
broken
volume
reconstruction,
so
this
is
an
important
bug
fix
that
Chakri
was
working
on
Chakri.
Do
you
want
to
give
an
update
on
this
yeah.
C
A
O
A
I'll
see
if
me
Jing
or
Michelle,
can
take
another
look
at
it
and
we'll
go
from
there
yeah
correct
appreciate
it
thanks,
thank
you
and
last
certainly
not
least
Apple
for
resizing
TVs.
We
kind
of
talked
about
this
in
conjunction
with
the
dynamic
update
of
PvP
PC,
so
I,
don't
think,
there's
anything
else
that
needs
to
be
added
here.
Yep.
A
Alright,
thanks
for
bearing
with
that
everyone
looks
like
the
big
features
are
all
on
track.
Some
of
the
minor
features
are
may
slip.
The
only
thing
that
I'm
fairly
concerned
about
is
the
longing
the
fixing
broken
volume,
reconstruction,
others,
man
I,
think
we're
fairly
antrax
all
right.
Let's
go
back
to
the
agenda
and
see
what's
next
PRS
to
digs
discuss
four
to
four
five
to
eight
six
who
put
this
on
there.
K
See
I
know
we
had
a
this
is
a
you
know,
pretty
long,
er
there's
a
couple
spots.
No
matter,
we've
had
a
pretty
long
discussion,
I
think
we're
pretty
good
on
it.
There's
a
couple
of
spots
that
I
we're
going
to
have
to
get
deeper
in
the
source
tree
to
get
merged
with
the
stuff
in
key
and
B
all
right,
but
I
just
want
to
call
it
out
because
we're
ready
for
merge-
and
this
was
something
that
urgent
for
us
cool.
A
N
What
do
you
need
hi?
This
is
my
first
attending
the
six
tours,
hopefully
I'll
be
coming
more
often
and
also
happy
to
take
notes
by
the
way
too
excited.
That's
that's
helpful,
so
cool,
so
I
kind
of
got
thrown
into
this
one
and
it
I
did
because
obviously
this
is
now
part
of
Microsoft,
and
so
this
is
what
a
future
this
scene
is
pretty
critical
for
the
one
to
the
milestone
from
the
Azure
crew.
N
So
I'm
not
sure
exactly
what
so,
why
it's
stalled
out,
but
it's
something
that
definitely
needs
some
attention
and
I
would
welcome
any
help
from
anybody
that
can
help
me
figure
out
what
needs
to
be
done
here,
dander,
the
sward
name.
Could
you
summarize
what
the
changes
here
are?
If
there's
a
lot
it's
really
about
getting
the
blob
storage
compatibility
in
place
and
with
the
azure
managed
disks.
Essentially
the
way
nazzer
storage
rolls
out,
and
it's
also
some
bug
fixes
I've
gone
over
a
little
bit.
I
mean
I'm
more
from
the
project
management
side.
N
A
M
I
reviewed
it
in
the
flow
path
on
the
table,
so
that's
a
lot
of
refactoring
actually
on
the
things
in
castle.
Car
provider
and
I
do
disk
to
be
restructured
and
it's
a
breaking
a
lot
of
things
so
and
I
don't
got
a
chance
to
see
the
cat
patter
have
not
consented
on
it.
This
can
be
fixed
and
the
shtetl.
We
can
understand
how
cursor
chase
of
the
liabilities.
So
far
we
haven't
corrected
yeah.
E
M
So
learning
so
the
review-
and
he
also
there
are
lots
of
suggestions.
Hopefully
they
are
actually
they
are
around.
This
issue's
edition
from
panics
can
be
communicated
the
around
some
PR
Australia
and
it
has
still
being
worked
on
so
assume.
This
is
some
something
they
have
the
internal
communication:
okay,
I.
M
I
think
this
is
a
relatively
small
change,
but
why
is
happened
in
such
a
large
scale?
That
is
something
I
haven't
got
a
chance
to
look
at
okay.
So
you
essentially
just
add
a
new
API
I
got
time,
I
read
about
it.
Imagine,
but
not
so
it's
result
in
the
huge
as
we're
messing
this
back
to
him
little
bit
up
to
get
introduction
to
a
smaller
scale,
so
we
can
manatees.
We
do
yeah.
D
M
D
Jason,
just
just
an
observation,
I
think
want
to
say
a
couple
things
one.
It
is.
This
is
a
huge
change,
huge
changes
always
but
slowly
to
to
get
merged
because
we're
always
a
bit
constrained
with
Revere
bandwidth.
So
it's
always
faster
to
get
sort
of
comic.
You
know
multiple
smaller
changes
in
and
then
second
I
think
there's
a
bunch
of
unresolved
question.
D
N
A
A
That
is,
has
not
been
addressed
and
looks
like
some
of
the
tests
are
failing,
and
this
key
of
this
refactor
is
fairly
large.
So
there
are
some
open
questions.
I
guess
you
could
relate
out
back
to
the
azure
team
and
hopefully
they
can
get
back
with
the
feedback.
Maybe
it
might
be
worth
setting
up
a
video
conference
or
something
with
bomb
in
and
maybe
that'll
help
move
things
along
more
quickly,
maybe
yeah.
N
That's
great
I
will
do
all
of
those
things
and
I
sincerely
appreciate
your
help
and
also
Quadra
dick
you
for
looking
at
that.
That's
great
thank
you
and
welcome
to
the
sig
Jason
thanks,
I'm
really
excited
I
love
and,
like
somebody
said
earlier,
this
is
a
super
successful
Pig,
so
I
want
to
learn
things
and
bring
it
to
cluster
ops,
because
that's
the
guy
learned
sorry
so.
I
Jason,
if
you
want
I
happen
to
be
dealing
with
five
SIG's,
a
bunch
and
I
can
give
you
lots
of
sort
of
case
stories
and
things
that
have
been
helping
other
people
or
not.
So
if
you
wanted
to
sort
of
reach
out
and
talk
about
it
feel
free
I
know
some
people
and
other
things
have
been
doing
that
and
they
found
it
useful
yeah.
N
N
A
C
P
P
P
I
She's
she's,
wonderful
right
now
and
she's
sort
of
I
think
she's
leading
the
effort
for
the
test.
Stuff,
cuz
I
got
her
to
say,
she's,
leading
a
weekly
meeting,
we're
in
an
effort
to
improve
the
testing
for
storage.
So
you
know
writing
the
test
is
great
keeping
her
in
the
loop
to
make
sure
that
it's
part
of
the
system,
they're
putting
together,
be
really
helpful.
L
I
I
I
can't
tell
if
you
like,
have
this
sort
of
sense
sound
in
your
voice?
Like
don't
be
sad,
you
know
they
are.
One
of
the
things
is
that
we
don't
have
a
lot
of
test
coverage,
so
it's
kind
of
cool
that
you're
reviewing
that
here's,
an
area
where
we
need
to
have
more.
You
know,
test
coverage
for
this
area.
So
you
know
it's
cool
man.
I,
don't
find
a
quick
jumped
on
it,
you're
doing
Illya.
You.
P
P
A
A
G
A
G
This
is
just
a
minor
one,
there's
another
one
coming
that
I
was
right
in
the
middle
of
with
my
laptop
hard
drive
failed,
but
the
irony
storages,
aha
and
then
the
new
laptop
had
an
it
hardware
issue
with
the
NIC.
So
it's
been
frustrating,
but
this
is
a
simple
one:
I
need
to
rebase
that
I
mean
that
I
just
done
that
this
morning
and
I'll
push
that
up
so
mostly
saw
it.
It's
just
on
to
get
the
asite.
Take
a
quick
look.
There's
been
a
couple
of
comments.
G
G
L
L
Q
So
this
is
artful
cord
work,
so
this
basically
removes
requirement
of
running
port
works
on
the
API
server
node.
We
really
have
a
gke
that
can
be
a
problem
for
us,
because
people
can't
run
stuff
on
the
API
server,
not
because
it's
hosted
one.
So
this
removes
it.
It's
not
a
bug
fix
its
kind
of
an
enhancement,
but
it's
critical
for
us
because
we
can't
run
on
gke.
If
you
don't
get
this
money.
A
A
C
I
R
Oh,
my
god,
I
think
that
just
doesn't
work
in
a
specific
case
in
the
season,
both
in
other
areas,
but
there
vx
is
required
to
run
in
the
master
and
the
master
is
listed
in
your
handing
that
I
would
not
call
it
as
an
enhancement
of
the
feature.
It
is
actually
something
similar
to
a
bug
picture.
A
Andreessen.
I
A
I
I
A
I
A
C
A
Could
do
a
separate
call
for
it?
How
about
we
kind
of
like
we
do
for
snapshots
and
other
designs?
It
sounds
like
this
okay
review,
so
how
about
yawn?
You
set
up
a
meeting
with
to
send
it
out
to
the
storage
sig
and
see
who's
interested
in
joining,
and
maybe
do
it
in
this
week
or
early
next
week.
Okay,.
K
I
I
I
Yes,
yes
and
again:
apologies
to
it.
I
may
not
it's
always
hard
for
me
to
come
to
this
SIG
now,
because
I
have
to
drop
my
kids
off.
So
if
you
know
I'm
like
Oh,
move
this
to
the
front
and
then
I'm
not
here,
it's
not
because
I'm
not
interested
in
it.
It's
just
because
I've
got
other
family
obligations
that
I
can't
change.
I.
Think.
I
On
that
beautiful
note,
hey
everyone!
Oh
its
deal
on
the
call!
Oh
yes,
I'm
here
Steve,
can
you
send
email
I'd
like
to
just
sort
of
chat
and
talk
to
you
about
a
few
things?
If
it's
okay,
nothing,
urgent,
but
I,
miss
you
so
I
didn't
know
how
to
get
in
contact
with
you
on
slack.
If
you
can
send
me
now-
and
we
could
do
that
now
to
be
be
okay,.