►
From YouTube: App Runtime Deployments Working Group [April 27, 2023]
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
Let's
go
over
the
notes
from
the
last
meeting,
so
we
started
with
the
Sea
of
Linux
fs4
integration
step,
one
which
is
I
think
now
completed.
So
we
have
now
fs4
as
the
default
stack
in
CF
deployment
Jamil
without
any
additional
Ops
files.
A
There
is
an
Ops
file
to
make
fs3
the
default
stack
again.
It's
used
here
in
the
bionic
stem
cell
validation
pipeline,
and
we
also
have
left
the
existing
experimental
Ops
files
in
place
just
with
an
empty,
except
for
a
a
short
comment.
Yeah
yeah
that
was
released
as
incompatible
major
version
28..
A
B
A
At
all,
it's
Jenny
with
bionic
stack
right
should.
C
A
B
A
Okay,
we
discussed
the
CEO
of
Linux
fs4
compatibility,
change
or
RFC,
not
our.
You
know,
change
RFC,
so
we
would
like
to
have
a
compatibility
stack
with
the
python
and
Ruby
binaries
included,
which
makes
the
transition
easier.
A
And
yes,
this
has
been
accepted
by
the
talk
and
they
are
starting
to
set
up
the
infrastructure.
Now,
a
new
repository
and
well
sooner
or
later,
the
compatibility
stack
should
appear
on
Bosch.
I
o
so
oh
grows
fast,
okay,
so
this
is
it
fantastic?
A
A
So
okay,
this
is
really
old.
So
when
we
we
have
here
a
working
version,
we
would
include
this
in
CF
deployment.
That
is,
provide
a
job
which
automatically
updates
that
new
Ops
file,
so
that
you
can
opt
in
to
use
the
compatibility
release
and
validate
it
in
one
of
our
scenarios
should
be
rather
easy.
So
in
the
RFC
we
discussed
a
few
other
possible
solutions
with
the
efforts
for
the
different
working
groups
and
yeah
I.
Think
for
us
this
is
I,
think
the
best
solution.
A
Yeah
good,
so
right
now,
there's
nothing
to
do.
We
have
a
story
to
integrate
this
when
it's
available
so
Stefan
in
case
you
have
missed
it.
The
the
the
the
scaffold
is
already
there.
D
Actually
contributed
it.
A
B
B
A
Good
yeah
cats
test
will
come
to
that
later
because.
A
A
A
That's
why
several
jobs
turned
red,
so
I
tried
to
summarize
everything
in
this
issue.
Where
is
my
chat?
Yeah?
Okay,
then
paste
it.
A
A
So,
basically,
the
the
state
files
from
our
existing
bubble
environments
are
not
competitive
anymore,
with
terraform
146.
A
A
So
these
are
some
of
the
incompatible
changes
I
already
found.
For
example,
terraform
override
files
are
not
accepted
anymore
because
they
really
have
to
overwrite
something
specified
in
the
base.
Terraform
file,
which
we
did
not
do,
but
here
is
simply
renaming,
is
sufficient.
A
A
So
these
are
the
the
configuration
files
which
we
use
in
so
wait.
Let
me
show
so
that
everyone
knows
what
I'm
talking
about
so,
for
example,
for
triloni
for
the
upgrade
scenario
we
are
using
an
external
database
and
blob
stores
and
these
files
had
to
be
renamed
and
also
the
syntax
had
to
be
changed
a
little
bit.
A
What's
a
bit.
Trickier
is
the
recreation
of
the
environments,
because
you
cannot
do
a
bubble
destroy
with
the
new
terraform
version
on
the
Old
State.
So
this
is
the
workaround
I
found.
A
You
have
to
use
the
an
old
version
of
cfd
program
and
Concourse
tasks,
for
example,
version
14,
which
contains
terraform
13,
something
and
bubble
eight
nine.
Then
you
can
run.
A
A
Then
you
can
run
destroy
and
then
you
change
it
back
and
then
you
can
run
bubble
up
or
maybe
we
should
use
two
Concourse
resources
of
this.
The
old
one
for
destroy
the
new
for
Bubble
Up
is
this:
did
you
understand
what
I
mean.
B
Yeah
yeah,
that
makes
sense
you
could
also
just
pin
the
resource
in
the
The
Concourse
pipeline
View.
A
Yeah,
okay,
but
with
this
procedure
we
should
be
able
to
migrate
all
the
long-running
running
environments.
So
trelawny
is
migrated.
If
you
want
to
pick
another
one
feel
free
to
do
so.
I
have
paused
the
main
pipeline
for
now,
because
it
simply
will
not
work
or
some
tasks
using
tariff
use.
Tasks
using
bubble.
Outputs
will
not
work
at
the
moment
like,
for
example,
oh.
A
This
one
is
here
this.
D
A
Next
thing
we
need
to
resolve
mm-hmm,
so
this
calls
bubble
outputs
which
interpret
some
of
the
terraform
files
to
extract
some
data,
and
now
it
says
plugins
are
installed,
which
is
true.
If
you
do
terraform
providers
in
this
container,
you
see
that
it
is
not
installed.
I,
don't
know
exactly
how
this
worked
before.
A
A
B
Yeah
I
think
that
makes
sense,
hi
Carson.
B
Yeah
running
terraforming
it
prior
to
running
to
General
from
outputs
doesn't
seem
to
problematic.
A
And
it
should
be
okay,
it's
just
a
few
special
tasks
which
are
affected:
yeah,
okay,
so
yeah
over
the
next
days.
We
should
try
to
migrate
the
remaining
environments
and.
A
Yeah
note:
the
next
tricky
thing
is
that
some
have
already
been
touched
by
a
new
by
the
new
terraform
version,
and
then
you
get
I
know
this
one.
Then
these
double
quotes
are
removed
and
then
it
cannot
be
interpreted
anymore
by
the
old
terraform
version
of
IC
version.
A
So
you
have
to
carefully
check
if
the
state
is
not
half
migrated
and
I
had
to
do
a
git
reward
in
some
in
in
the
trilani
environment
to
get
it
back
to
the
old
State,
probably
tear
it
down
with
the
old
tools
and
then
set
it
up
again,
and
then
it
works.
C
Makes
sense
is
I'm,
sorry,
I
missed
the
beginning
of
this.
Also
sorry,
I'm
late
I
miscalculated
the
time
again
that
the
event
is
still
wrong
in
my
calendar
is,
is
this
a
bubble,
terraformed
version
or
a
terraform
version?
That's
messing
us
up
like.
Is
it
the
terraform
version
in
our
image,
or
is
it
the
bubble
form
both.
C
B
So
they
have
decided
to
switch
back
to
Maine
from
the
the
8.4,
like
maintenance
branch
that
was
being
maintained
previously
and
they've
they've
just
cut
their
first
release
from
that
from
Maine.
Oh.
C
D
B
The
latest
terraform
everything's
now
up
to
date
and
it's
great
for
the
for
Bubble,
but
it's
a
huge
breaking
change
for
everyone
that
uses
it
right.
B
Recreate
them
right,
that's
what
we're
talking
about
is,
unfortunately,
because
I
think
you
and
I
a
month
or
two
ago
changed
the
pipeline
to
Auto
bubble,
rather
than
just
it
being
manual.
We
just
consumed
this
and
it
caused
chaos
so
we're
having
to
roll
bits
back
so
that
we
can
destroy
things
and
then
we
can
roll
it
Forward
again,
so
that
they
could
be
stood
up.
C
Yeah
yeah,
okay,
interesting
I,
saw
I,
did
see
the
failures
yesterday,
I
assumed
version,
I
didn't.
A
Realize
we
got
a
job,
but
no
yeah.
C
I
I
just
wanted
to
get
that
job,
yeah,
green
and
then
yeah.
A
Yeah
yeah
yeah,
usually,
but
no
so
here's
the
the
migration
procedure
condensed
in
a
few
lines.
So
if
you
like,
you
can
start
to
migrate
to
remind
remaining
environments.
If
not
then
yeah,
we
can
also
do
so.
A
Yeah,
it's
a
bit
tricky
because
you,
if
there
should
not
be
two
people
working
in
parallel
on
the
same
pipeline,
because
then
you
overwrite,
if
we
do
flyset
pipeline
two
different
people.
At
the
same
time,
this
will
be
important.
A
A
A
B
A
A
Things
look
already
a
lot
better,
so
maybe
could
it
be
that
the
cat's
suit
was
has
some
glitches
that
do
not
allow
two
catsuits
to
be
run
in
parallel
or.
B
A
A
There
was
just
one
thing:
I
did
not
really
understand.
Let
me
see
of
taking
some
notes
is.
B
A
C
So
agree
with
Dave
that
probably
running
two
two
cats,
both
at
12,
like
processes
is
was,
is,
is
probably
hitting
the
cfd
at
like
double
the
amount
that
we
expect
to
be
hitting
in
a.
A
C
A
Resource
which
is
used
by
both
and
depending
on
the
order
of
the
tests,
yeah
okay,
if
they
stay
green,
just
one
thing
about
this
timeouts.
A
B
B
A
Careful
this
is
the
maximum
amount
of
time.
When
you
start
a
CF
push.
I
think
this
is
for
the
CF
CLI.
A
B
D
A
A
D
If
I,
remember
correctly,
in
Cloud
controllers
of
the
health
check,
timeout
is
the
ones
that
you
configure
that
an
app
has
the
time
to
start
up
right.
It's
the
initial
so
for
Java
application
to
get
the
VM
running
Etc
and
the
maximum
hatch
check
time
out
is
the
upper
bound
that
a
end
user
can
configure
via
the
Manifest.
So
you
cannot
configure
a
value
bigger
than
180
seconds.
I
mean.
A
D
Our
Landscapes,
we
have
set
that
to
10
minutes
and
the
default
is
one
minute.
So
if
you
just
push
an
app,
you
have
to
say
60.
D
C
Yeah
I
I
had
a
question
about
that,
because
I
saw
that
the
latest
change
for
the
build
pack
cash
test
was
that
I
was
that
someone,
this
might
be
out
of
date,
the
one
you're
looking
at,
because
I
think
someone
added
I
think
so.
A
hard-coded
timeout
of
like
of
in
into
one
of
these
tests.
C
Yeah
I,
if
you
check
the
yeah
I,
don't
know
if
you
check
the
the
latest
on
on
Main
I,
think
that
that
test
includes
a
hard-coded
timeout
now
yeah.
A
A
C
Cappy
folks
started
the
thread
saying
hey
like
this
test
is
flaking,
and
then
we
started
looking
into
it
and
I
started,
dropping
it
into
a
lot
of
channels,
because
I
had
no
idea
what
was
going
wrong
and
it
ended
up
somewhat
identified
it
as
a
go
router
problem,
because
because
of
the
way
that
the
NC
thing
is
set
up,
it
doesn't
actually
it
responds
this.
The
NC
server
responds
so
quickly
that
the
client
doesn't
actually
get
an
opportunity
to
like
send
a
request.
C
It
just
receives
the
response
immediately
and
that
was
like
erroring
out
on
the
go
router
side,
somehow
and
I
think
go
router
folks
released
a
fix
for
that
in
the
latest.
Router
routing
release,
so
I
thought
that
was
actually
what
stabilized
it.
A
A
A
C
I
just
dropped
the
cloud
Foundry
thread
into
or
into
Zoom
I
mean.
C
Already
yeah,
look
if
you,
if
and
263,
had
its
own
problem,
so
we
didn't
actually
get
that
until
264
came
out,
but
when
264
has
come
out
and
we
got
that
one
but
the
yeah,
it
was
something
like
all
the
all
there's
like
an
actual
go
issue
where,
if
the
server
responds
before
the
before
the
client
actually
has
a
chance
to
send
something
which
the
NC
server
the
like
custom,
little
NC
server
does
then
go
just
like
fails
it's
it's.
It
thinks
something's
wrong.
C
Okay
and
yeah
I
think
Greg
Cobb
was
able
to
reproduce
it
with
like
a
little
app
like
test
app
and
the
go.
Router
folks
tried
to
push
a
fix
for
it.
So
I
I
don't
know
for
sure.
If
that
was
what
stabilized
the
test
or
whether
it
was
the
timeout,
but
it
would
be
interesting
to
try
ripping
out
that
time
out
again
and
see
if
the
tests,
okay,
still
work.
A
Time
out
of
good
news,
so
let's
do
so
and
then
we'll
see
if
it's,
if
it's
changing
anything
or
not,
okay,
but
that's
cool
I,
didn't
notice
that
they
really
tracked
it
definitely
down
and
released.
A
new
version
in
this
263
was
broken
with
yeah
yeah
right
now,
I,
remember
the
the
boss
release
was
broken
somehow
and
couldn't
compile
anymore
yeah.
B
A
Okay,
good,
then
last
point
for
today,
unless
you
have
something
is
renovate
for
our
Concourse.
A
So
this
is
our
Concourse
blueprint
project
from
the
runtime
interfaces
working
group
and
they
have
put
renovate
in
place
which
searches
for
the
Tarak
Grant
files
and
updates
yeah.
Well,
all
kinds
of
artifacts
and
yeah
question
was:
if
we
should
just
take
over
this
approach
to
our
con
course.
A
A
Don't
possibly
yes,
but
I,
don't
know,
so
we
at
sap
were
also
using
renovate
heavily
in
a
very
large
project
and
also
very
specific,
a
sophisticated
configuration
to
get
everything
working
properly
and
we
are
well
if
that
works
for
them.
We
could
just
take
over
the
approach.
Yeah.
B
Definitely
I
mean
I,
I've
used
it
as
well
on
on
projects
depend
about.
I
was
just
curious
because
I'm,
assuming
that
dependable,
is
automatically
scanning
these
things,
as
opposed
to
renovate,
where
I
think
you
have
to
run
a
job
to
do
the
scan
to
force
it
to
scan
every
time.
A
C
A
Quite
similar
yeah,
we
compared
the
two
and
yeah
each
has
his
pros
and
cons,
but.
C
A
Okay,
yeah
I
mean
this
depended.
The
renovate
stuff
is
probably
not
so
difficult,
but
of
course
you
have
to
reapply
a
Terra
Grant
and
that
all
from
time
to
time
to
to
get
the
changes
to
our
Concourse-
and
this
is
then
likely
the
challenging
part
okay
good.
But
then
we
would
take
over
the
solution
and
see
how
it
works.
C
D
A
D
A
D
C
C
Some
some
number
of
pipelines
broke.
It
was
an
easy
fix.
It
was
just
switch
branches,
but
if
we
for
those
kind
of
big
changes,
it
might
be
nice
if
we
cut
a
cat's
release
at
the
same
time.
So
the.
D
C
Is
tracking
those
two
latest
releases
also
against
the
the
corresponding
cats
change.
A
A
C
And
they
they
probably
shouldn't
be
tracking
it.
That
way
and
I
think
we
tried
to
make
that
clear
that
tracking
the
version
Branch
or
the
cfd
version
they're
using
or
tracking
release
candidate
is
probably
smarter
for
for
their
pipelines.
Yes,
but.
D
D
A
A
D
Yeah
yeah
yeah
yeah.
It
has
time
there
was
even
an
announcement
by
canonical
that
1804
will
get
one
month's
more
support
for
whatever
reason.
D
D
D
D
D
A
D
See
yeah
we
to
have
a
fixed
version
would
help
also
to
make
it.