►
From YouTube: Carvel Office Hours - March 11, 2021
Description
Carvel Office Hours - March 11, 2021
Topics included:
https://app.zenhub.com/workspaces/carvel-backlog-6013063a24147d0011410709/issues/vmware-tanzu/carvel-kapp/11
https://github.com/vmware-tanzu/carvel-ytt/issues/260
as well as in-depth discussion around a question from a community member
Details and notes here: https://hackmd.io/5Bh2IXwTShSrA0YdBY4AGg
A
All
right
welcome
everyone
to
the
first
edition
of
carville
office
hours,
we're
going
to
be
starting
these
every
second
in
fourth
thursday
of
the
month
at
11
30
a.m,
pacific
time,
2,
30,
p.m.
Eastern
time,
if
there's
anything
that
you
wish
to
discuss
or
if
there's
anything
that
you're
curious
about
for
cargo,
we
invite
you
to
attend
these
and,
and
that's
just
a
open
discussion
and
just
open
help
with
anything
that
that
you
want
to
bring
to
the
carnival
team.
B
Helen,
do
you
want
to
drive,
or
do
you
want
me
to
drive?
I
don't
think
aaron's
here.
I
think
we
can
start
with
the
cap
issue
that
we
have
listed
on
the
hackmd
there.
So
it's
a
cap
issue
number
11.
If
you
pull
that
up,
it's
about
providing
a
way
to
add
custom
rating
functionality
to
four
resource
types,
and
I-
and
I
think
we
still
have
you
know
pending
discussions
on
this-
to
decide
on
a
you-
know,
approach
how
to
solve
this
particular
problem.
So.
C
B
It
open
for,
I
guess,
the
engineers
to
talk
through
this
particular
issue
here.
Okay,
you
said
start
by
the
top
one:
the
cap
11.,
yes,.
D
I
can
I
can
help
talk
to
this
kind
of
tee
off
what
what
what
the
need
is
here
and
what's
going
on
so
in
general
kubernetes
wants
you
to
report
the
status
of
your
resources
in
a
particular
way,
but
I
think
like
to
keep
it
short,
like
largely
not
everything's
doing
that,
and
so
the
question
becomes
well.
D
How
do
you
help
out
the
operator
when
they're
dealing
with
a
third-party
component
that
isn't
behaving
the
way
that
it
should
to
spec
and
give
the
give
the
person
who's
at
the
switch
a
way
to
try
and
reconcile
that?
D
So,
if
the,
if
the
status
is
not
a
condition
status
like
what
can
they
do?
How
can
they
like
point
at
well,
it's
actually
over
here.
This
is
this
value
over
here
or
maybe
on.
Another
resource
would
be
a
sufficient
proxy
for
me
to
be
able
to
indicate
that
a
particular
resource
is
available.
D
D
E
One
kind
of
a
concern
I
have
this
actually
I'll
get
to
that
in
a
second
one
concern
I
have
here
is
you
know,
because
we
are
expanding
our
let's
say
ways
of
checking
on
stuff
stuff
right
so
right
now
we
only
check
on
conditions
right
and
you
get
to
specify
what
condition
means.
What
kind
of
a
thing
right!
E
E
E
If
this
field
x
is
value
y-
and
I
don't
know
some
other
field
is
something
else
then
do
something
right,
and
what
to
me
that
kind
of
signals
to
is
that,
because
this
is
a
more
advanced
case
scenario,
we
should
probably
consider
providing
a
solution
that
that
solves
all
of
the
possible
cases
in
bulk
right,
and
by
that
I
mean
a
fairly
specific
approach
and
that
approach
is
really
just
providing
programmatic
access
to
defining.
What
does
that
mean
for
this
resource
to
be?
E
I
don't
know
reconcile
successfully
or
something
like
that
right
and
I
think,
from
previous
conversations,
I've
had
with
various
folks
one
option
that
has
that
clear:
full
potential
is
ability
to
specify
a
function
like
a
ytt
function
that
describes
what
does
that
mean
for
a
resource
to
be?
You
know
complete
now.
One
interesting
thing
about
this
is
that
you
know
this
is
not
ytt
function
and
ytt
template.
E
Right
that
are
very,
are
going
to
be
extremely
hard
to
implement
in
any
kind
of
a
yaml
structure
and
thing
right
because
adap
at
some
point
you
start
if
actually
representing
code
as
yaml
right-
and
I
I
think
that's
that's
a
very
kind
of
that's
a
very
difficult
and
not
great
territory
to
be
in
where
suddenly,
your
yaml
starts.
Looking
like,
I
don't
know,
if
conditionals
and
structures
and
and
whatever
so
yeah.
E
So
that's
one
of
my
opinions,
there's
actually
maybe
semi-related
topics
that
there's
actually
other
places
where
I
think
this
kind
of
functionality
will
be
beneficial
to
cap
and
and
actually,
yes,
date
came
up
as
one
of
the
possibilities
in
the
image
package
where
we
could
be
using
the
execution
environment
in
a
particular
context.
The
other
example
in
cap
might
be
is
deciding
how
to
programmatically
rebase
something
right,
because
today
the
rebase
is
very
basic.
E
It's
like
either
pick
from
one
location
or
pick
from
another
location
right
versus
it
could
be
a
little
bit
more
complicated
right
where,
for
whatever
reasons
you
may
want
to
rebase
based
on
some
other
configurations
or
conditions
like
that
right
and
so
again,
providing
the
ytt
execution
environment
would
be
very
handy.
There.
D
I
think
I
think
one
thing
there
could
be
taking
like
vetting
an
approach
like
that
would
be
like
taking
the
expression
that
dustin
had
here,
for
example,
and
looking
at
to
see
what
would
that
look
like
if
that
were
the
execution
context,
so
like
how?
How
would
this
json
path
translate
to
a
a
ytt
expression.
E
Yeah-
and
I
could
imagine
you
know
just
kind
of
you
know,
roughly
speaking-
there
would
be
a
function
that
takes
that
resource
which
could
be
our
yaml
fragment
or
something
right
and
you're
able
to
access
whatever
you
need
to
access
and
use
conditionals.
If
you
need
to
use
conditionals
and
later
on
return,
a
boolean,
whether
something
is
or
maybe
it's
not
even
a
boolean,
you
know,
maybe
it's
actually
a
triple
value
kind
of
a
succeeded,
failed
and
uncertain.
Yet
right
with
additional
messaging
capabilities
where
you
can
actually
provide
something
useful
to
the
user.
D
E
E
E
You
know
into
a
configuration,
but
the
point
that
I'm
trying
to
make
here
is
or
really
I'm
trying
to
answer.
The
question
here
is
the
there
is
a
case
for
deployments.
I
think
where
it
actually
waits
for
replica
sets
or
something
like
that.
I
forget
what
it
is
right,
and
so
it
actually
looks
at
a
different
object
to
kind
of
get
a
better
view
on
what
it
is
now.
E
E
But
but
there
are
other-
you
know,
controllers
out
there
that
you
know
don't
necessarily
indicate
status
via
conditions
right.
They
indicate
status
via,
maybe
some
like.
Actually,
I
guess,
technically
stateful
set
doesn't
have
conditions
right
because
it's
such
old
resource,
so
they've
never
added
conditions.
It
has.
You
know
logic
about
like
based
on
the
status
that
this
number
status,
that
another
number
and
how
that
kind
of
compares
to
each
other.
E
Or
another
example
might
be
services.
Services
have,
I
think,
a
particular
field
set,
and
if
that
particular
field
is
not
set,
then
it's
waiting
for
a
particular
other
field
to
show
up
in
the
status.
I
think
it's
like
a
low
bounce
or
ap,
or
something
like
that,
so
that
logic
is
already
hard
coded
in
cap,
but
it
would
be
interesting
to
be
able
to
pull
that
out
as
a
configuration
thing
just
to
say
that
hey
we're
not
we're
not
really
coupled
like
whatever
we
need
to
do.
F
That
it's
like
the
ytt
looks
interesting,
but
in
the
end
it
feels
like.
Maybe
it's
too
much
to
allow
people
to
have
starlark,
because
what
are
we
expecting
them
to
wait
on
right?
So
we
talked
about
that.
We
expect
them
to
wait
on
something
particular
on
a
particular
resource
right,
like
some
field
of
a
particular
value,
and
that's
basically
it
right
or
to
wait
on
a
set
of
particular
values
on
a
particular
resource.
E
So
here
is,
like
the
logic
oh
come
on.
Why
is
my
zoom
so
slow
today
here?
Here's
the
link
in
zoom,
chat
or
logic
that
the
the
service
itself?
You
know
this
is
how
we
wait
for
a
service
right,
so
it's
not
only
waiting
on
anything.
It's
really
a
combination
of
things
and
some
of
those
things
are
conditional.
E
They're,
multiple
actors
in
the
system
right
and
they're,
ultimately
working
with
the
same
data
structure
right
conditions-
is
definitely
a
preferred
way
for
folks
to
describe
what
you
know
what
to
wait
for,
but
I
think
once
once
you
once
you
go
beyond
condition,
it's
really
there's
all
kinds
of
ways.
E
Right
like
you
could
you
could
probably
come
up
with
lots
of
different
examples
out
there
and
open
source
how
people
you
know
start
waiting
or
or
deciding
to
wait
for
certain
things
right,
and
it
also
might
be
very
much
environment
specific
right
like
for
some
customers.
They
want
to
do
very
particular
thing
for
a
particular
resource
compared
to
like
how
we
would
want
to
wait
for,
for
some
other
thing,.
F
Like
the
the
only
thing
that
I
was
thinking
before,
was
that
what
happens?
If
you
have
a
nor
statement
right-
and
I
think
that's
the
part,
that's
going
to
become
more
complicated
is
is,
if
you
on
your
waiting
rules,
you
have
ends
and
ors.
At
the
same
time,
that's
like
the
case
that
you
provided
with
the
service.
It's
just
like
a
a
bunch
of
ends
and
ores
right.
F
E
File
I
mean
this
is
this
is
what
I
kind
of
started
with
right
like
I
don't
think
we
want
to
be
creating
programming
language
in
yaml,
since
we
already
know
where
very
well
that
you
know
there
is
it
it
takes.
It
takes
effort,
and
we
also
have
already
a
good
investment,
for
example,
in
starlark
in
inside
ytt,
that
we
could
just
repurpose.
D
I'm
thinking
about
that
move
of
where
it's
like,
okay,
there's
complexity
in
the
code,
sort
of
like
a
core
principle
of
like
simplify
the
code
by
pushing
complexity
into
data,
is
sort
of
one
way
to
think
about
about
the
waiting
logic
as
a
separate
expression
and,
and
so
would
simplify.
This
would
basically
be
like
get
me
the
rule
for
a
service
and
evaluate
that's
what
this
code
would
end
up,
looking
like
in
fact,
sort
of
more
generally
it'd
just
be
for
that
service,
evaluate
its
waiting.
D
Logic
would
have
extracted
this
as
a
default
rule,
and
so
I
think
I
feel
like
I
feel
like
looking
at
that
way
it.
It
basically
sets
up
the
general
spirit
of
the
request,
which
is
hey.
I
have
this
thing.
That's
just
outside
of
what
the
kubernetes
api
spec
calls
for.
D
And
and
need
to
be
able
to
break
the
glass
to
work
around
that
so
as
long
as
it
doesn't
make
cases
like
the
simple
cases
painful
like
hard,
then
you
kind
of
get
the
best
of
all
worlds
like
the
tool
becomes
more
focused
and
simple
in
its
own
right.
D
Those
rules
are
then
more
general
and
then,
as
a
a
user.
If
I
absolutely
need
to,
I
have
the
ability
to
narrowly
express
the
weight
condition
or
more
broadly,
just
depending
on
my
needs.
E
I
like
that
that
you
mentioned
a
phrase:
break
the
glass,
because
this
is
to
a
certain
extent
similar
to
you
know,
to
like
ytt
overlays
right,
ytt
overlays
themselves
provide
that
advanced
capability
for
end
users
to
say
you
know
what
I
know
better.
This
is
a
very
special,
maybe
case,
I'd
like
to
I'd
like
to
do
the
thing
that
I
know
I
want
to
do
to
me.
This
is
kind
of
similar.
E
This
is
like,
where,
to
a
certain
extent,
once
you
kind
of
move
off
the
conditions
right,
you're,
just
in
open
land,
and
so
then
really
one
of
the
questions
then
just
comes
up
is
like.
Why
would
we
try
to
prevent
a
user
from
setting
up
whatever
logic
they
want
to
set
up
right?
Why
wouldn't
we,
let
them
just
say,
do
whatever
you
want
to
do
right.
D
Yeah-
and
I
I
don't
think,
there's
anything
that
prevents
like
if
you
have
that
more
generalized
mechanism
in
place
just
like
using
the
atm,
you
can
get
as
much
cash
as
you
want
within
limits,
but
there's
also
that,
like
give
me
the
40
bucks
fast
button,
so
we
could
also
provide
the
ability
where
it's
like
hey.
You
have
this
like
really
common
case
of
where
it's
just
another
field
on
the
same
resource,
and
you
just
need
to
be
able
to
provide
an
expression
to
do
that.
D
We
could
also
provide
a
convenience
eventually
provide
like
a
convenience
interface
for
that.
If
that
becomes
like
a
common
that
we
see
like.
Oh
that's
a
regular
thing.
Why
should
you
have
to
go
through
the
whole
learning
curve
to
like
figure
out
what?
What
what
is
this
world
you're
participating
in?
If
we
can
just
provide
you
a
string?
D
I
don't
like
the
way
that
dustin
illustrated
here
with
the
like
the
field.
Query
figure
out
how
to
work
in
a
more
convenience
ui.
On
top
of
it.
E
D
Almost
I
don't,
I
don't
think
it's
exactly.
I
think
it's
more
like
the
buddy
who
goes
into
the
store
to
get
the
money
for
you
like
you,
give
them
your
card
and
the
pin,
and
then
they
go
do
it.
You
come
back
and
they
got
money.
This
is
where
it's
like
it's
case
right
after
that,
where
it's
just
like,
doesn't
quite
align
it.
You
need
to
get
involved.
G
I
think
I
just
wanna
say
a
couple
things
real
quick.
G
I
only
got
a
few
minutes
and
I
got
a
drop
for
a
customer
call,
but
I
think
the
intent
to
this
was,
although
status.conditions
is
a
known
and
documented
best
practice,
even
an
upstream
project
like
cubebuilder,
which
a
lot
of
controllers
are
built
off
of,
did
not
introduce
status,
dock
conditions
and
until
about
six
seven
months
ago,
so
anything
prior
to
six
or
seven
months
ago
is
going
to
have
potentially
problems
with
status
up
conditions
in
the
actual
field
in
the
real
world,
understood
that
yes,
status,
dot
conditions
is
a
known
best
practice
at
this
time,
but,
like
there's
still
stuff
running
out
there,
that
people
aren't
going
to
be
able
to
change.
G
G
It
doesn't
have
to
be
json
path,
but
that
was
just
the
first
thing
that
came
to
mind,
but
just
some
simple
string,
query
that
that
I
could
put
in
and
get
and
be
able
to
tell
it
that
this
is
the
correct
condition
that
this
thing
needs
to
be
in.
So
those
are
the
two
thoughts
off
the
top
of
my
head.
Just
coming
from
an
outside
consumer
perspective,.
G
From
what
I've
generally
seen,
it's
a
field
on
the
same
resource
like
I
need
to
wait
for
you
know
I
got
a
custom
resource
definition.
I've
created
that
and
now
there's
some
sort
of
field
on
that
same
resource
that
I
can
look
for
so
that
that's
generally
my
experiences,
but
I
can't
say
that
that's
going
to
cover
all
of
them
either.
E
I'm
having
a
hard
time,
believing
that
cube
builder
well,
I'm
actually
not
quite
sure
how
cube
builder
would
even
play
into
this
right,
because
cube
builder
is
just
setting
out
a
bunch
of
you
know
stuff
right
for
for
your
operator.
I
I
don't
think,
there's
anything
that
you
builder
needs
to
do
to
enable
use
of
conditions,
because
it's
just
a
random
field
on
a
status
like
it's
not
in
any
part,
it's
not
special
in
any
way
compared
to
all
the
other
fields.
So
I'm
not
sure
there
is.
E
Actually
you
know,
I'm
sure
there
are
other
controllers
out
there.
That
just
happen
not
to
use
conditions,
but
conditions
has
been
like
part
of
the
convention
in
kubernetes
for
for
a
long
time
right.
So
I'm
not
sure
how
common
this
is
to
to
not
have
conditions
now,
but
even
so,
even
if
we
go
it's
it's
it's
it's
common
right.
If
that's
common,
that
probably
means
that
there's
other
cases
that
are
also
common,
like
the
services
case.
G
F
I
think
what
doesn't
mean
is
that
previously,
if
you
wanted
to
use
conditions,
you
had
to
use
the
duck
whatever
package
that
I
believe
came
from
the
k
native
controller
or
something
like
that,
that
contained
the
structure
and
that
I
believe
q
builder
now
has
the
structure
as
part
of
q
builder.
I
think
that's
just
the
reference
that
dustin
is
talking
about.
G
G
Anyways,
I
have
got
the
drop
for
a
customer
call,
but
if
I
need
to
hop
on
or
something
or
answer
some
questions
offline
from
a
consumer
perspective,
I'm
happy
to
do
that.
So
helen.
You
can
reach
out
to
me
or
if
I
miss
anything.
D
Thanks
for
being
here
and
providing
that
additional
context,
dustin
appreciate
it.
A
So
we've
spent
quite
a
chunk
of
time
on
on
this
discussion
topic.
I
would
like
to
move
on
to
other
items.
Where
do
we
wanna
leave
this
particular
item
there?
Any
action
items
moving
forward,
or
is
this
something
that
still
needs
to
be
discussed
at
length.
D
E
Yeah,
I'd
like
us
to
I'd
like
us
to
I'd
like
to
see
a
proposal,
maybe
in
a
more
fuller
form,
about
what
are
the
possible
approaches
and
maybe
yeah.
D
If
I'm
hearing
you
right,
it's
extracting
out
the
evaluate
starlark
functionality
out
of
ytt
like
exploring
well,
if
we
could
do
that,
and
now
the
carvel
suite
could
could
benefit
from
having
that
capability
in
multiple
tools
capping
one.
What
could
that
look.
E
Like
I'm
not
sure,
if
you're
a
man
like
extracting
like
code
wise,
I
actually
think
that
the
current
ytt
packages
are
usable
as
this,
for
this
kind
of
a
case,
sure
what
I
was
maybe
leaning
more,
is
just
trying
to
think.
Where
are
various
places
within
our
tools
that
would
benefit
from
this
break
the
glass
functionality
right
like,
for
example,
this
would
be
a
possibility
that,
like
potentially
more
advanced
rebasing,
would
be
a
possibility
image
package
copy
with
rename
would
be
one
there
might
be.
E
F
F
So,
in
the
end,
I
think
we
should
focus
this
also
on
the
users,
because
they
are
going
to
be
the
ones
that
will
benefit
more
from
this,
and
if
we
think
that
this
could
be
a
good
thing
to
add
internally
to
make
cap
like
do
other
things
and
just
extract
some
logic
that
we
currently
have
in
go
to
starlark.
It
might
be
a
good
idea,
but
I
think
we
should
separate
those
two
concerns
and
see
if
what
our
users
need
really
require
an
ytt
execution
environment.
F
E
I,
I
think,
there's
a
balance.
I
I
think
we
need
to
be
very
conscious,
because
I
also
don't
want
to
end
up
with
like
five
different
ways
how
you
can
wait
on
and
now
we
have
this
well.
We've
actually
have
a
very
fragmented
experience
of
how
to
wait
on
things
right
and
this
this.
This
is
also
like
true,
for
example,
like
with
an
image
package.
We
were
talking
about
having
like
multiple
different
ways
to
match
things.
E
We're
talking
about
that
advanced
user
or
at
least
advanced
use
case,
if
not
advanced
user
right,
and
I'm
not
sure
we
should
be
optimizing
that
five
percent
case
compared
to
like
just
providing
that
break
glass
functionality
and
saying
well,
if
you
are
in
that
five
percent,
you
gotta
put
in
a
little
bit
more.
You
gotta
put
in
a
little
bit
more
effort
in
actually
describing
that,
because
it
it's
such
a
special.
A
A
Sounds
like
what
the
consensus
really
is
is
just
writing
a
proposal
and
then
going
from
there
on
just
different
possible
approaches.
A
I'm
actually
going
to
just
skip
over
the
triage
help
for
now,
because
I
want
to
get
to
tyler
and
go
over
him.
I
want
to
make
sure
we
have
time
for
him
so
tyler.
A
If
you
want
to
just
go
over
what
it
you
know,
what
it
is
that
you're
looking
for
here
and
just
a
background.
H
Sure
yeah,
I'm
gonna
talk
real
fast
for
about
five
or
ten
minutes
and
then
see
what
falls
out
of
it.
So
what
I'm
hoping
to
get
out
of
this
conversation
is,
you
know,
is
ytt
a
potential
technology
that
cardinal
health
should
look
at
for
this
current
use
case.
That
is
not
necessarily
the
standard
ytt
use
case.
So
a
little
bit
about
me,
my
role
as
application
security
engineer.
H
This
might
not
be
the
conversation
that
you
expect
to
have
with
that
kind
of
person.
My
background
is,
I
spent
two
and
a
half
years
in
our
innovation
lab
doing
application
development.
One
of
the
things
that
fell
out
of
like
the
devops
revolution
is
that
a
lot
of
organizations
gave
app
teams
latitude
to
go
off
and
do
their
own
thing
and
determine
their
own
patterns
on
how
to
do
build
test
deploy
run.
We
did
that
as
an
organization
here,
and
we
found
out
that
we
learned
a
lot
of
things.
H
We
also
found
out
that
you
know
as
an
application
security
engineer
now.
Looking
at
it,
we've
got
150
app
teams
that
are
doing
everything
a
little
bit
different
and
that
sucks
you
know
I
I
want
to
be
able
to
have
a
single
extension
point
to
be
able
to
go,
provide
them
with.
You
know:
here's
this
new
tooling,
that
does
this
type
of
scanning
or
whatever.
H
So
one
of
the
larger
initiatives
that
cardinal
is
doing
right
now
is
trying
to
standardize
build
test
run
across
the
organization.
You
know
cardinal,
we
we
have
a
strong
relationship
with
pivotal
and
vmware.
We
adopted
pcf
very
early
or
tanzu.
I
guess
now:
we've
used
concourse
for
build
for
a
long
time,
so
I
don't
know
how
familiar
everyone
is
with
concourse,
but
that
is
currently
the
path
forward
for
our
build
tools
or
our
thing.
Doer,
it
uses
yaml
as
configuration
and
actually
do
you
mind
if
I
steal
a
share.
H
So
this
issue
was
opened
in
the
concourse
github
a
while
ago.
That
was
you
know:
hey
we've
got
a
whole
bunch
of
pipeline
yamls
that
are
running
around
they're
hundreds
and
hundreds
of
lines.
Long
and
everybody
is
doing
things
a
little
bit
differently
right.
So
I've
got
an
app
team.
I've
got
60
different
services.
You
know
I
I
don't
have
time
to
go,
reconcile
all
of
the
different
configurations
for
their
build.
I've
got
multiple
app
teams
that
I'm
now
managing
et
cetera,
et
cetera.
H
So
there
was
a
proposal
that
was
thrown
out
here.
That
was
like
how
do
we
break
this
up
and
templatize
things
so
that
you
know
your
deploy
jobs
all
use
this
template
your
you
know,
smoke
test
jobs,
all
use!
This
template
there's
been
some
back
and
forth
on
this.
I
don't
know
veto
threw
it
in
their
like
v10
roadmap.
A
year
ago
I
also
know
they've
been
very
busy
trying
to
get
to
kubernetes
native.
So,
like
I
get
it,
they
haven't
had
time
to
work
on
this
templatizing
function.
H
The
basis
of
how
a
pipeline
actually
looks
is
kind
of
here.
So
you
have
these
things
that
are
tasks.
You
have
these
things
that
are
jobs.
You
have
what's
considered
a
resource
which
is
it
could
point
at
git.
It
could
point
at
you
know
a
slack
web
hook.
It
could
point
at
a
pcf
environment
for
a
deployment.
You
define
tasks
which
are
the
set
of
steps
that
you
walk
through
to
do
a
thing
right.
H
So,
as
you
can
imagine,
a
lot
of
these
things
are
going
to
be
common
across
a
lot
of
different
projects.
I've
got
a
spring
boot
micro
service,
I'm
pushing
it
into
pcf.
The
deploy
is
going
to
be
relatively
simple
between
all
of
them,
so
we're
trying
to
figure
out
a
way
to
enable
an
internal
repository
to
hold
all
of
the
potential
configuration
and
and
drive
that
across
the
organization.
H
So
ytt
has
come
up
a
couple
of
times
as
a
templatizing
tool
and
before
I
go
and
do
a
bunch
of
spike
work
and
try
and
proof
of
concept
this
I
wanted
to
get
the
com
the
thoughts
of
the
community
right.
One
thing
that
is
a
major
concern
right.
We
have
probably
150
different
app
teams.
That's
just
a
guess:
I'm
an
application
security
engineer.
I
should
probably
know
that
and
I
don't
which
is
scary.
H
So
everybody
is
gonna,
have
a
couple
of
different
requirements
and
as
we're
building
this
thing
out,
we're
going
to
need
to
do
it
as
an
inner
source
type
of
product.
So
we're
going
to
need
to
have
tech
leads
from
different
app
teams
coming
and
pulling
request
pull
requesting
into
this
centralized
project,
and
you
know,
coming
from
an
application
engineer,
background
being
able
to
run
a
unit
test
on
a
pull
request
that
then
pushes
a
status
to
your
branch
and
you've
got
branch
protection
that
says
no.
H
You
can't
merge
is
going
to
be
very
important
to
us,
so
I
actually
ran
into
an
issue
in
helm
in
a
helm
chart
just
yesterday
that
kind
of
describes
what
I
potentially
foresee
happening.
If
we
use
ytt
to
do
this-
and
you
know
I'll
preface
this
all
by
saying-
I
haven't
looked
into
ydt
more
than
like
a
half
hour
of
googling
and
looking
through
your
github,
and
I
crashed
the
meeting
on
tuesday
and
nancy
reached
out
to
me.
H
So
that's
my
background
and
what
I
understand
of
the
product,
so
I
was
deploying
the
key
cloak
helm
chart
yesterday.
I
threw
a
really
long
name
on
it
right.
So
this
is
a
value.
It's
it's
going
to
be
in
your
values
or
you
know,
and
if
we
templatize
concourse
the
same
way,
I
foresee
us
using
a
values,
file
of
some
kind
that
then
gets
populated
into
all
of
your
templates
and
ytt
jams.
It
all
together
for
me
into
a
pipeline
dot
em.
H
So
the
way
their
chart
was
set
up
is
in
their
service.
They
point
they
use
this
helper
function,
key
cloak,
full
name
which
does
it
trunk
right,
because
apparently
I
don't
know
something
something
and
I'm
new
to
kubernetes.
I've
been
using
it
like
a
month
and
a
half,
so
I
don't
understand
what
that's
actually
doing
in
their
yeah
in
their
ingress
when
they
point
at
their
back
end.
H
They
point
their
service
name
at
something
else.
Guess
what
happened
when
I
used
a
really
long
name
that
got
truncated,
well,
cool.
My
service
got
truncated.
My
ingress
pointed
at
this,
and
now
it's
all
broken,
and
so
I
want
to
be
able
to
you
know
if
we
take
this
pattern
and
move
it
over
to
a
potential
concourse,
ytt
templating
type
of
thing,
I
want
to
be
able
to
run
a
pull
request.
That
says,
when
I
give
you
this
value,
assert
that
this
yaml
output
test
case
is
what
the
templatization
actually
outputs.
H
I'll
jump
in
if
nobody
else
so
for
what
it's
worth
one
one
thing
I'll
throw
out
in
initially
like
the
answer
of
ytt
is
not
the
tool
for
this
is
completely
acceptable.
If,
if
the
answer
is
go
write
your
own
templatizer
use
go
yaml
and
move
on,
I'm
I'm
a
hundred
percent
cool
with
that.
I
just
thought
you
know
I.
I
would
love
to
reuse
community
tools
when
and
where
I
can
and
contribute
when
and
where
I
can.
E
For
that
you
know
the
yaml
that
comes
out
of
somewhere
right.
So
I'm
not
sure
if
you're
kind
of
a
trying
to
also
once
or
if
there's
some
kind
of
an
interconnection
between
so
maybe
I'll
start
with
the
most
concrete
one
that
you
mentioned,
is
I
searching
on
the
value
on
some
kind
of
yaml
that
popped
out
of
somewhere.
E
So
let's
assume
for
a
second
that
we
have
no
idea
where
that
yaml
is
popping
out
from
we
just
have
yaml.
We
got
a
certain
some
values
right,
so
ytt
can
help
you
with
that.
There
is
a
feature
called
overlays
and
typically
overlays
are
meant
to
modify
things
in
the
yaml,
but
there
is
a
functionality
in
overlays
to
say.
Let
me
assert
on
a
particular
value,
so
you
could
write
some
overlays.
E
That
say,
I'm
just
gonna
make
sure
that
some
particular
value
is
there
right
and
you
know
feel
free
to
kind
of
ping
us
on
slack.
If
you
want
to
have
some
more
examples
of
that
or
or
things
like
that
right,
so
that's
kind
of
one
way
to
well
that
that's
a
that's
one
way
to
do
that.
One
particular
validation
thing,
maybe
then,
to
step
back,
so
you
were
working
with
some
helm.
Template
was
one
part
of
your
question
a
question
around.
E
Would
you
should
I
use
ytt
to
template
the
things
instead
of
helm,
template?
Okay,
so
that
was
not
it?
Okay.
So
sorry,
so
the
only
part,
the
only
part
of
that
question
was:
how
do
I
start
now?
What
was
the
bit
about
the
concourse?
How
does
that
play
into
this
whole
thing,
you're,
just
saying
that,
because
I'm
running
concourse,
that's
how
it's
in
the
picture-
and
I
want
to
be
able
to
report
back
the
status
to
github
or
something
like
that
right
so
so
concourse
is,
is
the
problem
statement.
H
They
they
define
job
configuration
in
yaml,
and
if
I
had
one
pulled
up,
I
could
show
you,
but
for
a
general
micro
service
pattern
where
you're
doing
multiple
stage
deployments
and
you're
doing
smoke,
tests
and
unit
tests
and
all
of
this
stuff
that
yaml
file
gets
to
be
eight
or
a
hundred
thousand
lines.
Long
yeah,
where
a
lot
of
that
stuff
is
shared
components
across
multiple
projects
where
it
should
be
so
that
that's
our
goal
is
to
instead
of
having
each
project
own,
a
thousand
line.
H
E
H
E
So
maybe
a
follow-up
question
to
that
would
be
so
assuming
that
you're
the
one
maintaining
this.
Let's
say
you
know
I
don't
know.
Maybe
you
have
a
set
of
templates
that
generates
you
a
basic
pipeline
with
the
unit
test
smoke
test
whatever
it
is
right,
I'm
assuming
that
you
know
you're
gonna
go
and
try
to
convince
bunch
of
teams
to
say
you
know
what
just
roll
with
this
configuration.
You're
gonna
love
it.
You
don't
have
to
maintain
all
of
the
stuff
yourself
right.
So
really
the
author.
For
that
configuration.
E
Is
you
right
and
so
then
the
question
I
guess
is
well.
If
you
are
the
author
right
and
you
know,
let's
say
using
ytz
for
that
stuff
right,
I
would
imagine
that
you're
probably
going
to
be
a
lot
more
conscious
about,
like
not
adding
code
like
you
know,
trim
the
random
name
over
here
or
over
there
right.
So
I'm
not
sure
if
you
actually
would
benefit
from
kind
of
a
testing
the
the
end
result
in
kind
of
the.
E
Let
me
run
ytt
with
this
particular
set
of
data
values
and
just
kind
of
assert
on
the
value
for
your
own
use,
not
for
the
use
against
what
kind
of
the
people
populate
in
so
it's
kind
of,
like
you
know,
just
like
you
would
ship
any
software
project,
you
might
have
some
tests
for
it
right,
but
they
kind
of
stay
internal
to
you,
they're
not
really
kind
of
shipped
off
into
other
places.
E
H
So
so
in
in
the
end
state,
yes,
one
thing
I
want
to
touch
on
is
actually
rolling.
This
out
is
not
going
to
be
convincing
app
teams
so
much
as
it
is
like
a
cto
edict,
so
that's
going
on
and
and
to
hit
the
kind
of
timelines
that
we're
hoping
to
hit
our
team
or
even
a
combination
of
teams
are
not
going
to
be
able
to
accomplish
it
alone.
H
E
E
Yeah
I
mean
depending
so
this.
Actually,
maybe
this
is
something
that
you
can
circle
back
with
us
on
flag,
given
I'm
guessing
we're
getting
closer
to
the
end
of
the
meeting,
but
this
sounds
like
a
question
of
how
is
it
how
what
are
the
different
approaches
to
testing
ytt
templates
just
in
general
right
and
so
there's?
Actually
a
few
answers.
Ytt
overlays
is
just
one
of
those
answers
may
not
actually
be
the
maybe
appropriate
in
your
case,
something
like
that.
So
I
I
I
think,
yeah.
E
I
think
it
will
probably
be
better
to
spin
up
a
thread
and
just
kind
of
throw
in
a
few
possibilities
there
and
see
kind
of
where
we'll
go
from
there.
H
A
Okay
awesome
thank
you,
tyler
for
for
sharing
that
and
we'll
take
care
of
it
on
slack.
Hopefully,
we
have
six
minutes
left
to
go
over
these
triage
items.
I'm
gonna
share
my
screen
again.
C
Anyone
doesn't
seem
like
I
remember
we
talked
about
this
previously,
so
I
can
give
like
a
high
level
overview
of
what
this
issue
is
describing
so
currently
in
cap.
There
is
a
feature
where
you
can
specify
a
directory
structure
as
input,
and
it
will
using
that
directory
structure
deploy
applications
to
the
cluster,
using
like
the
name
of
the
files
in
that
directory
structure,
as
the
name
of
the
app
that's
deployed.
C
C
C
A
I'm
not
sure
who
put
the
triage
help
in
so
this
was
moved
from
monday's
community
meeting
because
we
didn't
have
time
to
go
over
that
section.
So
I
moved
it
here
and
I'm
not
sure
if
it
was
aaron
that
had
put
the
triage
help
in
here.
If
he
could
give
more
contacts
as
to
what
the
goals
were.
I
I
was
part
of
the
pair
that
added
this
okay.
The
objective
of
this
one
was
just
to
come
up
with
some
next
steps
at
the
end
of
the
the
github
issue.
It
said
we'll
bring
this
up
with
our,
I
don't
know
with
our
pm
or
with
the
community
or
something-
and
I
just
wanted
that
to
I
wanted
to
close
the
loop.
C
E
I
guess
we
haven't
really
talked
too
much
about
this
kind
of
like
where
do
we?
E
I
think
this
is
something
that
even
has
come
up
in
the
recent
meeting
about
like
kind
of
a
formalizable
of
an
issue
and
where,
at
some
point
like
becomes
maybe
accepted,
to
be
included
in
the
product
or
something
else.
E
E
Wouldn't
want
to
implement,
but
I
remember
vaguely
from
our
previous
discussion
that
there
was
more
questions
and
discussion
to
be
had
about
it.
So
I'm
not
sure
how
to
might
be
worth
kind
of
dedicating
some
time
to
confirm
where
we
want
to
go.
A
A
C
C
A
We
have
no
time
left
so
I'll
move
the
last
two
for
sure
to
monday's
meeting,
and
we
can
touch
on
those,
then
all
right.
So
that
concludes
the
first
edition
of
our
office
hours.
Once
again,
we
have
these
now,
every
second
and
fourth
thursday
of
the
month
at
11
30
a.m.
Pacific
time,
2,
30,
p.m.
Eastern
time
we
also
hold
our
cargo
community
meetings
every
monday.