►
From YouTube: April 2022 OpenZFS Leadership Meeting
Description
Agenda: OpenZFS Conference; Encryption status; BRT update
details: https://docs.google.com/document/d/1w2jv2XVYFmBVvG1EGf-9A5HBVsjAYoLIFZAnWHhV-BM/edit
A
All
right,
let's
get
started,
welcome
to
the
april
2022
opencfs
leadership.
Meeting
I
looked
at
the
agenda
nice
background
jorgen.
I
looked
at
the
agenda
and
there's
not
too
much
on
it.
So
let's
we'll
apply
plenty
of
time
for
other
q,
a
or
other
items,
but
why
don't
we
get
started
with
I'm
not
sure
who
had
this
but
somebody's
asking
about
this
pull
request
the
blake
3
hash
support.
I
don't
know
if
I
think
tino
owns
this
pr.
A
A
C
C
C
A
I
have
one
agenda
item.
I
don't
know
if
there's
really
enough
people
here
to
to
create
a
significant
poll,
but
we're
thinking
about
the
the
next
open
cfs
conference,
which
will
be
in
october
or
november
like
it
normally
is,
which
is
about
six
months
from
now
and
we're
thinking
about.
Will
it
be
in
person.
A
A
Like
you
know,
if
there
was
an
in-person
conference,
you
know,
would
you
attend,
and
I
understand,
like
everybody's,
coming
out
different
situations
in
terms
of
risk
of
covid,
I
mean
that's
kind
of
the
main
concern
here
is
like
who
knows
what
covid
will
look
like
in
six
months,
but
obviously
like
there's
a
big
expense
and
and
organization
required
from
us
to
organize
it
and
from
other
folks
to
attend.
So
you
know
we'd
like
to
do
that.
A
A
A
A
C
So
I'm
helping
organize
an
event
that
will
take
place
next
month
and
for
us
I
mean
it's
probably
also
different
between
different
countries
and
so
on.
But
the
main
thing
was
that
the
venue
was
the
venue
is
giving
us
kind
of
the
the
guidelines
and
those
guidelines
are
generally
stronger
than
the
like,
legally
required
ones.
So.
A
Right,
yeah,
we'll
definitely
check
into
that
all
right
well
looks
like
a
lot
of
folks
are,
would
like
to
attend.
Some
people
didn't
respond,
I'm
guessing.
Maybe
those
people
might
not
be
interested
in
attending
or
or
be
able
to
attend.
You
know,
regardless
of
you,
know
covet,
so
we
will
get
started
with
planning
an
in-person
event.
I
think.
D
A
A
So,
given
the
effort
involved
with
an
in-person
event,
you
know
the
pa,
the
past
conference,
I
basically
organized
I
I
you
know
I
did
almost
all
the
logistics
in
terms
of
the
what
happens
on
the
day
of
myself,
which
I
would
not
be
able
to
do
for
an
in-person
conference,
so
we
definitely
need
more
help.
We
can
hire
some
of
that
help.
A
Like
you
know,
we
have
a
contractor
victoria
who
who
has
done
the
like
sponsor
working
with
the
sponsors,
so
you
know
she
may
be
able
to
do
some
more
stuff
for
us,
but
it
would
be.
It
would
be
really
wonderful
if
someone
from
the
community
was
interested
in
doing
some
of
this
logistics
stuff
to
help
make
the
conference
happen.
A
A
So
you
know
working
with
speakers
and
and
with
a
program
committee
to
do
selection
of
talks,
but
it
would
be
great
to
have
somebody
else
help
out
with
with
more
of
the
logistics
of
running
the
conference,
and
you
know
we're
definitely
open
to
like.
I
realized
that
with
asking
for
help,
it
also
means
giving
us
some
control,
so
you
know,
I
think
that
we
we
would
like
to
have
the
conference
in
the
bay
area,
but
you
know
we're
open
to
it
being
a
different
venue
or
a
different
city.
A
A
Oh
michael
was
asking
about
karen
yeah
karen
karen
is
not
doesn't
have
as
much
time
to
spend
on
this.
No
so
she
she
has
a
different
role
at
delfix
now,
and
so
she
doesn't
really
have
a
lot
of
time
as
much
time
to
spend
on
it,
but
karen
will
definitely
kind
of
karen
and
I
will
work
with
whoever
does
step
up
to
help.
This
organizing,
to
kind
of
you
know,
pass
on
all
the
info
that
we
have
and
then
you
know
to
help
with
any
decision
making.
That's
that's
needed.
E
A
Yeah
I
mean,
I
think,
that
I
can't
I
and
the
other
folks
that
I've
you
know
worked
with
in
the
past
can
can
handle
the
kind
of
content
program
community
stuff.
But
if,
if
you
are
interested,
if
any,
if
alan
or
anyone
else
is
interested
in
volunteering
on
that
front,
you
know
more
input
would
definitely
be
welcome.
A
A
F
F
A
No,
I
think
we're
definitely
open
to
like
if
another
company,
you
know,
as
I
think,
we've
talked
about
in
the
past,
like
the
the
model
for
this
with
belfix-
is
that
it's
like
delphics
becomes
more
like
a
normal
sponsor
and
not
like.
You
know
doing
a
bunch
of
special
things
that
nobody
else
is
doing.
So
you
know
we're
definitely
open
to
another
company
like
providing
space,
or
you
know
in
in
kind
sponsorship
where
it's
like
yeah.
A
A
F
I
just
quick
update
the
linux
namespace
patch.
I
think
we've
settled
on
how
we'll
change
the
user
interface
for
it
slightly.
Instead
of
providing
the
the
numeric
number
of
the
namespace
you'll
provide
a
path
to
the
namespace,
which
is
like
slash,
proc,
slash,
some
pid.
That's
in
that
namespace
slash,
ns,
slash
user,
and
by
doing
it
that
way,
we
can
turn
that
file
descriptor
into
a
reference
to
the
namespace.
So
zfs
can
hold
that
reference.
F
The
entire
time
a
data
set
is
delegated
to
that
namespace,
so
that,
even
if
that
namespace
goes
away,
a
new
namespace
can't
be
recreated
with
the
same
id
until
that
data
set
is
unzoned,
and
that
should
solve
the
remaining
problems
there.
We
should
have
a
patch
up
for
that,
hopefully
this
week
or
next
week
to
update
that
pull
request
with
that
change,
which
I
think
will
solve
the
final
problem
that
was
left
there.
F
Yep
and
then
we've
been
experimenting
with
the
suggestions
maven
and
matt
made.
I
guess
it
was
two
months
ago
now,
I'm
changing
the
right
throttle
stuff,
looking
at
maybe
having
a
second
counter
for
the
dirty
data
that
increments,
even
before
the
data
arrives,
so
that
we
can
just
say
you
know,
a
lot
of
threads
are
trying
to
write
all
at
once.
We'll
know
that
we
need
to
apply
the
the
throttle
sooner.
A
F
G
F
Yeah,
yes,
because
it
used
the
some
flag
to
know
that
you
weren't
asking
for
any
fields
that
required
actually
opening
the
the
data
set.
So
you
could
get
it
all
from
just
the
the
moss
or
whatever
there's
one
other
one.
Oh
I'm
still
not
I'm
having
trouble
fixing
the
failing
test
on
the
right
inflation
calculation
pull
request.
F
The
zfs
receive
with
the
rest
reservation
is
supposed
to
not
receive
more
than
one
transaction
group
too
much
data
like
beyond
the
quota
or
not.
It
wasn't
a
reservation.
Maybe
it's
a
quota.
I
don't
know
something:
you're
not
supposed
to
be
able
to
receive
too
much
data
and
it's
supposed
to
be
allowed
by
one
transaction
group,
but
the
test
doesn't
actually
test
for
that.
E
H
Oh
you,
but
you
don't
send
a
receipt.
I've
been
using
it
for
quite
a
new
number
of
years,
but
I
mean
it's
just
a
file
system,
that's
encrypted.
I
don't
send
or
receive
from
it
got
it,
and
I
was
working
with
dotto
to
have
imported
to
mac
os
window.
So
I
feel
I
guess
familiar
with
the
crypto
side
of
it,
but
yeah
the
send
and
receive.
Certainly
it's
had
some
issues.
I
Yeah,
that's
I
have
to
see
if
I
can
find
the
tickets
it's
filed
under
a
lumos
because
alex
wilson
found
it.
So
I
don't
know
if
it's
unique
to
lumos
or
if
it's
also
present
open
cfs.
I
know
like
in
the
past
when
we
were
when
I
was
porting
or
finishing
the
port
rather
of
the
encryption
stuff
to
lumos.
There
was
a
bug
there
that
turned
out
was
in
the
open,
zfs
one
as
well,
but
it's
one.
I
J
J
J
D
A
A
I
I
J
I
And
so
it's
unfortunately
something
you
know,
I
don't
know
if
anyone's
gonna
look
at
and
I
don't
know
it
may
be
unique
to
lumos
or
it
could
just
be
where
it's
more
easily
triggered,
and
I
like
that
issue
that
I
recall,
ran
into
when
finishing
the
encryption
port
it.
It
took
a
lot,
I
mean
as
it
was.
It
took
a
bit
of
work
to
trigger
it
on
illumos,
but
then,
as
it
turns
out,
it
took
a
lot
more
work
just
to
get
that
triggered
on
linux.
J
J
E
J
A
A
D
Oh
yeah,
thinking
about
now
I
actually
can
recall
the
tickets
one
of
tickets.
I
have
it's
actually
a
combination
of
re-keying,
some
changing
of
encryption
parameters
in
combination
with
replication,
so
something
on
receive
site
made.
It
panic,
but
I
still
haven't
looked
deeper
in
no
time.
That's
a
new
one.
J
A
Yeah,
I
don't
have
any
update
from
him
either.
There
was
somebody
asking
me
about
kind
of
similar
functionality
in
the
issue
tracker
recently,
and
I
pointed
them
to
paul
and
I
don't
think
that
he
responded
so
he
might
be
busy
with
other
stuff
right
now,
but
he
he
seemed
pretty
enthusiastic
about
and
making
a
lot
of
progress.
Last
time
he
gave
us
an
update.
A
F
D
D
C
A
K
K
Of
course
there
is
no
in
freebsd
there
is
no
interface
yet
to
do
that.
So
the
interface
is
loaded
with
the
zfs
module
in
freebsd.
At
least
you
can,
you
can
add.
New
system
calls
to
kernel
modules,
so
we
have
to
system.
So
I
have
two
new
system
calls
to
be
able
to
copy
entire
file
or
copy
just
file
range
one
one
missing
bit,
I
think.
K
Cloning
between
data
sets
with
the
same
keys.
So
if
data
set
was
cloned
from
another
data
set,
the
key
is
inherited,
so
we
could
clone
safely.
The
blocks
between
such
data
sets,
but
I'm
not
really
sure
how
I
can
compare
those
two
data
sets
and
tell
that
okay,
they
have
the
same
keys
or
should
I
just
check,
maybe
the
relation
that
this
one
is
clone
of
the
other
or
something
like
that.
F
H
C
A
K
K
There
is
some
edge
cases
which
I
was
trying
to
address.
There
are
probably
more
to
be
honest.
There
is
some
parts
I
not
100
sure
about
that.
Those
are
done
correctly,
but
there
are
cases
like
what,
if
you
clone
a
file
and
then
you
clone
a
clone
within
the
same
transaction
group,
so
I
was
trying
to
address
edge
cases
like
that.
K
So,
yes,
I
think
most
of
the
of
this
is
is
documented,
but
I'm
sure
it
can
be
probably
improved.
I
guess,
but
I
I
I
want
to
finish
the
design
documentation,
because
it's
it's
got
a
bit
more
complicated
than
I
than
I
that
I
wanted
to
save
more
space
like,
for
example,
we
have
separate
block
reference
tables
for
each
vdf
for
each
top
level
vdf,
so
we
don't
have
to
store
vdf
id
in
the
brt
entry,
so
we
save
eight
bits
which
is
33
percent
of
the
of
the
entire
entry.
K
B
C
M
K
C
K
So
on
freebie
is
the.
I
think
the
the
final
solution
is
to
have
interface
within
the
kernel,
so
like
more
generic
interface,
like
regular
system,
call
that
basically
there
is
a
v
up
to
to
file
system
and
any
any
file
system
can
implement
that
if,
if
they
want
to
and
for
example,
for
now,
it
will
be
implemented
only
for
zfs
and
it
will
just
return
not
supported
or
not
for
others
fire
system,
or
something
like
that.
K
K
M
I'm
still
looking
for
reviews
for
the
corruption
corrective
receive,
I
just
rebased
it
after
the
new
erlog
format,
changes
and
it
was
actually
surprisingly
few
changes
or
surprisingly
few
conflicts.
So
I
just
pushed
that,
and
I
don't
know
what
the
tests
will
look
like,
but
yeah
it'd
be
great
if
somebody
can
get
to
that
at
some
point.