►
From YouTube: 2018-03-27 Rook Community Meeting
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).
A
Yeah,
okay,
so
the
recording
is
started.
Welcome
everybody.
This
is
the
March
27th
2018
rural
community
meeting,
and
this
is
the
first
Tuesday
meeting
now
and
this
will
be
the
regular
cadence
going
forward
every
other
Tuesday.
So
we're
happy
that
the
Red
Hat
folks
can
join
us
today
and
some
of
the
folks
from
Nixon
as
well.
So
let's
go
ahead
and
get
started
on
the
agenda
here.
A
I
shuffled,
the
community
topics
in
questions
section
to
come
earlier
than
the
PR
section,
because
I
think
that
some
of
the
PR
stuff
is
can
be
kind
of
specific
and
not
necessarily
applicable
to
the
entire
community.
So
I
wanted
to
have
community
topics
come
first
and
earlier
in
the
meeting,
so
everyone
can
participate
in
those
earlier
on.
So
let's
go
ahead
and
quite
take
a
quick
look
and
see
what
we
have
still
lingering
in
0.7.
A
As
expected,
it
is
the
issue
with
the
rook
agent
flex,
volume
plug-in
sometimes
formatting
a
existing
volume.
Travis
do
you
know
the
latest
on
that,
like
if
I
believe
that
we
had
talked
about
that
properly
determining
when
a
existing
volume
is
already
has
data
on?
It
is
the
right
way
to
go
here,
because
the
fencing
and
locking
that
we
have
in
place
I
believe
is
appropriate.
A
C
B
C
A
Remember
too,
is
that
so
your
this
is
the
determination
of
if
there's
data
on
the
volume
already
is
done
by
the
common
kubernetes
of
you
know
the
upstream
kubernetes
mounter
and
there's
a
change
in
either
1.10
or
1.11.
That
now
uses
block
ID
instead
of
LS
block,
which
I
think
does
work
appropriately
so
going
forward.
This.
E
A
D
E
D
A
D
B
A
D
C
A
B
A
B
E
A
A
Okay.
So
let's
take
a
look
at
quickly:
the
0.8
milestone
and
projects,
I,
don't
think
I'm,
not
sure.
We've
had
a
whole
lot
of
specific
focus
on
some
of
the
stuff
and
0.8
recently
I
have
been.
This
is
in
review.
Actually
so
I've
been
focusing
all
of
my
work
time
on
the
design
for
multiple
storage,
backends
and
types
so
that
that's
definite
sort
out
eight
thing
and
that's
what
I've
been
focusing
on.
Is
this
board
fully
up
to
date,
like
I
use,
one
OSD
per
pod?
E
A
F
D
A
Dimitri,
normally
the
normal,
the
normal
flow
or
process
for
this
project
is,
if
you
know,
for
big
features
to
do
a
design,
one
pager
upfront
and
to
take
that
all
the
way
through
review
into
completion
before
you
know,
work
real
work
is
started
on
the
feature
itself.
You
know,
prototyping
effort
can
be
part
of
that,
perhaps
for
investigation
and
research
to
help
with
the
design,
but
normally
that's
the
flow
for
big
features.
A
C
A
I
know
we
definitely
have
steps
about.
You
know
what
the
contribution
flow
is,
but
I'm
not
sure
if
we
directly
address
that
specifically
so
I
will
take
a
follow
up
to
make
sure
that
we
have
good
guidance
and
documentation
on
that.
Good
actually
should
just
add
that
to
the
agenda
doc
then
so
the
community
knows
no
side
notes
here.
A
A
B
B
A
B
C
A
A
What
else
are
we
concerned
about
here?
I
would
think
that
so
you
know
Nixon
said
you
know,
that's
that's
something
that
we're
excited
about
to
be
able
to
add
support
for
in
rook,
but
since
we're
still
working
on
the
nailing
down
the
design
right
now,
I
have
a
feeling
that
you
know
actually
incorporating
next
sensor
would
not
necessarily
make
it
into
scope,
40.8
and
be
more
of
a
0.9.
F
F
C
F
F
A
That
sounds
great
I
would
love
to
see
that
artifact
Thank
You
Caitlin
and
to
meet
you
for
that
effort.
There
and
I
found
also
with
kubernetes
automation,
work
that
finding
what
the
manual
steps
are
goes
a
long
way
towards
what
the
code
that
will
be
implemented
inside
of
an
operator
controller
ends
up
being
so
that
that's
a
huge
step
towards
that
progress.
There
sounds
like.
Maybe
we
should
capture
that
in
a
issue
40.8
the
design
for
incorporating
next
Senta.
Oh.
E
C
C
C
B
C
C
B
C
B
E
C
A
C
And
it
looked
like
even
that
work
was
very
close,
so
I
mean
the
changes
there
were
to
reduce
the
number
of
service
accounts.
They
were
being
used.
I
should
maybe
even
bring
it
down
to
one
and
then
and
then
make
sure
that
namespaces
we
know
are
created
by
the
admin
and
honor
any
kind
of
PSP
policies
and
other
stuff.
So
the
end
goal
there
is.
C
A
Excellent,
a
lot
of
these
other
things
here
are
sort
of
opportunistic.
You
know
if
we
could
get.
You
know
help
from
the
community
or
you
know
spike
smaller
scope.
Bug
fixes
like
this
one
here,
for
you
know
racial
coding
for
a
block
and
like
those
things
you
know
it's,
it's
not
clear.
If
you
know
they're
gonna
get
to
be,
you
know,
taken
care
of
in
the
zero
date
milestone
but
opportunistically
getting
it
the
right.
You
know
getting
a
contributor
to
help
with
that.
C
E
G
F
E
A
G
If
you
have
a
concept,
if
I
want
to
erase
your
encode
this
object,
it
would
be
nice
to
have
some
sort
of
facility
where
you
can
essentially
ask
the
backend
storage
cluster.
If
it
knows
how
to
do
that,
and
you
invoke
it
directly
to
do
it.
Otherwise,
you
use
default
that
works
over
generic
piles
or
blocks.
A
We'll
get
back
to
that
in
just
a
little
bit
coming
up
in
the
meeting
here.
All
right,
so
is
there
anything
that
anybody
else
wants
to
bring
up
for
the
0.8
scope,
I
think
we
we
have
a
kind
of
an
idea
of
you
know
what
the
scope
is
looking
like
for
the
next
in
the
timeframe.
That's
left,
you
know
further
LeMay
yeah.
A
B
G
C
A
C
B
You
know
the
deployment
using
deployment
well,
actually
using
deployment
doesn't
specifically
address
these
issues,
but
there's
things
like
make
sure
we
never
put
two
pawns
on
the
same
node.
We
can
control
that
and
then
what
were
some
of
the
other
ones,
seeing
if
we
can
use
the
consistent
Mon
ID
per
node,
that's
gonna
require
some
prototyping
I'm,
not
sure
that
will
work,
but
I'm
not
sure
we
need
that
either.
A
C
A
B
B
A
Okay
sounds
good
to
me,
I
mean
hopefully,
Alex
will
be
able
to
adjust
the
schedule
so
that
he
can
join
us
for
future
meetings.
That's
unfortunate.
The
Gees
has
a
conflict
at
this
time
on
Tuesdays,
okay,
so
next
topic
is
there
any
reason
to
not
drop
kubernetes
1.6
support
now,
I.
B
E
B
A
C
E
C
Another
thing
to
do
is
to
slack
you
know,
like
four
hundred
plus
people
now
asked
who's
running
off
and
shift.
C
E
A
Okay
cool,
so
then
next
topic
I
just
wanted
to
just
bring
this
up
briefly,
there's
not
really
a
discussion
to
be
had
here,
but
so,
as
a
member
of
the
cloud
native
computing
foundation,
Brooke
was
able
to
be
participate
in
being
a
organization.
That's
as
students
apply
for
projects
this
summer
through
the
google
Summer
of
Code
program,
and
we
got
was
at
five
applications
Travis
from
five
different
students
to
try
to
take
on
some
of
the
projects
that
we
suggested
for
their
summer.
A
A
B
A
Alright,
so
multiple
storage
types
design
so
since
the
mixed
sense
of
folks
are
on
the
line-
and
you
know
we're
very
excited
for
you-
know
the
collaboration
we're
getting
from
you
guys,
let's
go
ahead
and
you
know
you
open
up
the
floor
for
some
of
the
discussion
topics
that
I
Caitlin
was
mentioning
earlier
Caitlin.
Do
you
want
to
go
ahead
and
take
that.
G
G
G
C
C
C
G
Things
like
open,
V
switch,
doo-doo-doo-doo
support
this.
The
the
problem
is:
is
that
they're
not
there
right
now,
they
haven't
been
told
to
add
a
network
to
an
existing
set
of
pods
and
what
I'm?
Actually?
What
might
Derry
right
now
is
you
add
you're,
not
the
pods
you're,
adding
our
each
in
these
scopes
networks.
But
then
these
front-end
pods
talk
to
back-end
pod.
C
C
E
E
C
E
G
G
C
So
assuming
the
dub
exists
somewhere,
whether
it's
calico
or
it's,
you
know
some
CNI
plugin
that
supports
this
at
the
rook
at
the
rook
level.
What
we've?
What
we've
done
is
we've
said:
look
you
could
decide
where
to
run
these
demons
in
a
namespace
that
has
you
know
some
PSP
or
network
policy
or
whatever
not
the
decision
to
of
which
demons
run
in
which
namespace
is
it
left
up
to
the
admin?
C
G
C
G
C
You
undocumented
I
think
so
it
feels
like
we
should.
We
should
have
a
you
know,
ensure
around
supporting
multiple
networks,
whether
it's
for
isolation
or
it's
for
performance,
and
that
that
should
be
a
you
know
if
we
need
to
factor
I
didn't
do
other
designs
that
we
should,
but
it
does
feel
like
orthogonal
to
to
what's
happening
here
and
the
local
backends.
F
So
guys
I
did
some
research
on
this
topic.
Apparently
signet
workgroup
looks
like
working
on
the
proposal.
At
least
that
is
loosely
based
on
Intel
down
with
their
plug-in
cm,
I,
apply
them
called
move,
and
so
I'm
hopeful
that
soon
google
makers
will
have
it
properly,
designed
and
degraded,
but
not
anytime.
Soon,
I
think
it's
gonna
be
the
end
of
the
summer,
but
a
birth
case
and.
F
C
B
G
A
Yeah,
this
is
something
that
I'm
very
happy.
You
know
with
the
community
that
we're
building
around
here
that
is
bringing
in
this
networking
expertise,
because
you
know
I've
been
working
on
the
design
for
a
multiple
search
back
in
so
far,
but
I
do
not
have
that
networking
background,
an
expertise
to
be
able
to
really
come
up
with
the
right
thing
to
do
here.
So
the
you
know,
investigation
that
we
can
do
you
know
as
a
community,
with
support
from
other
people
to
figure
out
how
this
can
be
properly
supported
is
very,
very
valuable.
A
G
Think
we
have
to
talk
about
some
taxonomy
and
in
the
design
on
things
like
is
the
storage
cluster
providing
its
own
for
its
own?
Is
it
using
local
store
or
a
sand
service
behind
it
and
I?
Don't
think
you
can
meaningfully
talk
about
configuration
or
steps
taken
by
the
store
service
without
differentiating
between
those
two
very
fundamentally
different
types,
and
it
also
has
big
impact
on
the
kubernetes
scheduling
our
type
of
storage
cluster,
where
we
want
to
heal
our
problems
ourselves.
G
A
Way
I
would
describe
it
is
that
you
know
there's
kubernetes
level,
redundancy
or
high
availability
operations
such
as
you
know,
filling
over
a
pod,
or
you
know
what
a
pods
goes
down
a
replication
controller
brings
it
back
up
right,
there's,
also
storage
back
in
specific
logic
to
you
know,
to
handle
replication
and
sure
redundancy
of
data
that
such
that
staff
does.
You
know
they
couldn't
OST
is
marked
out,
then
I'll
start
migrating
placement
groups
and
things
like
that
right.
So
it's
alright
handles.
A
C
G
G
G
I
would
like
to
think
of
Iranian
some
of
these
teachers
I'm
not
positive.
So
far,
most
of
the
people
we've
been
seeing
when
they
do
the
file
system
over
there
object
storage
system.
They
most
of
them
have
been
putting
the
files
namespace.
In
a
separate
space,
there
happened
some
attempts
at
unifying
the
namespaces
we're
not
just.
G
C
We're
where
there's
a
lot
of
flexibility
and
how
you
both
expose
that
as
an
abstraction
and
also
how
you
actually
implement
that
in
kubernetes
and
so
I
would
hope
that
you
know
such
design.
Differences
between
storage
backends,
you
know,
will
continue
to
work
as
they
do
today
and
I
think
Jarvis
mentioned
this
earlier.
So
there's
no.
We
were
not
trying
to
do
anything
on
the
data
path,
so
so
we're
not
forcing
any
changes
from
novice
buckles
right.
G
G
C
C
C
So
we're
trying
to
find
that
delicate
balance
between
the
two
and
you
know,
but
let's,
let's
maybe
I,
would
suggest
that
we
have
stuff
at
this
point,
I
think
Jarrod's
gonna
look
at
many
o
as
another
back-end
and
then
we
have
next
sentence.
So
that's
and
let's,
let's
come
up
with
the
you
know
the
right
abstractions,
with
the
right
amount
of
commonality
between
them,
but
not
and
take
that
up
in
an
endurance,
design.
G
A
If
it
needs
to
be,
you
know
it,
there
needs
to
be
any
silos
there
or
if
it
needs
to
have
something,
that's
more
common
or
you
know,
or
it
takes
to
set
the
same
approach
that
we
do
at
the
cluster
level,
where
you
know
we
have
a
general
commonality
across
clusters
and
then
backends
can
add
their
own
specific
properties,
sort
of
in
line
with
that.
That
may
be
the
way
to
go
for
a
file
and
object
as
well
and
then
something
to
explore
further.
So
it's
definitely
useful
that
you
brought
that
up.
Caitlin.
E
E
A
C
A
E
A
Think
another
comment
to
make
here
just
in
general,
is
I.
Think
we've
been
doing
a
better
job
recently
on
in
our
slack
conversations
of
using
threads
to
keep.
You
know,
conversations
kind
of
not
constricted
but
well
organized.
So
you
don't
have
you
know
interleaving
conversations
going
for
pages
and
pages.
You
have
a
particular
topic.
That's
you
know
well
organized
and
kept
together
in
itself,
so
I
duly
appreciate
that
people
using
threads
recently
that's
been
helpful.
I'm.
C
C
C
A
Well,
you
guys
obviously
have
lots
of
free
time,
as
bassam
mentioned
so
I
don't
know.
What's
here
guys
right,
ok,
so
I
think
that's
everything
we
have
here.
So
thanks
everybody
for
joining
hands,
a
definitely
beneficial
meeting
today
for,
for
you
know
the
community
as
a
whole.
So
thank
you
to
everybody.