►
From YouTube: October 2018 OpenZFS Leadership Meeting
Description
We discussed status of porting features to all the platforms.
Agenda and meeting notes: https://docs.google.com/document/d/1w2jv2XVYFmBVvG1EGf-9A5HBVsjAYoLIFZAnWHhV-BM/edit?ts=5bb3b66c
A
That
this,
the
point
of
this
meeting
is
to
to
carries
through
some
of
the
conversations
that
we
have
at
the
developer
summit
throughout
the
year
and
make
sure
that
we're
continuing
to
have
you
know
we're
conversations
that
are,
and
you
know
at
a
high
level,
not
just
about
particular
bugs
or
pull
requests
like
we
do
typically
nowadays.
So
the
first
thing
I
wanted
that
we
want
to
tackle
is
how
to
make
sure
that
all
the
features
in
ZFS
are
available
on
all
the
different
platforms.
A
A
It'll
probably
take
most
of
the
time
today,
but
if
you
have
ideas
for
future
agenda
items
will
have
the
opportunity
to
add
those
at
the
end
of
the
meeting
and
I'll
try
to
leave
five
to
ten
minutes
for
kind
of
an
open
discussion
of
other
other
topics.
At
the
end
of
the
meeting.
I
know:
30
minutes
is
probably
not
going
to
be
long
enough
to
actually
cover
everything
that
we
want
to
today
and
I.
A
Think
that's,
okay,
we'll
have
this
meeting
again
in
four
weeks,
so
we
can
pick
it
up
there
I
know
like
everybody's
time
is
precious
and
I
don't
want
to
you
know:
I,
don't
want
to
slip
an
hour-long
meeting
that
we
then
only
spend
20
minutes
in
each
time.
I'd
rather
have
the
consistency
of
having
folks
show
up,
not
just
the
you
know,
great
turnout
that
we
have
today
but
show
future
meetings
as
well
and
can
continue
this
conversation.
A
Okay,
so
dive
right
into
it.
I
have
the
spreadsheet
here
and
I
thought
how
the
spreadsheet
we
have
about
some
recent
features
that
have
been
integrated
into
all
the
platforms,
so
great
job
everybody.
It's
available.
All
these
are
available
on
and
Lumos
linux,
freebsd,
OS
10,
and
when-
and
the
next
section
is
the
one
that
I
want
to
discuss
today,
which
is
features
that
have
been
integrated
to
one
platform
but
not
yet
all
them
and
the
the
goal
I
don't
want
to
get
too
deep
into
the
technical
nitty-gritty.
A
A
So
the
first
one
on
here
is
channel
programs,
good
news,
it's
on
the
three
most
used
platforms
of
the
most
Linux
and
FreeBSD,
and
on
oh
it's
10
and
windows.
Your
again,
you
uragan
is
kind
of
the
one
man
as
far
as
I
know,
you're
gonna.
Let's
do
it!
You
want
to
correct
me
almost
the
one
and
the
almost
the
what
soul
engineer
working
on
these
two
platforms,
so
do
great
job.
If
you
guys
don't
know
about
urines
work
on
this
I,
you
should
go
check
out
the
his
talk
from
last
year.
A
C
A
A
D
So
this
so
Brian
and
I
are
kind
of
discussing
a
follow
on.
We
want
to
just
make
sure
that
the
following
changes
are
kind
of
at
a
good
state
before
we
integrate
the
initial
PR,
so
the
follow-on
changes
will
be
also
applicable
to
all
to
all
platforms,
but
effectively
their
their
performance
enhancements
to
the
initialized
code.
So
the
plan
will
be
to
integrate
the
initial
PR
to
kind
of
be
the
same
with
a
follow-on
PR
coming
shortly.
After
that,
we
can
then
distribute
across
all
the
platforms
and
then
are
you.
A
D
I
have
a
work-in-progress
PR
that
Brian
has
done
some
initial
performance
testing.
It
looks
like
it's
pretty
good.
I
need
to
go.
Do
some
cleanup
on
it,
because
we
wanted
to
just
kind
of
create
a
baseline,
I.
Think
there's
one
more
thing
that
I
want
to
investigate
to
see
if
we
can
make
a
further
improvement,
but
it
we're
pretty
close
to
like
having
it
to
a
point
where
we
can
start
doing
reviews
on
it:
okay,
cool,
oh.
A
A
But
my
understanding
Jurgen
is
that
you,
you
don't
really
have
the
the
timer
or
inclination
to
debug
on
all
the
issues
that
have
been
reported
so
I've
been
I've,
been
helping
out
on
that
a
bit
Tom's
been
helping
out
on
that
a
bit,
but
I
think
we're
we're
both
kind
of
getting
a
little
bit.
We
put
in
a
lot
of
effort
and
we'd
like
to
see
us
come
to
completion,
but
we
could
definitely
use
some
more
help
with
this.
So
is
anybody
else.
Has
anybody
else
started
working
on
this.
A
E
Looking
at
the,
we
received
a
crash
jump
from
a
camera
book
who
exactly
yes
and
we've
looked
at
that
and
I
think
hit
kind
of
a
dead
end,
because
the
octopi
just
to
not
appear
in
the
dump,
and
we
really
wanted
to
be
able
to
see
yeah
I
think
it's
basically
as
I
recall,
I
talked
to
Alex
about
this.
Yesterday
as
I
recall,
it's
failing
a
H,
Mac
validation,
something
that
we
don't
have
in
the
dump
to
have
a
closer
look
at.
E
So
one
thing
we
can
try
and
do
is
reproduce
it
alex
has
tried
to
trivially
to
reproduce
it.
We
have
a
ministry
that
yep.
The
other
thing
is.
We
can
maybe
go
back
to
go
know
and
get
him
to
do
a
dump
with
our
contents
included,
which
would
be
bigger,
but
at
least
would
probably
give
us
something
closer
to
look
at
so
we'll
try
and
follow
up
on
that.
Okay,
yeah.
A
His
own
bits,
or
is
it
yeah
so
he's
running
on
smart
OS?
Oh
he's
he's
done
a
port
or,
like
you
know,
merge
of
the
pull
request
into
smart,
OS,
I'll
kind
of
do
and
he's
done
the
build
himself.
He
doesn't
seem
from
from
what
I've
talked
with
him
only
very
little
bit.
It
seems
like
he
is
not
like
a
developer
per
se,
but
you
know
he.
He
knows
how
to
you
know,
merge
the
code
and
do
builds
and
stuff
like
that.
A
Okay
and
on
the
particular
issue
that
you're
mentioning
I'm
not
wanting
to
get
the
user
data
I
had
I
found
it
difficult
to
me
to
get
all
the
arc.
Data
included
in
the
dump.
I
need
a
way
that
works
right,
but
what
I
did
when
I
was
trying
to
run
down?
Something
else
was
I
just
to
change
the
code
to
add
something
that
was
like,
oh,
like
when
I'm
getting
this
failure.
Just
like
him
out,
you
know
Kim
and
Malik
somebody
buffer
copy.
A
A
E
A
E
A
That
wouldn't
surprise
me,
I
mean
he's
also
using
Z
vowels,
which,
like
I,
couldn't
see
there,
which
are
a
little
bit
special.
Only
because
the
like
a
lot
of
the
problems
that
he's
run
into
seem
like
it's
with
the
hmx
allegation
of
the
denote
block,
like
the
block
of
denotes,
which
is
which
is
handled
specially
by
encryption
and
by
kerger,
send
receive
and
in
with
Z
balls,
so
with
with
nods
evolves
with
file
systems.
A
Basically,
every
block
of
Dino's
has
bonus
buffers
which
cause
us
to
have
to
like
decrypt
stuff
earlier
in
using
a
different
code
path,
but
with
Z
balls
there's
no
bonus
buffers.
We
just
don't
happen
to
use
that
with
Z
balls,
and
so
the
do.
Katrina
and
Mac
validation
happens
later
and
it's
different
in
a
different
code
path,
and
so
it's
possible
that
that's
related
to
the
problem
all
right.
Well,
we
will
take.
G
H
F
According
to
Sean,
it's
working
for
him,
but
I'm
not
sure
how
much
sort
of
budget
taste
in
heat,
so
several
people
tried
it,
but,
along
with
rain,
meet
for
visiting
serious
for
plot
serious
production,
okay,
so
maybe
Steve
here
freeze
is
definitely
one
of
stopper
to
start
and
greatly
some
things
like
encryption,
algorithm
and
so
yeah,
but
most
astronomers
now
working
on
different
things.
Considering
that
part
is
done
just
waiting
for
integration.
F
A
Cool
and
encryption
is
kind
of
a
big.
It's
probably
the
biggest
project
on
here
and
probably
has
the
most
work
to
get
it
completed.
Does
anybody
else
have
kind
of
more
questions
or
comments?
I
mean
one
thing
that
one
thing
that
is
a
little
bit
concerning
to
me
is
that
you
know
we.
We
have
basically
one
person,
that's
been
reporting
bugs
on
Lumos,
my
encryption
bugs
on
on
the
bus
for
the
past,
like
six
months
or
so,
and
we
haven't
heard
like
similar
bugs
reported
on
Linux
or
in
the
freebsd
port.
H
G
Both
code
paths
or
liking
in
both
code
bases
but
I,
think
Linux
stresses
different
ones
differently
and
then
illumos
does
particularly
like
you
were
saying,
with
like
the
order
that
things
get
decrypted
and
things
like
that
I've
seen
before,
where
you
know
things
seem
to
exist
in
both
places,
but
they
just
never
happen.
They
never
seem
to
show
up
on
Linux
for
some
reason
or
another.
A
Don't
think
so
the
the
person
who's
been
testing
on
the
Lumos
is
using
smart
OS
tonight
it
looks
like
I'm
not
super
familiar
with
smart,
OS
and
kind
of
how
is
typically
used,
but
like
he
had
a
bunch
of
Z
balls
that
were
being
managed
by
some
higher
level
of
software.
So
it's
possible
that
somebody
else
doing
something
similar
with
smart
OS
would
exercise
it
in
a
similar
way.
A
Yeah,
definitely,
if
folks
do
have
tools
that
they're
using
for
testing
share
them.
I
suspect
that
a
lot
of
people
just
are
doing
their
particular
workload
and
not
like
actually
trying
to
test
an
encryption
very
much
they're,
just
like
turn
on
encryption
and
keep
doing
what
I've
been
doing
and
see
if
it
blows
up,
I've
been
in
personally,
like
I,
haven't,
really
tested
it
aside
from
running
that
running
the
test
suite
and
then
looking
at
the
dumps,
I've
gotten
from
other
folks.
A
Thank
cool,
so,
let's
move
on
to
the
next
item
was
multi
map
protection
or
multi
import
protection
as
its
called,
so
this
is
available
on
Linux
and
no
other
platforms.
I
will
bring
up
that.
You
know
if
we
just
if
we,
if
there's
something
you
that's
not
actually
applicable
to
other
platforms
or
people,
decide
yeah,
that's
cool,
but
it's
not
really
worth
it
for
us
to.
For,
like
that's,
okay,
I'm
I,
don't
think
that
you
know
we
sure
like
it
would
be
great
to
have
everything
out
of
your
platform,
but
we
are
balancing.
A
You
know
the
effort
that
we're
spending
with
the
benefit
from
it
and
with
encryption
like
the
benefit,
is
really
huge.
You
know
everybody
wants
the
encryption
on
their
platforms,
so
I
think
it's
worth
it
a
multi
mount
protection
I,
like
seems
like
you'd,
be
useful
on
every
platform.
I
know
you
know:
George
Wilson
I
know
you
debugged
a
lot
of
concurrent
pool
import
on
multiple
systems
that
caused
encryption
at
that
back,
caused
that
caused
corruption,
and
this
would
hopefully
have
protected
us
against
that.
A
A
H
A
I'm
gonna
go
ahead
and
mark
this
one
at
least
as
yellow.
Even
though
you
don't
have
it's
not
in
progress,
but
we
do
have
an
owner
who
looks
committed
to
working
on
it,
Icom
how
about
on
Bill
Lumos
we
at
del
phix.
We
don't
have
this
problem,
we
don't
have.
You
know
the
systems
aren't
that
we
ship
are
not
configured
to
be
able
to
import
the
same
pool
concurrently.
So
this
is
not
really
a
business
need
for
us.
It's
also
true
for
giant.
We
don't
really
do
network
storage
either.
A
It
sounds
like
those
are
going
to
potentially
different
paths
to
getting
this
in.
One
is
like
because
FreeBSD
and
the
most
ZFS
code
is
so
close.
You
could
wait
for
the
to
be
ported
to
FreeBSD
and
then
hopefully
that
patch
applies
and
more
cleanly
to
the
most
and
the
other
is
that's.
Some
kind
of
pull
sail
pulling
in
of
all
the
dips,
with
the
limo
with
Linux,
which
seems
like
a
big
lift,
I
think.
E
It
is
probably
gonna
be
a
big
lift,
but
I
think
we
in
in
the
in
the
limit
for
us
I
think
it
makes
sense
to
first
to
set
up
machinery
and
process
on
our
end,
to
start
to
at
least
assess
the
the
actual
size
of
the
lift
as
a
bulk
import
operation.
I'm,
not
saying
that
we're
gonna
just
import
everything
but
like
we,
we
haven't
really
tried
yet
some
like.
Maybe
it
won't
be
that
bad.
H
Considered
something
similar
for
FreeBSD
like
when
I
was
looking
at
pulling
over
some
code
runs
NFS
on
Linux
for
identifying
V
dev
names
in
the
zpool
Iost
at
command
the
amount
of
divergence
between
our
code
base
and
the
ZFS
on
Linux.
One
was
so
big.
It's
like
you
know,
pulling
things
out
of
order
seems
like
it's
gonna,
be
a
lot
more
work
overtime
than
trying
to
get
to
some
snapshot
and
then
work
forward
from
there.
Mm-Hmm
yeah
I
think
that.
E
If
you
consider
a
like
a
like
what
is
released
as
0
at
0
or
whatever
is
like
a
it's
a
fixed
point
in
time
that
we
can
go
and
stop
to
look
at
how
to
get
to
there
specifically
might
be
a
little
more
manageable.
But
again
we
haven't
haven't
started
that
process.
Yet
we've
got
some
a
few
other
in-house
bugs
to
get
to
and
then
walkin
I
try
and
set
up
some
testing
and
moach
infrastructure
to
see
how
that
goes.
A
C
Do
you
envision
just
kind
of
leaving
these
changes?
This
is
about
the
M&P.
We
were
talking
about
yeah,
yeah,
I,
think
I,
already
sort
of
started
on
putting
in
commits
I'm
set
up
to
color
in
commits
both
from
illumise
and
Linux
with
a
issue,
but
I
think
you
needed
parallel
mounts
first.
Does
that
sound
right.
D
D
C
A
C
A
F
E
F
E
G
One
one
thing
that
I'd
want
to
make
sure
because
I
think
this
did
kind
of
happen
on
FreeBSD
a
little
bit
so
I
believe
there
were
three
different
things
which
were
in
different
operating
systems
at
different
times,
because
the
original
code
came
from
Sasha
who
worked
on
it
and
then
I
took
that
ported
it
to
Linux,
and
while
I
was
working
on
it,
it
got
merged
into
FreeBSD
before
it
was
actually
merged
into
Linux,
which
is
where
I
did
all
the
work.
Then
it
was
eventually
completely
merged
into
Linux.
G
So
I
just
want
to
make
sure
that
we
have.
You
know
that
we
have
a
consistent
implementation
everywhere.
They
they're
not
like
all
drastically
different
that,
like
I
said
they
were
all
based
on
the
same
thing,
but
I
just
want
to
make
sure
that
everybody
has.
You
know
the
latest
version
of
it.
That's
all.
E
G
Sounds
correct,
the
only
thing
is
again,
it
could
just
be
the
like.
That
was
also
the
name
of
the
earlier
version
that
did
get
ported
to
freebie
FreeBSD
I'd.
Imagine
that
that
probably
I
think
they're
all
up
to
date,
but
I
just
wanna
make
sure
like
the
be
that
everything
that
writing
correctly.
It's.
D
G
H
J
A
E
F
E
A
G
I
have
one
thing
that
I'd
like
to
cover
I.
Think
a
lot
of
us,
especially
since
many
of
us
really
want
to
kind
of
look
at
0,
804
ZFS
as
or
I'm.
Sorry
for
ZFS
on
Linux
is
kind
of,
like
you
know,
kind
of
a
settling
point
where
we
can.
You
know,
get
a
lot
of
things
in
sync,
and
things
like
that
I
think.
G
A
A
A
All
right:
well,
then,
I'll
see
you
all
four
weeks
from
today,
four
weeks,
minus
34
minutes
from
today,
and
also
no
reminders,
if
you,
if
you'd
like
to
add
your
name
to
the
attendee
list
and
also
maybe
include
your
email.
If
you
don't
know
that
I
already
have
it
I
will
add
you
to
meeting
invite,
so
you
can
get
that
and
also
try
some
nut
emails
and
tweets
and
stuff
so
that
folks
are
reminded
next
month.