►
From YouTube: May 28, 2019 OpenZFS Leadership Meeting
Description
We discussed illumos encryption; RAIDZ expansion; BSDCAN; DRAID; DevSummit; Compressed ARC.
Agenda and meeting notes: https://docs.google.com/document/d/1w2jv2XVYFmBVvG1EGf-9A5HBVsjAYoLIFZAnWHhV-BM
A
We
have
mostly
status
updates,
but
there's
a
lot.
A
lot
has
happened
in
the
past
month,
so
I'm
just
wanted
to
keep
everyone
on
the
same
page
about
all
the
interesting
a
work
that's
been
going
on.
So
first
off
new
ello
died.
It
was
released.
So
congratulations
to
Brian
and
everyone
else
involved
in
these
Brian.
Do
you
want
to
say
anything
about
that.
B
B
A
A
A
D
This
is
Jerry
I
I
have
all
of
your
guns.
Work
pulled
over
and
built
and
I'm
running
it
right
now.
There's
a
couple
of
issues
with
some
ZFS
test
failures
that
I'm
looking
at
and
as
soon
as
I
could
get
those
resolved.
I
would
be
putting
it
out
for
the
illumos
car
review
and
then
you
know
trying
to
get
that
pushed
along
and
integrate
it.
So
I
think
we're
pretty
close
to
being
being
there
on
that
cool.
A
Cool
the
next
SAS
item
is
mine,
which
is
raids.
The
expansion
I
have
been
actually
working
on
implementing
raids,
the
expansion
again
after
a
long
break,
so
I
just
sort
of.
Let
folks
know
that
I
will
have
a
alpha
test
out.
Probably
this
week
or
next
week,
that'll
be
against
ZFS
on
Linux
and
it
will
sort
of
kind
of
work.
It's
still
alpha.
A
It's
still
don't
use
this
on
you
on
any
pools
that
you
care
about,
because
you're
gonna
have
to
destroy
them
later,
because
underscore
mana
is
gonna
change,
but
it
it
does,
let
you
add
a
device
to
a
raid,
Z
group
and
then
it
reflow
as
all
the
data
across
it
in
the
background,
and
then
you
have
more
space.
So
there's
a
bunch
of
caveats
which
will
be
in
the
in
dolefully
the
in
the
blow
requests
it's
still
very
early,
but
I
would
love
to
get
help
testing
a
fitting.
A
A
Cool
next
I
just
want
to
highlight
a
couple
of
talks
that
have
been
at
BSD.
Can
Alan
started
me
on
the
spot?
I
didn't
ask
you
all
this
beforehand,
but
I
was
wondering
if
you
could
just
kind
of
briefly
talk
about
the
the
talk
that
you
gave
about
the
future
of
DBS,
and
you
know
maybe
update
on
you
know
that
that
work
and
how
it's
being
received
in
the
freebsd
community
yeah.
E
A
You
had
in
mind
no
that
sounds
good
to
me.
I
know
one
of
the
one
of
the
things
that's
caused.
Some
contention
in
the
previous
key
community
is
just
the
l
word
being
used
and
associated
with
this.
Well
we're
being
Linux
and
I
think
to
a
large
degree,
it
seemed
like
that
has
been
addressed
through
education.
I
was
really
pleasantly
surprised
to
see
some
of
the
freebsd
internal
discussions
where
folks
were
kind
of
raising
that
concern
and
complaining.
A
You
know
the
Linux
kernel
development
with
the
FS
on
Linux
and
I
was
pleased
to
see
that
a
bunch
of
people
who
are
not
Alan
are
I.
Speaking
up
in
you
know,
kind
of
clarifying
that
it's
different
books
involved
and
that
you
know
the
folks
working
on
ZFS
on
Linux
and
FreeBSD
all
have
kind
of
the
same
goals
and
same
you
know,
code
quality,
sensibilities
and
you
know
we're
all
we're
all
working
towards
the
same
goals.
So
I
was
pleased
to
see
that
yeah
one.
E
Of
the
other
threats
was
just
differentiating
it
from
other
efforts
like
we
currently
in
freebsd.
We
pull
in
the
drm
graphics
drivers
from
linux,
and
we
use
a
Linux
kernel
compatibility
layer
to
be
able
to
have
those
drivers
run
on
FreeBSD
and
I
think
when
they
hear
ZFS
on
Linux.
They
assume
that
it
would
end
up
using
something
like
that,
and
that
would
be
just
more
linux
e-code
infesting
in
freebsd,
but
in
fact
it
all.
A
And
I
think
we,
as
you
discussed
previously
this
meeting
a
is
super
cool
and
I.
Think
it's
like
a
it's
super
cool
that
we're
creating
one
repo
that
you
know
will
have
the
ZFS
bits
for
both
for
two
different
operating
systems
and
that
you
know
you'll
be
able
to
test
both
of
those
with
pull
requests
with
one
pull
request
and
I.
Think
we've
discussed
as
meeting
previously.
A
You
know,
after
the
work
is
done
rather
than
while
the
work
is
kind
of
still
in
progress.
So
just
wanna,
let
folks
know
like
at
least
for
me
and
I.
Think
I
mentioned
this
briefly
with
Brian
like
it
is
still
in
our
radars
to
consider
doing
something
like
that.
But
you
know
we
we
might
take
action
after
this
is
done.
B
E
E
A
Cool
then
I
also
gave
a
talk
at
BST.
Can
it
was
about
how
snapshots,
work
and
scalability
problems
and
solutions
on
there?
A
This
is
what
kind
of
rehash
rehash
the
content
here,
but
if
that's
something
you're
interested
in
or
you
know
other
folks,
you
point
them
to
the
the
slides
are
posted
in
the
video
will
be
eventually
I
guess
my
take
it
might
take,
it
seems
like
usually
takes
a
few
weeks
or
months
for
that
to
happen.
It.
A
Cool
anything
else
from
abuse,
Deacon
Islander
anyone
else
who
is
there.
F
F
She
seems
like
a
year
ago,
but
I
think
it's
like
two
weeks
ago,
or
so
thanks
to
del
phix
for
hosting
it
at
their
San
Francisco
office,
and
we
had
about
eleven
participants
thanks
Matt
for
creating
link
there
to
the
meeting
recap.
H
H
So
we've
got
it
all
just
completely
done
before
that
event,
and
that's
about
it
and,
in
the
meantime,
any
questions
that
are
coming
up
or
kind
of
trying
to
keep
track
of
them
in
a
separate
dock,
to
make
sure
that
we
refer
back
to
those
over
time.
But
hopefully
people
will
do
that
for
themselves
that
they,
they
think
of
a
comment
that
they
want
to
make
sure
it
gets
in
there
that
they
reach
out
to
us
during
that
period.
They
don't
see
that
change
questions,
comments.
A
Thanks
Karen
I,
realized,
I,
don't
think.
We've
actually
announced
the
developer
summit
in
this
meeting
so
developer
summit.
We
have
a
date.
It's
gonna
be
November,
4th
and
5th,
which
is
a
little
bit
later
than
it
hasn't
been
in
the
past
mm
that
date
just
kind
of
worked
better
for
the
venue
and
also
for
avoiding
other
giant
conferences
in
San
Francisco.
So,
hopefully,
your
hotels
will
only
be
expensive
and
not
in
exam
really
expensive.
A
A
Yes,
so
the
well
for
registration
will
officially
open
July,
8th
and
then
you'll
need
to
send
me
your
presentation,
proposals
by
August
19th
and
then
we'll
have
the
conference
in
November
4th
and
5th
we're
also
looking
for
sponsors.
We
have
a
little
bit
different
sponsorship
offerings
this
year.
So
if
your,
if
your
company
is
interested
in
sponsoring,
get
in
touch
with
Karen
and
she
can
send
you
the
details,
yeah.
A
A
A
G
G
Forget
where
I
saw
it
I'm
a
visual
thinker,
so
there
is
a
have
the
URL
somewhere
there's
a
nice
URL
that
describes
how
the
file
system
is
layout,
laid
out
what
data
structures
point
to
what,
but,
with
encryption
being
increasing
logs
of
the
possible
change
to
the
de
doop
table.
All
that
stuff
is
starting
you
to
the
point
we're
having
that
updated
is
going
to
make
it
easier
to
understand
what
is
going
on.
B
G
C
C
C
A
So
I
think
the
challenge
there
is
that
that,
like
I,
don't
know
that
we
have
the
source
to
that
document.
It's
like
there's
just
a
PDF
and
I,
don't
know
what
the
license
is
associated
with
that
if
we
wanted
to
copy
and
paste
it
somewhere
else
and
then
and
then
update
it,
but
I
do
agree
that
it
would
be
great
to
have
an
updated
on
this
format
guide
for
the
like,
because
I
think
that's
you
know
it's
definitely
well.
A
J
C
I
also,
I
think
this
is
more
just
to
like
kind
of
put
the
scope
of
work
into
perspective.
I
think
it's
more
of
like
an
aggregation
and
a
beautification
kind
of
thing,
because
I
think
we
do
have
this
documentation.
It's
just
spread
out
in
that
kind
of
places.
So
I
know
like
spa,
dot
h
lists
what
an
embedded
block
pointer
is
and
what
a
normal
block
pointer
is
and
like
and
the
I
have
the
encryption
design
document,
which
is
just
like
a
google
doc.
C
G
Just
aggregation
because
it
is,
as
you
point
out,
spread
out
all
over
the
place,
which
is
why
I
thought
of
it,
because
I
just
reread
that
I
might
just
reread
the
encryption
changes
and
then
I
was
looking
at
the
space
map
log
changes
and
there
are
all
these
things
that
say
yeah.
This
is
what
this,
this
data
structure
does,
how
this
file
uses
it,
and
but
in
order
to
find
it,
you
have
to
track
it
down
into
whichever
file.
A
Yeah
I
definitely
agree
sounds
like
one
way
to
go
would
be
to
like
scrape
the
content
out
of
that
PDF
and
you
know
reformat
it
and
update
it,
and
you
know
copy
all
the
stuff
from
the
code
comments
back
into
there.
I
don't
know
what
maybe
we'll
put
out
a
call
on
the
email
list
to
see
if
we
any
volunteers
to
do
that,
but
that
would
definitely
be
useful,
especially
for
folks
new
to
ZFS
development,
yeah.
G
B
A
Totally
I
think
it
would
be
great
to
have
that
in
a
format.
You
know,
that's
not
just
a
PDF
that
we
can.
Actually
you
know
whether
it's
markdown
or
whatever
is
something
new
we
can
check
in
to
some
repo
and
then,
whenever
you,
you
know
make
it
changes
on
this
format
for
the
review
processes,
somebody
reminds
you
to
go
update
that
file
as
well.
K
G
K
C
A
A
A
E
There
was
quite
a
bit
of
movement
on
the
disabling,
the
ability
to
disable
the
compressed
arc.
Pr
issue
on
github
it
turns
out.
There's
is
actually
a
modest
performance
impact
with
compressed
arc,
mostly
having
to
do
with
the
fact
that
when
you're
reading,
when
you're
getting
a
cache
hit
from
the
compressed
arc,
you
decompress
one
block
at
a
time,
there's
no
parallelism
at
all,
and
so
you
get
bottleneck
down
to
a
single
threat.
How
fast
a
single
thread
can
decompress.
A
A
read
you
realized
I
mean
right,
it's
really
the
same.
If
you
have
a
single
threaded,
you
get
one
thread
reading
then
that
one
thread
is
the
thread
doing
the
decompression
of
the
blocks
that
it's
requesting,
as
opposed
to
in
the
like
old
way.
The
decompression
would
happen
at
the
pretended
prefetch.
So
if
you're
reading
sequentially,
then
the
prefetcher
is
going
to
go.
Kick
off
all
those
reads
in
the
background
and
those
background
reads
includes
the
decompression
versus
now.
A
The
background
reads
just
includes
the
read
from
disk
and
then
when
you
actually
do
the
demand
read
then
the
decompression
happens.
So
if
you
have
one
thread
that
a
bunch
of
stuff
that
one
thread
is
the
one
doing
the
impression,
if
you
have
a
whole
bunch
of
threads
reading
a
whole
bunch
of
different
stuff,
then
they
all
get
to
decompress.
You
know
concurrently,
so
it's
not
that
it's
see
realize
it's
just
that.
If
you're
making
cereal
requests,
we
don't
we
aren't
magically
making
them
parallelized,
as
we
did
in
the
past.
E
That's
a
much
better
description
than
what
I
was
trying
to
say
dad
it,
okay,
that
makes
more
sense
yeah.
So
maybe
we
could
have
some
kind
of
read
ahead
thread
decompressing
stuff
into
the
debuff
cache,
but
at
the
same
time
we
don't
want
to
burn
a
lot
of
CPU
doing
that
so
I
don't
know
how
to
find
the
right
balance,
but
it
seems
like
that,
won't
be
as
easy.
A
win
is,
as
we
thought.
A
E
Right
so
then
I
guess
there's
actually
two
different.
The
first
one
is.
We
thought
we
could
just
get
rid
of
the
ability
to
disable
confess'd
arc
because
it
wasn't
that
useful,
but
maybe
it
is
and
then
separately
we
probably
have
a
new
issue
to
look
at.
What
can
we
do
to
make
the
performance
of
single
thread
and
single
sort
of
reads
from
compressed
arc
better
and.
E
We
could
consider
it
I,
don't
know
how
much
value
it
has.
The
original
impetus
for
it
was
dealing
with
the
interactions
with
the
l2
arc
and
I
have
to
look
at
a
little
more
closely,
but
some
of
the
changes
Tom
made
when
he
did
the
encryption
stuff,
where
there's
now
of
what
the
l2
are
untransformed
function
and
so
on,
might
have
obviated
the
need
for
some
of
it.
Okay,
all
right
so
I
have
to
go.