►
Description
A
So
many
times,
okay,
thanks
I'm,
sorry
about
that,
so
trying
to
fetch
the
tool
to
show
you
in
case
you,
you've
seen
the
store
before
so
I
was
using
this
tool,
which
I
don't
have
it's
not
that
I'm
really
trust
in
it,
but
it
did
help
me
go
move
forward
with
the
errors
you
would.
I
was
just
looking
at
this
and
looking
at
other
examples
of.
A
Deployments
and
things
and
I
was
like
I
was
able
to
move,
keep
moving
forward
and
I.
Think
one
so
adding
this
init
container
here
was
part
of
that
process
of
going
through
the
errors
to
get
to
a
point
where
I
could
deploy
this
with
customize.
So
what
no
one
is
saying
is
not
something
that
I
know
for
sure,
if
you
know
for
sure,
but
I
would
investigate
because
it
might
be.
The
customize
has
restrictions
that
keep
control
Desiro.
Is
that
a
thing.
C
A
All
right,
I,
just
I,
didn't
want
to
be
shooting
in
the
dark
too
much.
That's
why
I
wanted
to
have
this
quick
cause?
It's
just
to
get
some
some
information.
If
anybody
had
any
but
I'm
pretty
sure
I
had
to
add
this
back
here
to
move
forward
and
maybe
once
I
get
through
and
clean
everything
up.
It
might
be
that
I'm
being
able
to
go
back
and
remove
things,
and
that
was
not
actually
the
issue.
I
don't
know,
but
that
might
be
an
issue.
B
C
I
I
don't
know
that
it's
stepping
back
I,
don't
think
it's
that
it
requires
the
pod
to
have
in
it.
Containers
I
think
having
worked
on
some
base
tools
for
ansible
I.
Think
that
it's,
if
you
specify
a
path
in
in
some
overlay
that
the
path
may
already
need
to
exist
like
it's.
It's
not
necessarily
intimate
containers
by
themselves.
It's
something
to
do
with
the
mo
pathing,
and
maybe
they
don't
want
to
add
something
to
it,
but
I'm,
not
for
sure
that
that's
the
Hays
yeah.
A
A
A
A
A
A
Sorry
this
actually,
this
is
how
I
started
out
the
deployments,
but
customized
was
complaining
about
the
format.
So
I
was
looking
at
other
ways
to
format
the
mo
file,
and
this
worked
so
I
removed
these
top
two
items
here.
These
top
two
elements
here
and
just
put
three
dashes
the
API
version
until
it
did
it
this
way
and
then
I
separated
everything
with
three
dashes
and
this
this
format
Mart
worked.
However,
I
had
to
change
a
few
things.
One
significant
one
was
excuse
me
I
had
to
follow
this
format,
which
is
not
the
way.
A
It's
not
the
way
that
that
we
have
the
outputs
from
the
install
the
yellow
outfits.
So
basically,
there
is
schema
Eadie's,
and
this
is
under
versions.
This
is
also
something
I
needed
to
add,
so
I
might
be
doing
this
wrong.
I
mean
it's
not
as
wrong
in
the
sense
that
this
is
valid,
but
this
might
not
be
what
we
want.
If
we
do
want
to
go
this
way,
then
my
question
is
I'm
not
really
clear.
A
So
it
was
the
same,
the
so
a
good
question.
The
this
was
in
the
process
of
trying
to
make
the
customized
deploy
this
to
my
cluster
and
I,
wouldn't
get
any
specific
failure.
I
will
just
get
go,
get
the
CMO
would
be
highlighted.
I
was
using
a
different,
a
different
editor,
but
you
highlight
some
like
an
element
with
a
red.
You
highlight
red
and
say:
okay.
A
A
A
A
C
The
the
reason
I'm
asking
this
is
so
the
one
on
the
left
looks
correct
to
me
in
looking
at
the
all
the
schemas
that
I've
been
looking
at
for,
what's
necessary
for
a
for
a
CRD
to
define
a
schema
at
least
in
v1
is
it
would
have
the
version,
it
would
have
spec
dot
versions
and
then
each
one
of
those
versions
would
have
a
name
here.
It's
b1
and
then
underneath
that
each
version
of
the
CRD,
so
we
have
v1
of
backups
and
v1,
would
have
its
own
schema,
which
has
an
open,
API,
v3
schema.
C
However,
that's
based
on
v1
I,
don't
know
for
sure.
If
V
1
beta
1
had
a
different
layout
of
the
fields
because
I
know
there
is
a
structure
in
the
file
called
backup,
custom
resource
validation,
mm-hmm,
there's
a
structure,
that's
neat,
I
think
this
game
feel
the
struck.
The
field
is
named
schema,
but
the
structure
is
called
valid,
backup
resource
validation,
which.
C
A
C
A
A
Well,
I
give
like,
if
you
know,
or
if
you
can
recall,
like
six
minute
thinking
and
recall,
and
give
me
some
pointers,
is
fine
I
just
needed
a
break
before
it
cut
is
about
this
again,
but
in
what
I
really
wouldn't
need
to
know
is
when
like
what?
What?
What
is
the
output
that
we
want
from
our
commands
once
we
have
the
C,
so
let's
say
it's
version
1.4
that
that
this
gets
into
version
1.4
of
the
layer.
A
B
So
the
the
version
like
in
line
76
and
the
versions
at
lines
92
and
93
or
two
separate
things.
So
the
first
one
is
like
you
know,
for
the
CR
D
itself.
Are
you
creating
it
as
a
v1,
beta,
1c
Rd?
The
stuff
is
very
hard
to
explain
clearly,
but
for
like
for
now
we
need
to
create
them
as
V
1,
beta
1
or
at
least
have
the
option
to
create
them
as
V
1
beta
1,
because
any
cluster
prior
to
1.16
doesn't
have
the
v1
API
extensions,
API.
C
Okay,
oh
that
seems
to
violate
the
stuff
that
they've
mentioned
in
their
Docs,
because
it
shouldn't
have
gone
straight
to
preferred
version
without
incubating
anyway.
V1
beta
1
is
where,
where,
where
line
line,
76
should
be
V
1
beta
1,
because
that
gives
us
the
broadest
segment
of
kubernetes
versions
and
then
on
line.
C
Oh,
that's
only
for
custom
resource
definitions
because
custom
custom,
resource
definitions
are
our
V
1
beta,
1,
okay,
the
rest
are
versioned
on
their
own
and
then
similarly
we're
line
93
is
our
version
of
the
backup
CRD.
So
we're
saying
the
backups
here
or
yeah.
This
is
the
backup
security.
So
we're
saying
the
backup
CRT
is
version
one
and
that
that
line
we
control.
So
we
would
say
like
this
is
this
is
a
line
that
we
would
bump
and
say:
hey.
We
have
a
beta
version,
two
backup
or
we're
removing
the
v2
backup.
Okay,.
A
C
A
A
Because,
depending
on
the
version,
the
the
the
keys
can
have
a
different
spec
right,
the
the
what's
gonna
be
allowed
here.
It's
going
to
depend
on
this
version
that,
where
am
I
wrong
by
the
way
I
don't
have
a
cut-off
anymore.
Christine
canceled,
I
wanna
move
down
one
for
Friday.
So
if
you
have
time,
I
have
time
yeah.
C
I,
don't
have
anything
else,
yes,
so
there's
two
two
things
that
come
to
mind
that
rely
on
that
version.
The
fields
that
are
in
there
and
usually
in
kubernetes
excuse
me
I.
Think
the
CR
bees
have
been
the
exception
so
far
that
I've
seen
me
being
usually
from
beta
to
the
next
version.
It's
not
there's
not
significant
changes
so
usually,
but
not
always
from
beta
to
the
actual
version.
C
The
fields
will
stay
pretty
much
the
same
and
then
the
other
thing
that
will
change
is
the
import,
because
the
import
path
is
linked
to
the
version
in
the
code.
So
those
are
the
two
main
things
that
come
to
mind.
I
think
we
had
it's,
maybe
in
the
helm
chart
we
had
users
change
this
particular
one
from
Apps,
V,
1,
beta
1
to
AP's
V
1,
so,
like
our
helm,
charts
aren't
necessarily
in
sync
with
the
Yamal
that
gets
pushed,
which
we
we
know.
A
But
any
if
you
answer
this
I'm
sorry
do
we
know
do
we
have
an
opinion
on
what
this
would
be,
but
do
we
know
for
the
purpose
of
this
news
year
like
organization
for
for
deployment
is
specific
because
deployments
I
think
I.
Think
the
point
is
it's
big
enough
and
important
enough
that
we
need
we
should
nail
what
the
version
and
make
sure
that
all
the
fields
are
performance.
B
C
A
A
A
B
There's
probably
not
a
super
straightforward
way
to
do
this,
I
mean
because
we're
basically
constructing
kubernetes
deployment,
API
objects
and
they're
now
putting
them
as
yamo
I
feel
like
we'd
probably
have
to
do.
You
know,
use
some
kind
of
e
mo
processing
library
to
do
like
a
yeah
mold.
If
I
don't
know
what
do
you
think?
No
one.
C
A
B
A
Exactly
so
now
that
actually
I'm
thinking
about,
if,
if
we
get
a
but
like
all
a
bunch
of
fields,
they're
not
relevant
for
this
and
let's
say
they're
empty
or
now
I
anyway,
say
I'll,
just
patch
the
deployment
with
this
new
llamo,
then
you
will
obviously
overwrite
so
the
the
one
way
I'm
thinking.
That
is
that
we
would
have
to
fetch
the
deployments
builds
their
outputs
with
the
with
the
current
values
that
are
there
replace
the
ones
we
want
to
replace
and
announce.
Put
the
whole
thing.
C
Yeah,
the
the
fetch
runs
into
the
case
where
you
don't
necessarily
like
it
still
stumbles
on
the
case
where
you
don't
necessarily
have
a
kubernetes
server
up
and
running
yet
because
I
mean
we
can
similar
to
how
that
install
command
works.
Today
we
can,
we
could
spin
up
an
empty
deployment,
struct
fill
in
these
fields
and
then
give
it
to
a
yeah
mol.
C
C
C
So
you
imported
a
pre-populated
image
from
somewhere
right
now.
I'm,
sorry,
I'm,
sorry,
a
pre-populated
deployment
struct
from
a
Valero
library.
So
you
didn't
have
to
fill
in
all
of
that
stuff
each
time.
So
we
already
had
stuff
like
namespace
and
stuff
like
that
I'm
just
thinking
of
a
shortcut
helper
yeah.
A
C
B
Mean
today,
I'm
just
stepping
back
and
thinking
about
the
general
flow
here.
So
the
idea
is
that,
if
you're,
a
user
who
basically
just
wants
to
pre,
generate
some
yamo
and
check
it
in
to
get
you'll
essentially
have
a
dry
run
flag
for
all
of
these
new
COI
commands.
So
you
can
dry
run
bolero
in
it
to
get
the
base
install.
You
can
dry,
run
Valero
plug
an
ad
to
get
the
patch
to
add
the
plugins
that
you
need.
A
1.4,
let's
say,
and
here
is
we
I
mean
we
can.
This
is
not,
for
example,
this
file
here
is
not
just
the
ployment,
it
has
services,
it
has
the
our
back
stuff.
So
we
can
rename
this
something
else.
But
basically
this
is
everything
for
the
basic
deployment.
Then,
if
they
run
like
I,
said,
set
the
backup,
okay,
Sarge
location
and
that's
all
they
need,
then
you
go
into
customize
fire.
It
just
leaves
those
two
files
super
simple,
and
then
you
have
the
overlays,
which
is
going
to
be
what
plugins
you're
going
to
need.
A
B
A
Okay,
so
so,
let's
say,
if
I'm
not
so
I,
don't
know
how
people
really
use
this
but
I'm
thinking
that
this
is
how
you
work.
So
the
answer
is
no.
They
should
not
be
checked
into
our
repo
unless
we
want
to
have
like
an
example
yeah.
We
could
have
an
example,
but
some
I
mean
the
credentials,
for
example,
would
be
bogus.
They
have
to
replace
them
anyway.
A
So
the
idea
is,
if
I'm,
a
user
I
do
I
run
out
of
those
four
commands
like
in
it
Guerra
llamo,
which
which
would
be
this
this
one
here
deployments
or
if
we,
my
name,
is
something
else
or
whatever
like
they
just
gonna
pipe
it
to
a
file
right,
and
we
can
document
this,
so
the
parties
to
a
file
they'll
pipe
the
backup
location
that
they
just
create
with
a
dry
run
to
a
file.
They
dump
everything
somewhere.
A
We
can
give
documentas
and
say:
hey
Dom,
it's
worth
for
the
names
based
base
and
then
they're
gonna,
add
a
plug-in
they're
gonna,
pipe
that
into
a
mo
file,
they're
gonna,
leave
it
or
whatever
and
then
gonna
have
another
custom
is
customization
yamo
die
structure
like
this.
That's
going
to
determine
okay,
my
base
is
here
and
then
this.
This
is
the
plug-in
that
I
want
to
add
patched
on
that
base
and
that's
it.
B
Okay,
yeah,
that
that
makes
sense
to
me
it
seems
pretty
reasonable
I
guess
the
one
thing
is
I
would
want
to
go.
You
know
again
try
and
look
at
a
couple.
Other
projects
and
see
like
is
this
similar
to
structures
to
flows
that
other
projects
are
using,
where
you
have
kind
of
a
CLI
to
generate
various
parts
of
Yemen.
That
can
be
all
stitched
together
through
customize
yeah.
A
I
know
sort
sort
manager
has
is
using
customized
to
and
they
have
actually
looked
it
up.
They
have
both
a
repo
I
wear
this
to
the
pier
and
they
have
a
blog
post,
which
I
haven't
finished,
really
explaining
how
their
work
for
is,
but
looking
through
the
repo
is
yeah.
This
is
the
idea
that
they
are
using.
A
A
C
A
If
the
patching
works,
the
patching
of
this,
the
plugin
works
with
customized,
it
looks
very
simple:
just
give
them
the
ammo,
they
were
gonna
I
mean
if
you
don't
customize,
you
know
where
you
need
to
put
things.
You
know
how
how
to
use
the
customization
customization
memo,
it
seems
so
simple,
I
mean
the
hard
thing
is
coming
up
with
a
properly
formatted
llamo
file
or
output.
C
What
but
in
my
mind,
it
makes
sense
to
use
the
features
that
are
built
in
acute
control,
because
that's
what
most
people
will
have
on
hand
if
they
want
to
reach
for
helm,
we're
already.
We've
got
that,
but
I
don't
see
any
I
don't
see
a
benefit
to
unless
it
without
without
knowing
the
features
for
sure
I,
don't
see
a
clear
benefit
to
like
going
all
in
and
saying
yeah
we're
gonna
support
the
customize.
That's
not
gonna
keep
control.
A
C
C
C
A
C
C
A
B
A
Okay,
I
was
laughing
when
I
saw
this,
so
basically
I
wrote
it
down
here
but
I'll
just
say:
I
took
the
image
images
swag
from
flower,
install,
don't
that
in
here
and
didn't
think
about
it
again,
but
so
why
you
said
Steve
is
afraid
for
you
both
said
so,
but
but
then
you
had
a
questions
do
do
we
want
to
support
the
ability
to
add
multiple,
multiple
plugins
like
we?
We
supported
that
with
install,
but
now,
with
the
actual
plug-in,
add
commands
so
now
I
have
this
question
I
think.
B
Did
you
know
it'd
be
nice
to
support
it
and
you
can
also
just
do
it
as
multiple
positional
arguments,
so
you
can
just
save
a
little
plugin
add
you
know:
falero
plugin
for
AWS
flair,
plugin
for
g
CP,
which
is
consistent
with
our
existing
commands.
Like
you
can
say,
arrow
back
up,
get
back
up
one
space
back
up
to
and
it'll
you
know
get
both
of
those
backups
for
you.
So
it's
just
it's
just
basically,
two
positional
arguments.
A
A
So
that's
a
quick
I
agree.
Okay,
so
this
is
a
just
want
to
get
a
clarity
on
this
cuz
just
to
make
sure
so
what
I
was
thinking
was
to
have
Valero
in
the
verb.
So
with
that
intense
I
was
thinking
to
rename
the
server
to
win
it,
but
the
functionality
will
be
the
same.
So
this
was
a
deprecation
in
name
and
now,
functionality
and
I,
don't
know
I
wasn't
sure
if
it
was
clear
or
not.
If
you
were
is
if
this
is
now
right,.
A
B
Velaro
in
it
creates
deployment
like
creates.
The
Valero
deployment
like
generates
yeah
mol
and
applies
it
to
an
API
server
and
the
Valero
server
command
is,
is
the
command
that's
run
inside
the
Valero
bot?
It's
just
the
process
that
Valero
runs
to
run
on
all
the
controllers,
and
everything
so
like
one
is
something
that
a
user
runs
to
get
Valero
set
up
and
the
other
one
is.
What
is
the
process
that's
running
inside
the
container
on
the
server
yeah?
B
You
know
it
probably
would
have
been
better
named,
like
Valero
serve
rather
than
server
I
guess
if
we
wanted
to
have
a
verb
there,
but
this
is
not
something
a
user
ever
runs
unless
they're,
developing
and
running
it
locally.
This
is
just
what
the
process
that's
wrong.
In
the
kubernetes
cluster
in
the
pod.
C
They
may,
if
they
edit
the
deployment,
so
the
the
most
common
one
is
log
level.
So
if
you
go
in
right
now,
the
way
you
edit
it
is
you
go
in
edit
the
deployment
and
add
two
extra
yellow
lines
to
say
space
one
level,
then
a
new
line,
space,
debug
or
whatever
log
level
you're,
changing
it
to
or
you're
changing
the
log
format.
So
these
are
a
lot
of
them.
I
mean
users,
don't
use
them
directly
and
and.
C
C
So
when
the
process
starts
up,
it
grabs
that
and
just
hold
so
in
the
process,
we've
talked
about
putting
any
annotation
on
the
locations
and
things
like
that.
But
at
the
moment
right
now,
all
it
does
is
either.
It
just
looks
for
one
called
the
false
looks
for
a
single
one
or
grabs
this.
This
argument
should.
A
B
Also
one
one
thing
to
think
about
here:
right
is
when
you're,
when
you
run
Valero
in
it,
your
output
is
basically
a
kubernetes
deployment
and
it
would
probably
be
nice
to
say
to
you
know
when
you're
running
Valero
in
it
to
have
an
option
to
say
you
know:
I
want
a
client
QPS
of
50
and
then
basically
that
flag
just
gets
passed
through
and
and
configured
as
a
flag
on
the
Valero
deployment.
So.
A
You
yeah,
so
the
deployment
has
the
arguments
and
Zach
Valero
server.
That
starts
at
the
server.
So
basically
we
could
edit
the
deployments,
but
then,
when
we
have
to
to
read
the
poets,
you
write
because
you
wouldn't
be
just
there.
We
have
to
do
two
things
edit
deployments.
If
we
could
add
these
things
to
the
I
mean
some
of
these
makes
sense
to
me.
But
then
how
can
we
trigger
the
redeploy.
B
Yeah
I,
don't
know,
that's
I,
don't
know
that
we
need
to
expose
them
all
through
bolero
in
it
I
mean
many
of
them
are
kind
of
our.
You
know,
advanced
things
or
things
that,
like
99%
of
folks,
are
never
going
to
tweak
like
the
plug-in
door
or
the
metrics
address
you
know,
and
so
adding
them
to
Valero
and
knit
mate
means
that
there
are
just
more
flags
on
that
command.
Right.
A
C
C
It
it
so
that
you
say:
Valero
can
figure
this
thing,
Valera,
configures
set
or
whatever
client
burst
mm-hmm
and
you
abstracted
it
and
we
then
like
for
now.
We
set
it
as
a
thing
on
the
we
set
it
as
a
flag
on
the
deployment,
and
then
we
abstract
it
so
that
maybe
later
we
move
it
and
store
it
somewhere
else
right
where
we're
just
instead
of
making
them
at
at
the
deployment,
we
could
change
how
we
store
it
all.
A
B
C
B
Could
also,
you
know,
consider
adding
a
level
of
indirection
here
so
instead
of
having
each
of
these
as
as
like
a
proper
flag
on
Valero
in
it,
you
could
say
Valero
in
it
and
then
have
a
single
flag,
called
server
flags
and
then
have
like
a
string
of
key
value
pairs.
So
you
could
say:
client,
QPS,
equals
20.
Comma
restore
only
equals
false,
but
it
would
just
be
a
string
value.
So
then,
then,
you
don't
have
to
add
a
new
flag,
double
arrow
in
it.
Every
time
you
add
a
server
flag,
but.
B
A
A
A
C
A
Okay,
this
is
the
question
which
is
also
same
sort
of
question
a
little
bit
further
down
in
how
to
store
the
information
for
defaults.
So
basically,
what
I'm
thinking
is
would
be
nice
to
start
in
the
back
of
the
location
because
it
is
an
attribute,
but
if
that
means
we
would
have
to
fetch
all
of
the
locations
in
iterate
through
them
to
find
out
which
one
is
the
default,
then
that's
not
that's,
gonna
be
the
way
to
do
it
so
I'm
thinking.
Alternatively,
we
will
have
an
annotation
that
we
can
query
like
an
annotation
on.
A
C
A
C
A
C
C
So
it's
a
convention
that
exists
and
I,
don't
I,
think
field.
Selectors
are
kind
of
new,
so
I,
don't
know
how
many
versions
back
they
were
introduced
but
again,
like
I,
think
they're
restricted
to
a
certain
number
of
fields
like
it's
only
like
the
stuff.
That's
in
object,
metadata
or
something
like
that.
So
I
don't
think
we
can
put.
We
can't
create
a
new
arbitrary
field
and
put
put
a
label
selector
on
or
not
feel
it's
Lecter
on
it,
but
we
can
create
an
arbitrary
label
and
put
a
label
slaughter
on
it.