►
From YouTube: 20200212 - Image Builder Office Hours
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
A
Just
to
start
put
together
a
little
bit
of
an
agenda
for
today
to
talk
about
the
image
builder.
If
you
don't
have
a
link
to
this
document,
I
can
drop
it
in
chat,
and
anybody
can
add
anything
to
the
agenda
that
they
would
like
for
sure.
We
drop
a
link
to
this
in
chat
just
in
case
there.
We
go.
The
first
thing
that
I
had
on
the
list
here
was
last
week:
I
got
maintainer
access
to
the
repo
and
I'm
very
interested
in
applying
in
the
beginning,
0.1
tag.
A
This
is
come
on
a
bunch
of
times.
Before
most
you
had
no
objections
to
tagging
the
repo
I
just
kind
of
wanted
to
confirm
that
we
don't
really
have
anything
in
place
about
when
the
right
time
to
do.
It
is
but
I
think
we're
at
a
place
where
I
have
downstream
consumers
that
want
to
be
able
to
depend
on
something
not
changing,
and
it
is.
It
is
very
much
a
0.1
State,
but
I
think
we're
in
a
place
where
we
can
do
that.
A
Yeah,
okay,
you
know
I
the
thing
that
I
always
feel
weird
about
is,
as
you
point
out,
there's
multiple
tools
in
the
repo
there's,
the
cluster
API.
You
know
Packer
based
build
or
risk
your
config
Adams
stuff
and
the
cubic
deploy
stuff.
So
it's
kind
of
weird
to
tag
multiple
projects
without
some
coordination,
but
we
got
to
start
somewhere,
yeah,
okay,
so
I'm
gonna
do
a
V,
zero
dot.
One
this
week
might
be
today.
A
A
A
The
MD
book
stuff
is
in
place,
but
I
haven't.
We
had
a
link
to
it
from
like
the
top
level
reading.
So
I
was
wondering
if
the
right
thing
to
do
there
would
be
at
the
top
level.
Read
me
call
out
the
different
projects,
meaning
you
know
whether
it's
the
cluster
API
builder,
the
cube,
deploy
stuff,
whatever
is
happening,
they're
linked
to
their
you
know,
down
into
images,
folder
path
and
link
to
their
documentation
or
getting
started.
I
think.
A
That's
probably
what
I'll
try
to
submit
as
a
PR
to
kind
of
point
people
to
the
different
projects.
So
they
understand
what
actually
is
in
the
repo,
because
I
know
I'm
confused
about
it.
If
it's
not
the
cluster
API
stuff,
I,
don't
know
anything
about
cube,
deploy
or
cops
or
any
of
that
kind
of
or
those
various
other
projects
so
like
I'd
like
to
put
placeholders
in
there.
So
someone
who
is
knowledgeable
about
them
can
fill
them
in
and
help
me
as
well.
So.
C
A
A
Some
of
those
tools
had
different
goals,
and
so,
for
example,
one
of
them
in
in
there
and
so
like
I,
can
just
kind
of
let's
do
since
we're
sharing.
Just
if
I
go
look
at
the
sorry
because
of
my
screen.
Sure
I
can't
actually
see
my
address
bar
anyway,
and
this
kind
of
what
we're
talking
about
exactly
is
that
the
documentation
doesn't
really
explain
this
very
well.
But
there
are
a
few
different
tools
in
here
and
one
of
them.
A
It's
for
building
images
for
consumption
by
what
we
call
Kathy
or
capi,
which
is
the
cluster
API
project
in
images
built
for
cluster
API,
have
a
very
specific
goal
of
being
able
to
be
deployed,
whether
you're,
using
a
provider
for
Amazon
for
Google,
for
digital
ocean
for
vSphere,
and
the
images
that
we
build
and
publish
from
them
are
considered
to
be
immutable.
So
there's
no
package
updates,
like
we
bundle
up
kubernetes,
pacific
kubernetes
version
and
all
the
dependencies
to
run
it
and
whatever
you
need,
but
you
the
image
is
never
supposed
to
be
changed.
A
Tenets
of
the
cluster
api
images,
but
other
tools-
don't
do
it
that
way.
There
are
other
tools
in
here
and
which
is
kind
of
what
I
was
talking
about
with
motion.
There's
this
configure
atom
stuff
that
that
motions
been
leading
for
a
while,
and
this
cube
deploy
project
as
well
and
I'm,
not
very
knowledgeable
about
them.
So
how
would
we
describe
it
much
so.
A
Now
say
for
the
cluster
API
based
images:
yes,
that
the
that
there's
nothing
additional
that
needs
to
get
installed,
that
there
is
an
easy
way
for
each
provider
to
build
images
tailored
for
their
needs
or
for
the
specific
cloud
provider.
So,
for
example,
the
providers
that
are
support
here
right
now
or
Amazon.
So
you
build
your
a
ice
age
or
digital
ocean
Google.
Compute
and
Oba
is
for
a
piece
here
based
infrastructure.
B
B
So,
what's
probably
going
to
happen,
is
that
a
little
bit
of
the
courage
that
currently
sits
in
config
admin
will
go
into
what
we're
calling
a
CLI
that
will
wrap
all
these
the
different
methods
of
holding
image,
so
your
other
body
in
image,
using
Packer
and
spinning
up
an
instance
or
you
building
an
image
using
a
local
VM
and
and
doing
it
that
way,
and/or
using
ansible
on
and
we're
not
using
ansible.
So
there's
two
conner
all
three.
B
C
C
B
This
is
its
kind
of
been
very
much
demand
driven
so
as
people
have
a
pressing
demand
and
and
kind
of
downstream
users
who
want
a
specific
image
for
a
specific
cloud.
Somebody
goes
out
and
does
the
work
and
then
adds
it
into
here.
So
there's
nothing
stopping
somebody
from
from
creating
an
openstack,
a
pecker,
bold
for
OpenStack
in
submitting
it
into
the
kepi
folder.
For
example.
B
That's
that's
exactly
what
the
user
folks
did,
for
example,
so
the
Amazon
and
the
OVA
were
done
a
lot
by
Travis
and
team,
and
then
the
azure
images
were
done
by
Susilo
Robert
from
Microsoft
I'm,
not
sure
he
did
the
digitalocean
one,
but
there's
nothing.
Stopping
anybody
who
has
a
need
for
OpenStack
for
creating
an
open,
stack,
folder
doing
whatever
is
necessary
and
committing
it.
B
The
problem
we
do
first
today,
however,
is
that
nothing
that
India
is
actually
tested
or
published,
which
was
Connor
where
we
want
to
move
towards,
is,
is
getting
this
work
tested
and
published
so
that
it's
consumable
from
an
image
not
from
a
tool.
So
at
the
moment,
if
you
want
to
build
an
image
you
have
to
download
or
clone
the
repository
and
run
the
tooling
to
create
your
own
image.
Oh
I
think
some
of
the
images
have
been
pre-created
and
linked
to,
but
that's
very
much
an
out-of-band
process.
It's
not
a
it's,
not
image.
A
Now
the
only
thing
I
was
good
in
the
ad
is
that
some
wood
is
is
not
automated.
So
the
a.m.
eyes
that
are
built
with
this
tool
are
published
publicly.
The
OVA
is
are
published
publicly
and
they
can
be
downloaded
and
used
as
they
are
and
and
they
are
not
part
of
any
automated
taste
testing
framework,
it's
basically
people
using
it
and
saying
yeah
it
works.
And
if
it
doesn't,
we
hear
back
so
there
isn't
a
lot
of
automated
testing,
but
but
we
do
publish
a
Mis
and
OVA
s
regularly,
but
it
is
manual.
C
Manual
is
fine
at
this
stage,
but
what
I
found
interesting
was
metal,
cube
is
doing
its
own
and
that
is
metal,
3
dot
IO,
they
are
doing
their
own
images.
So
what
is
the
difference
in
sense
that
if
I
have
to
install
a
cluster
and
for
that
cluster,
I
need
hosts
or
instances
and
those
I
need
to
create
using
bare
metal
and
then
I
have
to
install
the
Horus
and
then
I
have
to
partition
them
into
whatever
the
pods
kubernetes
pods
and
clusters?
So
there
is
a
whole
slew
of
steps.
C
B
There
is
no
minimum
requirement
or
assumption
so
I'm,
actually
not
familiar
with
what
metal
cube
is
doing,
but
this
there's
nothing
stopping
them
from
joining
this
this
meeting
and
adding
their
solution
into
the
mix
and
OH
horrifying
what
we
have
to
make
it
suitable
through
Raman
parametric.
Any
reason
why
this
Mart
not
wrong
there
metal
today
is
because
nobody
uses
actively
participating
in
the
project
has
downstream
users
who
are
asking
them
to
produce
pyramidal
images,
but
there's
nothing
in
the
project.
Word:
map
or
goals
and
non
goals
that
says
a
metal
obscure.
B
C
C
Based
on
your
comments,
I
see
there
is
scope
for
airing
both
OpenStack
and
metal,
cue
to
possibly
candidates
for
image,
building
who
possibly
can
participate
here
unless
they
have
some
reasons
which
I
don't
know.
So
what
I
would
say
is
I
would
like
to
request
those
things
to
see
and
engage
here,
at
least
to
figure
out
if
there
is
a
possibility
and
what
is
that
makes
it
difficult
or
what
problems
they
have,
that
we
can
leave
with
them
and
how
they
can
leverage
vortexes.
C
C
Because
yesterday,
when
I
had
attended
the
trustor
EPA
I,
don't
know
there
are
many
meetings
of
that.
One
of
them
probably
was
office
hours
or
something
or
something,
and
then
I
was
referred
here
that
if
you
want,
if
you
have
problems
for
image,
why
not
attend
this,
and
so
I
came
here
searching
for?
Is
there
anything
that
I
can
leverage
from
this
for
what
we
need
in
metal,
cube
and
I
am
do,
although
I
don't
write,
the
Medi
cube.
D
C
Gate
which,
which
were
basically
short
of
resources,
and
so
if
I,
could
liberate
somebody
from
there
to
bring
here
and
have
commonalities,
so
we
can
reduce
over
both
floor,
etc.
So
it's
optimization
possibly
opportunity.
I
am
trying
to
look
it.
Hopefully
that
explains
my
jumping
into
this
car
happy
to.
A
Have
you
for
sure,
and
and
as
you
can
tell
by
the
number
of
attendees
on
this
call,
there's
not
a
lot
of
additional
contribution
happening
right
now,
so
anybody
like
Moshe,
it's
very
demand-driven
someone
kind
of
needs
to
create
images
for
a
particular
provider,
every
provider
right
now
with
some
type
of
cloud
provider
or
virtualization
provider.
So
there
hasn't
been
anybody
who
is
contributing
a
bare-metal
implementation,
yeah.
C
So
this
is
where
I
will
look
into
bare
metal,
how
why
etc.
There
is
a
Quay
or
something
they
have
image
library
where
they
deposit,
all
that,
and
so
I
want
to
find
out
from
the
team
and
airship
team.
Can
we
leverage
this
as
the
you
have
reported,
where
you
deposit
your
images
and
all
or
is
it
its
channel
to
it.
B
Are
tied
to
the
vSphere
cloud
in
it
provider?
Ok,
but
the
the
config
admin
images
shouldn't
be
tied
to
the
the
vSphere
provider
at
all
and
I
will
actually
want
to
try
publishing
him
on
github
itself,
so
creating
a
release
on
github
and
attaching
those
images
as
artifacts
to
the
release
other
than
what
you
think
of
that
Travis.
C
B
B
Integrate
at
a
girl,
level
and
and/or
use
some
of
the
curve
some
of
the
configured
incurred,
especially
from
girl.
That's
what
it's
designed
for
I
saw
the
contract
admin
tool.
What
it
does
essentially,
is
it
you
give
it
a
spec
and
it
takes
that
spec
and
applies
a
whole
bunch
of
transformations
to
it.
That
will
then
spit
out
a
set
of
bash
commands
and
falls
to
to
save,
which
is
then
created
in
a
in
a
cloud
init
file
or
run
directly.
B
C
So
I
think
I'll
have
to
go
through
the
whole
thing
before
I
can
really
contribute
anything,
but
at
least
let
me
review
it
and
see
if
I
can
get
some
people
to
help
us,
because
on
my
own
I,
don't
think
I
will
be
able
to
do
it
fully,
because
it's
a
little
bit
more
complex,
then
on
documentation
side.
Yes,
I
can
do
it
on
coding
side.
C
So
those
are
the
so
I
will
investigate
this
and
suddenly
like
to
see
something
happening
here,
because
it's
quite
complex
for
me
to
understand
the
whole
process.
The
multiple
layers
here,
metal,
cube
as
its
own
ionic,
has
its
own.
Then
API
is
its
Raman
and
metal
operator
has
its
own,
so
bit
more
complex
than
I
can
understand
my
level
of
understanding
but
I'll
try
to
coordinate
and
see
if
I
can
bring
something
to
the
table
to
if.
B
If
you
look
at
that
configured
images
reaper,
there
are
bunch
of
images
that
have
been
published
there
being
color
using
that
as
a
testing
ground.
Okay,
that
will
come
into
this
Reaper.
So
now
that
we
have
TCP
infrastructure
available
for
building
the
images
that
config
admin
images,
Reaper
will
merge
into
this
one
and
we
should
be
able
to
get
some
images
published
for
whatever
provider
is
available.
That's
it!
That's!
That's
that
helps.
B
C
A
C
C
Let's
take
one
tool:
let's
go
through
the
tool.
You
have
one
tool
you
take
in.
Not
you
explain
me
what
is
on
configured
mean
image
building.
He
said
if
you
have
expertise
under
or
whatever
choose
me
one
flow
of
image
building.
So
if
I
have
to
build
image,
I
need
to
first
have
some
ml
five
or
some
Jason
Jason
file
with
the
packaging's
our
manifest
and
packages
whatsit.
We
need
to
do
to
get
English.
A
B
E
D
B
So
basically
it
previously,
it
was
just
a
side
project
of
mine,
but
I
actually
just
started
a
company
called
flank
sauce.
So
it's
been
moved
to
the
flank
source
repository
and
I'm
gonna
be
spinning
up
some
more
resources
to
kind
of
contribute,
Rochester
config
admin,
but
to
hopefully
image
builder
as
well.
So
we
can
get
a
bit
more
support
into
it.
That's
kind
of
my
goal
at
the
moment
and
then
at
the
cent
it
is
cnc
of
license
before
you
wanted
to
bring
it
into
this
repo
part.
A
B
So
even
now,
there's
a
frank
sauce,
config
admin,
repo
that
does
integration
testing
using
docker
containers,
so
it
will
spin
up
a
docker,
systemd
container,
install
kubernetes
and
and
all
the
the
dependencies
run
some
tests
on
it
and
spin
it
back
down
after
it's
done
with
that,
it
produces
a
banner,
and
I
can
do
that
across
like
five
or
six
different
operating.
So
there's
there's
quite
a
bit
of
unit
testing
and
integration
testing
both
into
that
directory.
B
B
Are
the
the
specs
that
define
what
goes
into
the
image?
So
what
the
spec
says
is
deploy
all
the
dependencies
of
kubernetes,
deploy
all
the
dependence,
vadhaka
and
Iran
cubed
and
config
images
pool
and
there
and
then
do
a
cleanup
of
that
image
so
that
it's
it's
temporary
so
that
at
its
core,
then
what
that
will
do
is
it'll
then
split
that
out
into
a
whole
bunch
of
packages
package
repos
that
it
needs
to
install
a
bunch
of
commands
and
needs
to
run,
and
then
it
does
that
and
they'll
go
either.
Do
that
in
line.
B
So
you
know,
config
admin
apply
with
a
spec
and
it'll
actually
create
the
files
and
run
the
commands
for
you
or
you
can
go
conflict
admin
images,
bold,
which
will
download
a
cloud
in
each
file
spin
it
up
using
KVM
with
the
cloud
in
it
already
created.
As
that
as
that
image
spins
up,
it
runs
all
the
cloud
in
it
and
then
shuts
itself
down,
and
then
you
have
a
baked
image
with
Senshi.
So
there's
no
there's
no
ansible
involved
is
no
ssh
involved.
C
B
All
config
admin
requires
is,
is
kvn
or
nested
virtualization.
If
you
build
in
image,
if
you
just
want
to
test
the
configuration
that
config
admin
uses,
then
you
can
run
that
on
on
download,
binary
and
run
it
with
a
spec
in
it
will
apply
it
on
that
operating
system
and
and
do
whatever
it
needs
to
do.
But
I'll
run
it
on
on
a
Mac
in
on
circle
CI
and
a
nested
virtualization,
so
it
the
number
of
dependencies
that
it
requires
is
just
the
binary
and
and
KVM,
and
it
actually
produces
a
docker
image
as
well.
B
A
B
D
I
think
honestly,
we
have,
we
don't
even
have
an
image
yet
so
we
just
need
to
like,
let's
see
if
we
can
create
the
image
based
on
Bionic.
That
basically
has
the
same
set
of
changes.
That
is
that
is
in
like
if
we
go
into
the
cube,
deploy
repo
most
of
the
or
the
cube
deploys
I'll
rush
free.
Most
of
the
changes
are
isolated
to
like
a
script.
I
don't
know
if
yeah,
if
you
can
click
around
sorry
image
is
deploy
and
then
image
builder
and
then
templates.
D
And
then,
if
we
look
at,
like
115
stretch
dot
yml,
you
can
see
that
like
pretty
much,
not
a
lot
going
on,
and
then
we
have
like
a
bunch
of
packages.
We
pre
install
that
are
important
and
like
some
cloud
in
it
tweaks
and
some
like
installing
a
best
CLI
and
like
doing
a
bunch
of
stuff
and
like
make
sure
we
don't
making
sure
we
can
boot
things
like
that,
and.
D
But
this
seems
very
compatible
with
what
you
have.
The
syntax.
I
saw
right
where
a
second
runs
a
class
who
just
would
like
I
guess,
but
what
we
should
evaluate
is.
Can
we
effectively
translate
this
logic,
to
the
extent
that
it
applies
between
stretch
and
Bionic?
Can
we
just
translate
that
to
a
to
a
configure
diem,
build
and
create
an
image
that
works,
and
then
we
could
be
like
all
unified
behind
behind
this?
D
E
D
In
the
same
way
that
we
don't
want
cube,
deploy
to
be
like
a
different
thing
like
we
should
probably
try
to
have
like
one
unless
there's
a
valid
reason
to
like
in
general.
We
should
not
have
different
solutions
for
the
sake
of
it,
but
we
should
try
to
have
like
more
consistency.
If
we
can,
somebody.
D
B
B
We
set
up
the
automated
pipelines
and
automated
testing
so
that
we
can
produce
images
from
birth,
and
then
we
can
iterate
on
that
and
try
and
bring
the
changes
that
are
common
down
and
to
make
them
common
and
then
make
the
solutions
extensible,
so
that
if
I
want
to
use
ansible
to
post
configure
my
image
I
can
use
ansible.
The
first
configure
my
image
and
if
I
want
to
start
with
an
existing
AWS
VM
I
can
start
with
an
existing
AWS
and
configure
it
for
for
kubernetes
and
I
can
do
that.
A
So
most
kind
of
framing
this
into
Tim
st.
Clair
task
of
cleaning.
This
up
and
by
this
I
mean
the
repo
like
it
still
sentenced
me
like
all
the
projects
that
are
there.
I
have
a
purpose
and
well
I
mean
we
can't
really.
He
was
asking
us
to
like
carve
things
out
or
move
them
a
little
bit
deeper,
so
they
weren't
as
I
think
right
in
your
face
right,
it's
public
if
there
was
a
project
that
wasn't
being
actively
used,
but
I'm
still
hearing
that
they
kind
of
already,
even
if
it's
very
nice,
so.
B
D
I'd
say
like
I
would
expect
that
I
will
have
some
requests
to
go
like
like
that.
There
will
be
that
when
we
try
bionic,
we
will
find
some
things
that
we
would
like
to
enhance
or
add
to
come
to
give
you
a
Liam
Packer
in
the
past
has
proved
not
the
right
approach
from
a
reproducibility
point
of
view,
type
thing
for.
B
Us
yeah,
so
I
think
we're
always
going
to
have
that
situation
where
some
people
prefer
a
building
locally
and
some
prefer
spending
oblivion.
So
I
think
we
do
need
to
support
both
but
I
think
once
we
have
both
general
ways
of
building
an
image,
we
should
try
and
converge
as
much
as
possible
on
with
the
actual
bits
that
go
onto
it
and
make
those
as
common
as
possible.
B
C
B
So
it's
not
even
not
they're,
not
common,
even
at
that
level.
At
this
point
and
I
think
we
need
to
make
them
common
at
that
level,
so
that
if
I
choose
container
D,
whether
I'm
using
Packer
or
whether
I'm
using
config
admin
to
produce
my
image,
I
get
the
same
container
D,
that's
on
disk
and
the
same
container
D,
whether
if
I'm
on
a
Debian
system
that
was
produced
out
of
this
project,
the
container
D
will
be
installed
and
configured
same
way.
C
B
Commonality
in
the
spec,
so
if
I
choose
a
version
of
container
D,
whether
I'm
building
via
a
VM
or
whether
I'm
moving
by
a
k,
vm
and
and
but
Saunders
I,
should
get
the
same
output
when
I
at
the
end
should
look
to
say.
I
shouldn't
be
able
to
distinguish
the
two
at
the
moment.
I
can
distinguish
the
two,
because
the
wires
is
installed
by
our
package
repository
and
the
other
is
installed
by
a
binary
downloaded
from
a
service,
and
that
is
something
that
we
want
to
converge
and
all
of
those.
C
I
got
answers
but
I
love
to
steal,
study
and
make
sure
that
I
approach,
metal,
cube
image
builders
team
and
see
how
they
can
bring
into
this,
and
the
other
one
is
of
course
the
open
stack
which
I
am
looking
at
is
a
cloud.
Does
it
have
anything
and
can
any
grief
bring
any
value
here,
so
it
we
at
the
single
point.
C
Otherwise
it's
very
hard
to
trace
what,
when
how
so
I'm
I'm
with
you
all
I
will
try
what
best
I
can
to
bring
people
here
to
coordinate,
and
since
yesterday,
I
Jason
we
had
in
the
I
think
what
does
that
cluster
API
meeting
I,
don't
remember,
was
it
a
where
I
came
for
the
probably
first
or
second
time
so
I'm,
not
very
conversant
still
with
the
kubernetes
way
of
doing
things
trying
to
learn
and
see
if
I
can
coordinate
couple
of
things
from
a
mystic
side
of
it?
So
the
meeting
that
you
had
yesterday.
B
That
was
the
cluster
large
circle
office
hours
that
happens
every
two
weeks
and
that's
for
all
of
the
sub
projects
of
custom
life
cycle,
and
then
there
was
a
meeting
just
before
this
one,
which
was
casta
API.
That
happens
every
week.
And
then
this
meeting
is
the
image
folder
office
hours,
and
this
one
happens
after
two
weeks.
C
And
learning,
but
I
did
see
that
value
for
image
is
important
for
us,
because
that's
the
deployment
point
for
this,
even
whenever
I
do
redfish
either
whether
I
do
otherwise
cloud
in
it
so
and
for
all
the
clouds,
basically
not
just
one
cloud.
So
it's
a
good
place
to
gather.
Where
can
I
get
a
image?
How
can
I
bring
people
to
produce
image
and
if
we
can
standardize
the
process?
That
will
be
super,
but
it
will
take
time.
I
know
and
I
mean
until
then
we'll
have
to
see
how
we
can
keep
maintain
the
tempo.
A
A
A
Don't
know
if
you
were
anyone
on
your
team.
Who
is
some
learn
more
with
like
the
cops
and
and
cube
deploy
stuff
this?
How
to
look
at
it
look
I
mean
the
goal
in
tear
is
just
dead.
Simple,
like
I
should
just
slash,
prove
it
like
I
know
it
fixes
it,
but
still
we
we've
kind
of
been
taking
like
I,
only
kind
of
understand
that
stuff,
that's
under
the
the
cluster
API
folder
and
I'm,
happy
to
lgt
and
stuff.
There's
a
couple
things
they're
looking
for
a
proves
that
you
might
take
responsibility
for.