►
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
C
B
B
B
B
B
B
B
Dong
dong
is
still
working
on
that
Josh
Cullen
is
taking
a
lead
on
interacting
with
down
dog,
to
see
that
that
gets
shepherded
through
it
really
kind
of
depends.
If
dong
dong
has
time
otherwise,
Joss
is
going
to
take
over
that
PR.
But
even
if
we
do
finish
that
I
think
we
still
want
to
raise
the
help
20
mb/s
or
not
first
matched.
There
could
be
unexpected
issues.
B
A
Yeah
could
be
so
the
approach
that
we
take,
where
the
other
demons
is,
there's
a
there's,
a
require
min
release
field
for
the
USD
map
and
now
for
the
Mon
map.
Do.
Is
there
a
similar
thing
for
the
MVS
map
already
or
once
the
upgrade
completes
to
Nautilus
say
if
you
try
to
start
mimic
it'll
just
refuse
to
let
it
join
like
that.
B
I
believe
we've
been
using
that
Sophie
that
competitive,
mvs,
smarter
and
yes
have
to
do
that,
but
we
haven't
been
rigorous
about
adding
any
bits
to
that
at
least
recently.
A
C
D
B
D
A
D
A
D
A
D
A
Might
be
that
if
the
compat
said
is
a
pert
MDS
map
and
not
an
FS
amount
property,
then
you
could,
just
before
you
assign
a
standby,
just
make
sure
that
that
particular
envious
that
you're
considering
thatis
Phi's
the
compat
set
or
you
assign
it.
That
might
be
sufficient
at
least
four
major
versions.
They
would
help
four
minor
versions,
but
four
major
versions:
no
job,
if
I
think,
if
you
want
something
more
granular
I
would
suggest
adding
an
explicit
require
major
release.
A
require
minor
release
at
field
on
to
the
MDS
map.
B
A
B
A
Yeah
I
mean
I
think
it
depends
on
whether
the
goal
is
to
say
that
octopus
adds
this
new
thing,
and
so
therefore
we
just
need
to
make
sure
that
only
in
the
s
is
that
supports
n,
you
think
it
run
or
if
the
version
is
or
if
the
goal
is
that
I'm,
a
conservative
operator
and
I
only
want
to
deploy
the
new
point
release
on
this
file
system.
There's
my
more
staging
one
before
I
roll
it
out
to
the
more
critical
one,
because
there
I
think
it
is
it.
Isn't
it
isn't
a
feature?
E
So
I
will
say
that
we
actually
get
requests
for
this
in
Rados.
Pretty
often,
people
want
to
be
able
to
roll
out
to
like
portions
of
a
pool
or
a
cluster
in
the
OS
DS,
which
is
basically
not
practical,
but
if
you
have
multiple
file
systems
and
Southwest,
there's
actually
like
easy
boundaries,
and
people
are
definitely
gonna
ask
for
it
and
I
think
it's
pretty
reasonable
of
them
to
do
so.
C
A
B
A
That
would
be
up
here
there.
That
was.
A
B
A
A
B
B
A
B
Yeah
and
dug
up
the
PR,
but
it's
in
the
description,
yeah.
A
Alright,
so
there's
a
scheduling,
the
stats
in
the
mirroring.
Yes,
the
mirroring
might
be
a
stretch,
one
perhaps,
but
hopefully
not
we
don't
I
mean
that's.
We
don't
actually
know
what
the
we
haven't
picked
the
release
date
for
octopus.
Yet
so
we
don't
know
if
we're
talking
about
November
or
March
or
summer
in
between
both
who
knows
exactly
what
all
we're
gonna
get
through,
but
at
least
these
alright
well
is
there
anything
else?
What
else
is
anything
else
that
you
think
is
in
scope
here?
C
C
C
B
A
C
A
A
Yeah
sure
right
so
that
I
think
that
the
meta
discussion
was
that
there
was
there's
discussion
going
on
inside
learn
about
what
their
longer-term
archival
file
storage
strategy
should
be.
They
have
this
crazy
distributed
file
system
that
they've
written
themselves
and
they're.
The
only
news
arrived
at
stores,
like
whatever
300
petabytes,
manages
like
300
petabytes
of
tape,
data.
B
But
I
think
what
we
can
do
is
have
the
MDS
manage
the
actual
migration
that
is
reading
the
old
data
out
and
writing
it
out.
That
should
be
really
easy
to
multi
thread
as
well.
Yeah
and
the
way
I,
envisioned
and
working
is
setting
an
update
to
the
each
I
nodes
file
layout
and
the
MDS
will
lock
file
or
put
files
and
in
queue
to
be
to
be
migrated.
B
And
then,
while
it's
actually
in
the
process
of
migrating
a
file,
it
remains
locked
in
clients
can't
read
her
right
from
it
reads:
maybe
are
okay,
but
no
right
and
the
MDS
is
in
the
background,
moving
it
to
the
new
or
the
other
pool,
or
even
while
any
kind
of
layout
transformation
yeah
and
then
once
it's
complete.
The
files
unlocked
and
right
gaps
can
be
obtained
again
and
the
file
layout
is
atomically,
updated.
A
A
I
think
that
well,
the
first
part
just
migrating
between
radius
pools
sounds
straightforward
and
like
easy
to
understand,
within
the
context
of
a
soft
cluster,
the
the
large
erases
HSM
discussion,
around
tape,
archives
and
having
pluggable
backends
and
so
on,
where
the
NDS
basically
manages
the
metadata
for
a
huge
file
hierarchy,
but
individual
files
get
archived
and
you
set
policy
on
directories
and
so
on.
That
is
going
to
be
some
weird
intersection
of
like
DMAP
ii,
which
I
think
is
the
semi
quasi
standard
interface.
D
A
Could
probably
implement
it
on
top,
also
where,
if
you
just
have
like
you
know
the
equivalent
of
a
actuals
or
magic,
etc,
stew
like
quiesce
access
to
a
file,
then
process
just
copies
it
to
the
archive
and
then
says
that's
an
attribute
that
says
it's
over.
There
I've
seen
your
nut
call
exit
mechanism
so
that
somebody
tries
to
access
the
file
you
get.
D
D
A
Think,
that's
I
think
that's
a
well
what
it's
about
to
say,
think
we
should
have
maybe
have
a
targeted
conversation
with
Dan
and
whoever
else
papi
the
list
about
like
what
it
would
take
to
do
like
a
full
HSM
type
solution
like
that,
because
that
might
be
a
better
investment
of
people's
efforts.
That's
open
source
and
used
by
a
broad
community
as
opposed
to
them.
Building
there.
D
D
C
A
Would
be
awesome,
yeah,
the
other
part
of
this,
which
is
probably
face
the
phase
one
is
just
migrating
between
rid
of
schools.
Phase
two
is
probably
the
ability
to
set
policy
on
a
subdirectory
that
is
sort
of
propagated
to
things
nesib,
anything
that
you
could
say
like
you
know
a
home
Patrick.
Please
move
this
to
the
lowest
II
seat
here,
because
whatever
and
then
have
the
MBS
go
and
apply
that
to
Indian
neutral
files,
yeah.
B
B
Please
you
know
what
this
would
set
off:
SP
PE
are
his
bone.
This.
A
Was
this
is
a
fallout
from
this
ongoing?
Should
we
have
with
liberate
us
where
the
C++
ABI
keeps
changing,
but
the
see
ABI
is
stable
and
because
they're
combined
into
a
single
thought?
Oh
so
we
didn't
want
to
rev
the
the
ABI.
We
didn't
run
a
rev,
the
Esso
name,
because
I
would
break
all
the
linkage
and
builds
and
whatever
for
like
you,
which
uses
the
linkage
and
meanwhile
the
C++
one
was
just
breaking
and
so
keep
you
did.
A
This
split
so
then
are
now
two
dot
esos,
there's
one
that
has
C
binding
someone
that
has
C++
ones,
and
so
you
can
read
the
C++
ones
independently
FC
once
and
the
assumption
was.
We
just
want
to
repeat
that
for
liberty
and
looks
out
of
this
I,
don't
know
how
much
it
matters
I
mean.
Maybe
an
equally
valid
approach
would
be
to
just
drop
the
C++
bindings,
perhaps
for
your
slips
F
of
s.
A
A
B
B
A
A
That's
probably
that
will
fly
work,
fine
for
all
the
sub
volume
style
once
it
won't
help
a
large-scale
users
that
aren't
using
some
planes
explicitly
they
have
their
own
directory
structure
yeah,
but
it
might
be
able
maybe
the
code,
the
manager
code
that
is
written
to
do
this
can
be
written
so
that
it
just
has
a
set
of
sub
directories
that
it
does.
This
on.
That
happened
to
be
defaulting
to
the
sub
volume
based
directories.
C
B
B
B
A
A
A
C
A
Doug,
do
you
want
to
kick
off
a
discussion
with
well
you're,
going
to
talk
to
your
lab
friends?
Maybe
I'll
join
also
reset
to
Dan,
or
you
want
us
to.