►
From YouTube: 20200408 - 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
Okay,
I'm
recording
and
we're
live
hello.
Everyone-
and
this
is
the
image
builder
office
hours
on
Wednesday
April
8th,
just
wanna
remind
everyone
that
this
meeting
is
recorded
and
therefore
should
adhere
to
the
code
of
conduct
guidelines
for
kubernetes
in
the
CN
CF
light
attendance
and
light
agenda
today.
So
thank
you
for
attending
Prakash
and
one
of
the
first
things
that
has
been
asked
about
is
an
open
stack
of
builder
I
wanted
to
point
out
that
one
was
merged,
and
so,
if
I
happen
to
go
too
quickly
and
I'm
gonna
check
this
out
there.
A
B
C
B
C
B
C
A
B
A
A
C
A
Yeah
sure
what
this
is
is
starting
with
cluster
API
v1,
alpha
3
and
the
VC
reimplementation
of
it,
which
we
refer
to
as
Kaffee
cluster
API
provider
vSphere.
It
requires
a
load,
balancer,
Oba
and
right
now.
The
only
thing
supported
is
H,
a
proxy
I
think
that
will
change
in
the
future
as
part
of
that
they
built
an
OVA
for
for
that
H,
a
proxy
image.
But
it
wasn't
built
into
the
image
builder.
A
It
was
actually
built
into
the
cat,
V
repo,
and
so
when
that
was
originally
created,
the
structure
to
build
that
opa
was
kind
of
copied
from
the
image
builder.
But
of
course
that
means
that
there's
a
lot
of
copying,
which
means
that
they
diverge
and
I
am
proposing
that
I
want
to
take
over
building
of
that
OVA,
because
it's
required
deploy
to
deploy
a
cluster
with
cat
V,
and
the
image
builder
has
the
intent
of
building
images
required
for
cluster
API
deployments.
A
And
this
came
up
because,
within
the
last
couple
of
weeks
there
were
lots
of
errors
or
issues
that
came
up,
because
the
tooling
for
building
images
was
different
between
the
two
and
so
I
had
to
make
commits,
and
the
two
different
repos
and
so
I
filed
its
issue
with
the
cat,
the
repo
directly
and
all
the
feedback
that
I've
gotten,
which
is
only
two
people
and
has
been
yes,
please
take
it
over,
but
I
at
least
wanted
to
bring
it
up
here.
So
someone
can
comment
on
it
as
if
there
was
a
good
reason.
So.
B
B
Yes,
so
the
octant
needs
to
connect
with
copy.
So
I
was
telling
him.
Are
you
going
to
connect
to
bare
metal
directly
or
are
you
going
to
go
through
copy
if
you
go
through
copied
and
that
will
be
better?
And
since
you
have
brought
the
image
here,
I
think
this
is
the
right
point.
There
is
some
convergence
between
the
two
and.
A
B
A
In
the
OVA
is
your
messages
at
our
file,
and
then
there
is
an
OVF
which
is
all
the
metadata
which
is
an
XML.
It
defines
defaults
for
the
virtual
machine,
specs
number
of
CPUs
Ram
disk
network,
all
of
that
kind
of
change
when
you
import
it
or
launch
it,
and
then
it
contains
the
disk,
the
V,
if
you're
doing
it
really
well,
it
also
includes
a
dot
in
F.
A
manifest
file
contains
shop
sums
for
both
of
those
files
make
sure
everything's
not
corrupted,
and
that's
it.
It's
only
three
files
is
what
we
is.
A
What
we
package
up
in
there
there's
a
couple
options
for
other
files,
what
we're
not
using
them
right
now,
but
yeah.
This
is
just
a.
This
is
really
about
convergence
and
building
you
with
the
same
tooling
in
the
original
thought
was
well.
This
is
very
specific
to
to
vSphere,
so
it
doesn't
belong,
but
since
it's
required
with
a
v1
l43
and
the
image
builder
is
supposed
to
build
images
specific
for
cloud
providers
like
I-
think
that's
where
it
belongs.
You
know
everything
in
the
adder
stuff
is
specific
to
Azure.
A
B
A
So
there
are,
there
are
two
images
that
end
up
being
used
when
you're
deploying
a
cluster
with
cap,
v4
vSphere
one
is
the
node
image
used
to
create
your
master
nodes,
your
your
control
nodes
and
your
notes.
The
other
image
is
the
the
load
balancer
node
and
right
now.
Yes,
theoretically,
could
be
any
load
balancer.
But
when
you're
using
cap
V
the
only
supportive
one
right
now
is
H
a
proxy
that
will
change,
but
that's
the
state
of
it
right
now.
So.
A
A
B
A
And
the
same
thing
on
vSphere
lease
fear
supports
a
couple
different
modes,
so
you
upload
the
OVA
and
then
you
have
to
convert
it
to
a
template
in
vSphere
and
then
from
there.
You've
got
a
couple
of
options.
If
you
want,
if
you
want
the
fastest
possible
way
to
bring
up
your
cluster,
what
you
end
up
having
to
do
is
before
you
create
a
template.
You
have
to
take
a
snapshot
of
the
disk
and
then
convert
it
to
a
template
and
then
and
your
Yamma,
you
can
specify
a
link
by
the
dog.
B
A
Will
do
a
fast
clone
the
gotcha
with
that
is
that
you
can't
change
the
disk
size
if
it's
a
linked
clone.
So
whatever
the
disk
size
is
when
you
create
your
template,
and
all
of
that
is
what
the
size
will
be.
The
the
other
way
around.
That
is,
instead
of
doing
a
clone
mode
of
the
link
glow
and
you
can
do
a
clone
of
a
full
clone
where
it
copies
the
disk.
B
I
did
see
one
of
the
OpenStack
example
confusion
just
put
in
the
chat
window.
That's
what
I
saw
the
how
they
do
it
for
creating
the
DML
file
for
the
configuration,
and
it
has
got
many
options.
Yeah.
It
has
got
many
options
good
at
the
top.
It
tells
you
how
to
create
the
cluster
cuddle
file
with
the
flavors.
B
So
when
you
create
the
first
time
itself,
the
cluster
configuration
that
includes
the
provider
that
includes
the
version
of
the
kubernetes
that
includes
the
either
you
can
use
config
map
or
you
can
so.
This
is
generic.
This
is
not
generally
not
specific
to
OpenStack,
but
can
be
used
by
any
provider.
Yeah.
B
B
A
B
It
is
the
source
right,
so
there
is
our
cat
free
template
and
how
we
pull
it
out
from
yeah,
but
this
is
worth
looking
at
thanks
for
the
information.
I
think
I
got
now.
So
the
question
is:
do
you
think
H
a
proxy
now
Suffolk
is
sufficient
for
the
load
balancing,
because
we
are
going
to
use
the
I/o
right.
B
A
B
Just
I'm,
not
at
this
time,
I'm
not
enough
knowledgeable
to
basic
concern,
I'm,
seeing
the
possibilities
that
are
there
like
bit,
because
in
VMware
they
use
NSX
T
as
the
baseline
for
telco
networks.
And
then
the
question
is
okay,
that
is,
for
the
transport
I
can
understand.
But
when
it
comes
to
again
the
instead
of
flannel,
if
I
want
to
use
calico
for
the
networking
related
to
container,
how
does
it
get
supported
in
this
because
load
balancing
of
trawl?
You
have
to
go
across
the
young.
A
Yeah
I
mean
so
a
couple
things
I
would
say,
is
a
good
I.
Don't
want
to
make
this
meeting
be
specific,
but
since
it's
just
you
and
I
we
can,
we
can
chat
about
it
most.
Your
questions
are
more
relevant
to
cat
V
clustering
provider.
Folks,
here
the
images
that
we
build
that
get
used
by
Kathy
don't
have
no
don't
have
a
like
an
overlay
or
a
CNI
plugin
installed
that
is
deployed
as
an
add-on
by
Kathy.
A
A
B
Think
calico
is
good
enough.
That's
what
I
would
go
so
basically
that
means
it
meets
the
requirement
mostly
and
also
H
a
proxy.
Then
when
you
talk
about
at
what
protocol
are
we
going
to
do
balancing
your
balancing
layer,
three
transport
layer,
four
in
terms
of
what
are
the
protocol
supported
in
HT
proxy
by
default?
That
would
be
the
question
then
Thanks
yeah.
A
B
So
the
who
will
be
the
caffee
has
to
specify
what
are
the
environment
variables
and
what
are
the
various
variables
that
we
need
to
include
as
part
of
the
generation.
So
will
there
not
be
a
requirement
for
building
the
image
from
Kaffee
asking
captain?
What
should
we
do
for
configuration?
How
is
it
now.
A
The
thing
we
think
we
need
or
the
thing
that
we
create
and
even
in
I
mean
I
mean
jump
over
here.
This
is
within
Cathy.
I
can
kind
of
show
you
like,
but
what
gets
rated
is
a
an
OVA
that
just
has
AJ
proxy
in
it,
without
really
any
configuration
and
it's
up
to
cat
B
to
create
it.
So
it's
using
this
something
called
data
plane,
API
and
it's
and
so
I
and
I
in
here
I
might
have
misspoke.
It
says
it's
remotely
configurable.
B
B
A
B
The
reason
is
where
I
am
focusing
from
cluster
API
point
of
view,
what
the
providers
need
and
specifically
I'm
focused
on
two
of
them.
One
is
OpenStack:
there
is
V
in
both
okay
since
I
work
in
open
source.
My
focus
right
now
has
been
on
the
OpenStack.
That's
why
I
asked
that
question,
and
this
is
also
being
relevant,
because
I
am
definitely
connected
with
this.
A
B
A
This
is
again
yeah.
There
was
just
something
I
wanted
to
put
out
there
to
see.
If
anybody
had
feedback,
though
one
of
the
things
that
happens
is
I
can
show
you
one
of
the
things
that
happens
in
go
to
the
right.
One
here
dig
in
the
way
that
we're
building
the
actual
OVA
is
right.
Now
is
a
little
interesting.
A
It
works
fine,
but
when
we
actually
go
to
build
an
OVA,
it's
a
Python
script
and
the
way
that
this
ends
up
working
is
we
read
in
a
bunch
of
information
from
the
Packer
build
itself
from
the
manifest
file.
There's
a
couple
things
that
we
can
override
like
where
the
disc
is,
but
in
general
it's
pretty
much
as
simple
as
we
take
an
XML
template,
fill
in
a
bunch
of
the
fields
that
need
to
be
filled
in
and
then
tar
up
a.
A
It
also
has
a
base
template,
but
you
can
feed
it.
A
ya
know,
file
that
says
here's
what
I
want
you
to
set
things
in
and
then
we
can
build
up
that
template
and
in
if
the
yamo
says,
hey
and
I
also
want
you
to
fill
this
field
like
within
the
product
section.
There's
these
properties
but
say
I,
want
to
add
a
property
in
this
product
envelope
that
isn't
there
already
there's
no
way
to
do
that
today.
A
I
have
some
Python
that
will
do
it
actually
and
I
started
to
go
down
that
road,
but
it's
just
not
as
flexible,
because
what
I
end
up
having
to
do
is
add
a
flag
to
the
Python
script
and
say
add
this
like
for
everything:
I
want
to
add
I
need
to
add
like
the
flag.
I'd,
rather
just
do
really
y'know
file
and
I
can
do
it
in
Python
as
well,
but
I'd.
A
B
B
Know
yeah,
so
it
is
not
a
performance
issue,
so
yeah,
maybe
immutability
the
the
most
of
the
pythons,
will
do
substitution
of
a
new
of
a
given
field.
Here
we
are
talking
of
adding
a
new
parameter
in
itself,
mm-hmm
like
if
I
just
substitute
the
property
with
a
value.
So
it's
a
key
value
pair,
so
replacement
of
values
is
fine,
but
if
you
have
to
add
a
key
which
is
like
changing
the
whole
structure,
yes,.
A
Manipulated
yeah
I
can't
speak
today.
You
can
change
them
with
Python
very
easily
like
it
doesn't
have.
This
doesn't
have
to
be
gone
as
a
string
template
where
we're
replacing
variables.
You
would
treat
it
as
an
actual
XML
document
and
then
export
that
document
and
write
it
out
at
the
end.
It
is
doable
and
I
have
code.
That
does
that.
That
is
unpublished
right
now,
but
it's
working.
It
was
just
more
flexibility
that
I
have
to
add
and
remove
things
a
little
bit
better
if
I
can
take
it
from
something
like
a
yamo
file.
B
A
It's
making
decision
on
it's
something
I'd
like
to
discuss
with
most
but
I
kind
of
know
where
he
falls
on
this,
because
when
I
told
him
I
was
going
to
add
a
bunch
of
Python
to
do
that
or
really
when
I
told
him
what
I
was
trying
to
accomplish.
Yes,
are
you
writing
and
go
and
I
said
no
Python
years
ago?
Okay,
he
wants
to
see
it.
B
B
Two
weeks,
every
two
weeks:
okay,
so
8
+,
7,
15,
+,
7
22.
So
when
the
next
meeting
occurs,
I
will
try
to
see
if
I
can
get
somebody
to
join.
Who
is
motivated
to
do
it
in
go,
because
if
you
have
something
in
go
a
lot
of
people,
you
want
to
do
this.
Just
for
the
language
say
that
is
one.
The
second
is
obviously
it
does
serve
the
purpose
for
us.
A
B
So
the
same
question
people
ask
and
I
I
think
if
it
is
not
a
performance
issue,
it
is
more
of
a
following
line
with
the
kubernetes
to
use
go.
Second,
is
it
has
got
some
advantage
like
you
can
easily
do
template
modifications
and
all
that
with
the
ease
of
using
and
that
those
to
itself?
So
there
are
some
pros
and
cons,
we
can
say
are
not
Const,
but
more
than
that
generally,
when
we
think
of
go,
we
think
of
performance.
B
B
A
Of
course,
I.
Don't
think
here
we
go
so
bi-weekly
on
Mondays,
so
so
now
in
it,
especially
if
you
got
questions
about
h8
proxy,
how
that's
configured
templates,
how
to
configure
templates
this
would
be
the
meaning
to
go
to
and
but
it
might
all
be
covered.
You
know
you've
always
hoped,
but
as
you're
saying
like
as
you
go
through
things
and
you
find
gaps
in
the
docs,
there
might
be
things
that
you
can
get
feedback
you
can
give
or
improvements.
But
there's
a
pretty
good,
getting
started
guide
here.