►
From YouTube: CF on K8s Working Group Forum Jan 25 2023
Description
2023-01-25
[PC] Reminder: We’d like to spend some time next week focusing on peer feedback. [Slack]
[GC] How does pushing with a manifest actually work?
Sequentially:
Apply the bulk manifest
All resources in the manifest will be created
Push sources for each app
Create build and droplet
Assign droplet
Start
Shall we improve the manifest application so that it doesn’t return before resources are created?
[GC] Mel C
[RI] Korifi blog post: Please fill in responses
A
All
right
well
so
this
first
one
I
put
on
here,
I,
don't
know
it
was
really
specifically
a
working
group
thing,
but
since
I
had
the
audience,
I
figured
I'd
just
remind
everyone
that
we
want
to
spend
some
time
next
week,
focusing
on
peer
feedback,
collecting
it
and
delivering
it
so
post
about
it
in
slack
again,
just
really
just
kind
of
a
heads
up.
B
Yeah,
just
a
very
technical
one,
so
I
am
working
on
making
money
fast
with
more
than
one
app
work
and
that's
a
really
trivial
change,
because
we've
got
everything
in
place
and
we
just
needed
to
put
in
a
for
Loop.
But
then
I
wanted
to
also
add
an
end-to-end
test
that
pushes
some
sort
of
moderately
realistic,
manifest
and
seat
working.
But
when
we
push
from
the
CLI,
you
just
push
with
the
Manifest
and
it
works.
I
just
don't
know
how
it
works
when
it
comes
to
like
API
calls.
B
So
maybe
someone
here
knows
and
saves
me.
It
can
save
me
some
time
of
research
like
because
if
I
just
apply
the
Manifest,
of
course,
the
the
apps
are
going
to
be
created,
but
I
haven't
pushed
the
sources
yet.
So
how
does
it
work?
Do
I,
apply
the
Manifest
and
then
do
I
need
to
push
the
sources
afterwards
and
then
do
the
builds
of
the
droplets
afterwards
yeah
and
but
what
happens
to
like?
If?
B
C
All
right,
so
let
me
caveat
this
with
my
knowledge.
Here
is
probably
for
at
least
four
years
old
at
this
point,
but
the
way
it
works.
Fine,
you
have.
You
have
path
keys
in
your
manifest
right
that
point
to
the
source
code
for
each
of
the
apps.
C
Okay,
you
need
those,
so
the
the
flow
the
CLI
does
is.
The
very
first
thing
is:
it
will
apply
the
bulk
manifest
and
then,
if
we're
talking
about
CF
or
VMS,
then
Cloud
controller
will
create
any
resource.
C
That's
defined
in
that
manifest
up
front,
so
it
will
go
and
either
create
or
update
every
app
and
then
create
processes
like
skeleton
processes
just
based
on
the
the
type
and
command
you
have
and
in
any
routes
or
anything
for
the
apps
and
then
after
that,
the
CLI
essentially
just
does
a
loop
over
the
apps
and
we'll
go
in
and
like
push
the
the
code
at
the
path
for
each
app
and
then
do
its
normal
flow.
And
then
it's
like
basically
just
a
loop
that
does
the
single
app
flow
for
each
app.
C
That's
in
the
Manifest
is
that.
B
Not
like
trying
to
do
it
all
at
the
same
time,
but
like
why
don't
I
see
any
processes
so
inquiry
if
I
apply
a
manifest
with
one
I
I
would
I
was
expecting
to
find
the
web
process
because
of
the
normalization
which
creates
the
web
process
by
default.
If
there
are
no
processes,
but
if
I
hit
slash
processes,
I
don't
see
anything.
Is
it
because
we
filter
out
processes
that
are
not
radio
whatever?
That
means
well
I.
A
C
C
Should
still
show
up
like
the
processes,
endpoint
should
show
everything
it
knows
about
the
processes,
including
whether
or
not
there
they
should
be
started
or
stopped.
C
It's
it's
possible
that
there's
also
like
some.
How
long
are
you
waiting
before
hitting
the
endpoint.
B
B
Right
right,
right,
okay,
thank
you,
I'll,
try,
harder,
I!
Guess
it's
gonna
be
nice
to
have
a
test.
That
actually
does
this,
because
at
the
moment
I
think
we
do
use
manifests
in
some
tests
here
and
there,
but
we
don't
have
like
a
dedicated,
manifest
end-to-end
test.
C
Yeah
I
think
this
could
be
also
a
case
for
improving
that
apply
manifest
job.
So
that
said,
it
doesn't
return
immediately,
but
it
could
potentially
like
wait
and
and
see
if
these
things
exist
or
anything
before
it
says
that
manifest
is
applied
so
like
right.
D
C
C
Because,
like
I
think
I
think
this
is
more
than
just
like
an
into
in
test
writing
issue.
It
could
be
like
a
CLI
bug.
Essentially
if
it's,
if
it's
trying
to
do
the
multi-manifest
and
it
returns
too
soon,
no.
C
B
C
C
B
E
B
E
B
E
E
To
support
it
right
now
just
to
write
the
test,
thank
you
that
is
Trivial
just
to
implement,
but
maybe
there's
something
hidden
there,
but
yeah
I
guess
just
give
it
a
try
and
and
if
there's
a
problem,
I
can
probably
look
at
the
code
see
if
it's
suffering
from
the
same,
maybe
it's
trying
too
soon.
Maybe
it's
yeah
we'll
have
to
look
at
the
details.
I
guess,
if
needed,.
A
A
B
Yes,
I
was
thinking
while
we
were
discussing
that
this
thing
is
very
convenient
to
test
on
a
like
traditional
CF,
just
to
see
how
it
behaves
and
compare
so
a
bit
of
history.
When
we
were
working
on
Garden,
the
team
needed
a
whole
bunch
of
environments
right
and
we
had
an
environment
which
we
called
spicy,
which
was
the
one
with
all
the
like
experimental
features
enabled
and
like
when
it
came
to
the
to
naming
that
environment.
B
Someone
had
the
idea
of
naming
it
after
a
Spice
Girl,
so
it
was
called
Mel
B
and
we
played
with
that
environment
even
in
with
irini
or
even
with
crify
like
when
we
wanted
to
try
on
real
CF.
We
just
try
on
Mel
B,
and
it
was
there
conveniently
and
like
even
if
it's
another
team
now
owning
it,
it
would
still
worked
for
a
while
and
then
the
other
day
we
realized.
No,
it's
been
torn
down
because
the
working
group
has
moved
the
resources
or
whatever
rotated
their
credentials,
so
we're
like.
B
Oh,
we
actually
do
need
a
like
a
traditional
CF
around
it's
just
for
convenience.
When
you
want
to
try
something
so
of
very
naturally,
the
successor
of
male
B
is
Mercy,
which
conveniently
is
also
an
existing
Spice
Girl.
So
we
have
deployed
Mel
C
in
the
corifi
working
group,
gcp
project.
So
if
you
want,
if
you've
ever,
if
you're
very
curious
of
like
how
does
traditional
CF
behave,
you
can
check
them
out,
see
I.
Think
everything
like
there
is.
There
are
scripts
to
actually
Target
it
I
think
in
the
in
our
I.
B
Think
we
put
it
in
our
secrets
in
the
creepy
Secrets
repository
Maybe,
so
you
should
be
able
to
just
Target
it
and
play
with
it,
which
is
what
we
do
like
even
for
acceptance
or
when
I'm
writing
stories
when
I'm
wondering
oh,
how
does
traditional
CF
behave
in
this
particular
instance?
I.
That's
what
I
do
I
play
with
the
existing
with
the
traditionals?
B
Yes,
and
it's
done
with
the
Bosch
light,
not
even
the
new
Bosch
light,
I
think
they're
all
deprecated
Bosch
light
just
because
we
already
had
the
scripts
lying
around,
so
it's
basically
a
VM
with
an
entire
CF
in
it
deployed
with
containers
instead
of
VMS.
So
so
it's
it's.
Basically,
a
VM
that
has
the
Bosch
director
using
a
garden
based.
B
How
is
it
called
like
backhand?
So
when
you
try,
when
Bosch
creates
VMS?
Actually
the
back
end
calls
guard
and
then
creates
containers
instead,
which
means
you
can
have
a
full-blown
like
CF
all
on
one
VM,
so
it's
cheap
and
we
can
keep
it
running
so
that
when
you
need
it,
you
can
just
you,
don't
have
to
wait
for
it
to
come
up
or
anything
you
just
you
just
it's
there.
B
A
Next
panel,
wrong.
D
Hey
so
I
just
wanted
to
Circle
back
to
that
topic
that
we
were
speaking
about
last
week
and
basically
doing
that
top
post.
Thank
you,
Tim
for
being
the
Lone
Star
in
our
lineup,
who
I
am
asking
those
comments
on
time,
but
if
other
folks
want
to
pitch
in
as
well
and
and
have
your
thoughts
heard
and
have
something
to
share
with
the
community
in
terms
of
the
great
engineering
effort
that
you're
a
part
of
I,
really
invite
folks
to
add
something
to
their
talk.
D
So
if
you
could,
please
take
some
time
today
or
tomorrow,
time
box
it
if
you
can
and
fill
in
some
stuff
in
the
talk
about
what
you
have
in
terms
of
feedback
or
thoughts.
I
would
really
appreciate
it.
D
If
you
have
any
thing
you
want
to
discuss
I'm,
obviously
available
on
slack,
you
can
ping
me
and
we
can
chat
about
it
before
filling
it
into
the
top.
So.
D
Help
anyone
then
I'm
available
for
that
as
well.
So
just
a
reminder
for
folks
to
contribute
in.