►
From YouTube: kubernetes kops office hours 20190426
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
Hello:
everyone:
it
is
Friday
April,
26th
2019.
This
is
cops
office
hours.
I
am
your
moderator,
just
in
Santa,
Barbara
I
work
at
Google
a
reminder.
This
meeting
is
being
recorded
and
will
be
put
on
the
Internet.
Please
be
mindful
of
our
code
of
conduct.
If
you
would
like
to,
we
haven't
a
we
have
an
agenda
if
you
would
like
to
add
items
to
the
agenda
or
add
your
name
to
the
agenda
for
the
sake
of
posterity.
That
would
be
very
welcome
and
helpful.
A
B
Basically,
we've
cut
a
few
and
all
four
releases
for
new
companies
versions
and
I'm
wondering
if
we're
gonna
do
the
same
for
game
eyes,
so
that
each
kubernetes
version
has
its
own
ami
and
if
and
what
our
release
process
for
am
eyes,
looks
like
and
they
change
as
a
result
of
our
many
cops
releases.
Yes,
okay,.
A
So,
yes,
we
should
cut
an
ami.
This
is,
as
I'm
sure
you
know,
we
haven't
even
fully
figured
out
the
credentials
for
doing
releases
of
like
our
binary
artifacts.
The
releases
of
the
images
are
even
more
sensitive
right
now.
The
process
by
which
that
happens
is
basically
fairly
open.
The
the
repo
it's
built
from
the
cube
deploy
repo
guitar
kubernetes
to
deploy.
A
So
if
you
want
to
build
your
own
image,
you
can
do
so.
The
challenge
is:
how
do
we
push
those
image
images
in
a
community
way
without
you
know,
without
it
being
insecure
and
honestly
I,
don't
think
we
will
get
to
that
until
after
we
figure
out
our
binary
artifacts,
which
are
much
simpler
and
I'm
actually
hoping
we're
gonna
make
some
actual
progress
on
that
next
week
at
next
week's
working
group,
kids
in
from
meeting
so
then
I
guess
next
up
would
be
this
topic.
A
It's
not
entirely
dissimilar,
except
that
we
need
a
basically
a
very
lockdown
AWS
account
so
currently
I'm
doing
that
it
is
a
totally
different,
a
diverse
account
from
the
one
I
use
day-to-day.
It's
a
different
like
everything.
Different
billing
account
different
everything,
different
email
address.
Everything
is
different.
The.
A
It
would
be
great
to
get
that
under
community
workings,
but
honestly,
that's
probably
a
ways
away
in
terms
of
the
actual
release.
We
can
certainly
collaborate
on
the
specifications
in
that
cube,
deployed
repo
and
I
see.
There
are
five
pull
requests
which
I
presume
are
against
me.
So
I
should
probably
do
something
about
those
yep.
They
are
mostly
against
me,
so
I
will
work
on
those,
and
that
is
the
right
way
to
collaborate
and
I
am
being
bad
at
not
merging
those
and
or
at
least
reviewing
them.
A
In
terms
of
your
question
around,
do
we
want
to
split
the
way
releases,
work,
I,
don't
think
we
have
to
once
I
know.
The
heartbeat
is
basically
getting
a
new
kernel
version
on
the
versions
where
we're
building
a
kernel
of
our
own,
which
is
we
were
definitely
doing
on
Jesse
the
older
Debian
and
we
are
not
I,
don't
think
we're
doing
that
on
the
new
Debian,
which
is
called
stretch.
A
Someone
will,
let
me
know
if
I
got
the
name
strong,
hunk,
the
sir,
once
once
we
have
the
kernel,
we
basically
push
that
up
to
all
the
releases
and
the
only
difference
in
the
or
the
primary
difference
in
the
images
is
just
about
the
darker
version
that
is
by
default
installed.
So
that's
the
primary
advantage
of
this
images.
It's
it's
honestly.
It's
no
big
deal
to
it's
a
script.
There's
it's
no
big
deal
to
run
ten
of
them
or
six
of
them
doesn't
make
any
difference.
A
A
You
know
the
stock
Debian
image
or
the
stock
CentOS
or
Alison
system,
Linux,
RL
or
whatever
those.
What
images
should
all
work,
so
I
guess
at
some
stage
and
we
used
to
have
to
build,
we
used
to
have
to
build
those
images.
We
might
not
have
to
build
them
in
the
future.
It
is
still
in
my
septum
ization
2,
pre-installed
docker
or
in
future
container
d,
but
maybe
we
can
get
out
of
the
business
of
am
eyes.
I
do
think
that
the
there
is
also
a
sin
in
cluster
lifecycle
or
cluster
API.
A
There's
also
a
separate
effort
to
start
getting
into
that
business
of
maintaining
images.
So
hopefully
we
can
all
work
together
and
maintain
those
images
together,
but
that
is
a
very
long
answer
to
the
idea
that
yes,
I,
should
do
that
and
will
be,
will
be
updating
the
images
so
expect
a
new
set
of
images
with
a
kernel
bump
and
all
the
updates
and
I
will
put
I
will
put
B
I
will
try
to
package
the
correct
doctor
versions
in
there.
A
That's
not,
obviously
that's
not
terribly
difficult,
and
if
anyone
wants
to
help
that
is
super
welcome
yeah
in
general,
it's
a
bit
of
the
bootstrap
VC
project
which
we
were
using
or
which
we
are
using
in
order
to
build
those
images,
I
think
it's
been
deprecated
now
and
so
now
there's
a
different
way
to
build
images.
So
I
don't
know
whether
we
want
to
stay
on
this
business
at
all.
But
hopefully
we
can
have
that
discussion
more
in,
say,
faster
life
cycle.
A
A
A
Yes,
I
will
take
another
look
at
it
as
I
recall.
The
issue
is
that
AWS
region
is
not
necessarily
or
my
concern
was
that
the
a
diverse
region
and
var
wasn't
automatically
gonna
be
the
same
as
the
the
region
in
which
your
bucket
was
and
that's
what
I
just
want
to
make
sure
that
we
were
locking
together.
A
D
A
A
A
What
sort
of
language
there,
but
if
I
actually
try
it
myself,
I
often
discover
what's
going
on
so
maybe
if
I
try
wiring
out
the
s3
region,
I
will
discover
that
this
is
in
fact
safe
or
that
this
is
in
fact
a
nightmare,
and
we
should
just
do
this
or
that
a
diverse
region
is
in
fact
fine
anyway
yeah,
because
if
it's
just
China
yes,
if
it's
just
trying
it
should
be
fine,
but
anyway,
I
will
have
a
look
at
that
I
apologize
for
not
moving
that
forwards.
What,
where
is
that
targeted
to
release
milestone?
A
E
This
was
the
blocks
next
off
of
alpha
2
that
we
decided
with
we
couldn't
replicate
with
that.
Cd
manager
were
on
doing
a
rolling
upgrade
after
successfully
upgrading
to
the
officer
to
the
betas
and
then
doing
another
upgrade
on
the
rolling
update.
The
new
master
won't
join
so
after
hitting
it
again
and
doing
some
debugging
I
realized
what's
happening
is
on
the
running
master.
The
Etsy
host
file
is
being
corrupted,
and
it's
just
it
looks.
E
My
guess
is
between
that
CD
main
and
fcd
events
they're
both
trying
to
do
updates
and
at
some
point
it's
reading
the
middle
of
an
update
and
doing
in
trashing
the
file
it
ends
up
with
an
invalid
host
file,
which
then
prevents
it
from
ever.
Getting
the
new
IP
address
for
the
new
master.
That's
coming
up,
so
I
put
a
fix
a
patch
in
for
a
CD
manager
that
will
prevent
it
from
writing
out
a
bad
file.
E
A
Especially,
thank
you
for
joining
that
down
as
well.
Like
that's
really
great
to
know.
I
I
had
I
think
I
had
hoped
that
we
could
not
find
mount
the
same
hosts
file,
but
I
will
have
a
look
at
what's
going
on
there
and
I
will
look
at
your
PR
as
well.
I
think
this
yeah
yeah
and
I
emailed
you
a
copy
of.
A
Maybe
we
can
figure
out
a
way
to
only
to
make
the
Etsy
host
file
scope
to
the
pod,
in
which
case
yes,
we
should
avoid
that
class
of
bug
there
may
be
other
bugs,
but
we
should
hopefully
avoid
that
classified.
It's
also
supposed
to
the
I'll.
Also
look
into
the
semantics
of
how
files
like
exclusive
opening
of
files
when
they
are
bind
mounted
I
honestly
have
absolutely
no
idea
how
that
works
or
does
not
work.
A
F
F
The
scripts
that's
running,
tries
to
use
a
make
Travis
CI,
but
when
I
check
to
make
the
Travis
Toyama
far
in
the
release,
113
repository
a
branch
I
see
the
script
says:
Mik
SIA
not
make
Travis,
CI
and
I
do
not
know
how
this
happened
and
if
I
just
want
to
see
it
was
a
way
if
someone
has
seen
this
before
and
how
to
resolve
it.
Yes,.
A
I
broke
this
recently,
we
split
Travis
off
to
avoid
doing
any
phase
or
something
some
basil
task,
camera
which
one
it
was
Mike
might
know.
I
don't
remember,
but
yes
that
was
Travis.
The
I
think
there
are
two
things
here.
First
of
all
in
general,
you
can
we
in
general,
it's
better
to
Targetmaster
and
we
can
cherry-pick
back
so
you
you
can
sort
of
sidestep
this
issue
and
well
actually,
I
have
on
the
agenda
later
to
show
the
the
cherry-pick
spreadsheet.
With
all
these
released
brunches
we
I've
had
to
create
a
spreadsheet,
but.
A
You
can
you
can
sort
of
if
it's
more
normal
to
target
master
or
even
if
you
want
to
get
it
into
one
thirteen
and
then
we
can
cherry-pick
it
back
once
it's
merged.
Okay,
that's
the
sort
of
short
answer,
I
think
and
then
the
long
answer
is
yes.
We
know
that
some
of
the
branches
have
like
we
should
have
a
separate
CI
and
okay
thanks.
I'll
have
a
look
at
that.
There's
a
common
in
chat,
sorry,
but
we
should
have
a.
A
We
should
have
a
separate
CIA
build
for
all
the
feature
branches
and
we
should
have
a
set
for
et
built
for
the
future
branches
and
I
think
honestly
that
they
are
not
necessarily
working
correctly.
That
is
something
that
Travis
might
not
I,
don't
know.
I,
guess
the
table
is
good.
I,
don't
I,
don't
know
whether
Travis
can
do
it
I,
don't
to
say
they
can
do
it,
but
we
probably
just
have
to
set
up
some
more
jobs
and
proud
to
make
that
work
correctly.
I.
A
F
Okay,
the
master
is
easier
to
know
right.
Okay,
so
that's
great.
We
re
based
on
what
stop
and
see
about
that.
Thank
you.
Thank
you
very
much
cool.
Thank
you.
Many
wait.
Can
we
like
create
a
an
issue
for
the
changes
you
suggested
for
Travis
so
that,
if
anybody
wants
to,
you
know,
carry
on
with
the
effort
put
in
their
brilliant
yeah.
A
F
Yes,
so
something
like
as
a
continuous
improvement
document
or
something
pretty
much
was
done,
but
now
synchronise
does
they'll,
be
very
helpful
to
you
know,
drive
cops
develop
mentors
where
we
would
love
for
it
to
be
yeah.
That'd
be
great.
Thank
you
all
right,
so
I'll
ping
you
later
and
set
up
with
the
suggestion.
F
G
A
A
I
I
mean,
if
verify
based
basil
fails,
the
the
trick
is
or
the
command
is
to
run,
make
basil
gizelle
she,
the
e,
maybe
two
CS
I,
don't
remember,
but
the
we
also
needs
yeah.
Thank
you
for
drawing
attention
to
that,
because
that
the
error
output
is
not
helpful.
It
should,
we
say,
run,
make
basil
gazelle
and
it
should
probably
tell
us
why
it
failed.
It's
all
verification,
but
yes,.
A
Okay,
next
on
the
agenda
is
Oh
me
well,
why
don't?
Why
don't
I
show
the
cherry
fix
spreadsheet
first
well,
I'll
swap
the
order,
because
the
cherry
fix,
appreciated
sort
of
comes
more
naturally,
which
is
I,
will
try
to
share
a
screen.
Let's
see
if
I
can
do,
this
should
I
pick
see,
share.
D
A
Sorry,
this
is
I,
guess
I
have
to
share
my
whole
other
go
okay,
so
I.
Can
you
see
a
spreadsheet
with
lots
of
pretty
colors?
Yes,
okay,
so
we
now
have
I
think
at
least
four
branches
that
are
very
active.
We
also
have
master.
Of
course
we
have
111
112,
113,
114
and
I
am
trying
the
way
I'm
sort
of
trying
to
do
this.
I.
Imagine
we
will.
This
is
I.
A
In
my
mind,
this
is
the
beginning
of
a
process,
and
a
spreadsheet
is
a
great
way
to
sort
of
prototype
a
process,
but
so
for
every
PR
that
goes
into
the
master
branch.
Oops.
Sorry
didn't
necessarily
won't
click
on
some
PR,
okay,
so
every
PL
that
goes
into
the
master
branch
I
capture
it
from
the
logs
and
then
do
we
want
to
port
it
to
one
for
gene
and
do
we
want
to
put
it
to
114
I
mean
this
one.
A
So,
for
example,
this
one
report
it
to
both
or
cherry-pick
to
both
this
one
to
neither
this
one
to
113
only
but
not
to
114
this
one
to
114,
but
not
113,
and
then
the
colors
are
green,
I
hope
everyone
can
see.
Reading
green
by
the
way
this
this
color
is
green
and
green.
This
color
is
for
ones
that
are
all
done,
and
these,
though
the
whites
are
aware,
I
have
not,
or
the
terrific
I've
done
the
cherry-pick.
So
it
has
a
why,
but
it
hasn't
necessarily
emerged.
A
It's
not
a
feature,
but
the
risk
is
always
if
we
don't
cherry
pick
something
and
the
same
with
that
like
there
are
some
ones
that
are
updating
constantly
the
risk
is
if
we
don't
cherry-pick
things
that
it
causes
another,
a
future
PR
to
not
trip,
including
me,
because
it's
a
conflict.
So
it's
not
trivial,
and
maybe
we
should
just
cherry-pick
everything
until
we
do
the
releases
or
until
we
get
them
to
beta
I.
Don't
I,
don't
know
how
people
feel
like.
A
Anyone
have
any
views,
or
should
we
just
keep
doing
this
and
see
where
it
goes?
You
don't
happen
to
have
this
automated.
Do
you
it
does
not
automated.
Oh
boy.
Well,
yeah,
it's
not
I
mean
I
have
a
script
where
I
can
type
in
these
numbers,
but
yeah.
It's
not
automated.
So
I
am
wondering
whether
it's
better
to
just
rethink
everything.
It
would
be
more
I
guess
be
more
automated
yeah.
We
could
actually
have
a
script
that
did
that
and
was
automated.
A
Okay,
anyway,
I
wanted
to
share
that
spreadsheet
and
put
that
in
people's
minds.
If
you
wanted
to
have
any
have
any
thoughts,
please
do
let
me
know
and
yeah.
We
can
think
about
how
to
do
this
it.
This
will
get
much
less
bad
when
like
when
we
release
all
of
these
versions,
I
mean
I,
think
111
is
released,
but
as
we
do
that
as
we
right
now,
we
are.
We
did
this
sort
of
feature.
A
We
did
the
sort
of
catch-up
move
where
we
are
basically
cutting
future
branches
for
everything
I'm
guessing,
so
we
don't
become
bogged
down
and
cherry
picking.
We
should
start
typing
everything,
but
I
don't
know.
I
I
can
I
hope
that
answers
to
that
anyway,
I
wanted
to
share
that
and
then
the
real,
exciting
thing
I
hope,
which
is
the
next
thing
on
the
agenda,
is
also
mine,
which
is
about
why
we
did
all
these
feature.
A
A
So
we
are
more
able
to
do
things
that
are
we're
not
playing
catch-up
on
the
we're,
not
playing
catch-up
on
the
like
behind
on
every
kubernetes
release,
in
terms
of
not
being
able
to
merge
features
because
it
will
delay
the
new
release.
But
so
here
we
are
so
the
the
features
I
am
hoping
to
get
in
our
ones.
We've
been
talking
about
I,
don't
think
any
of
these
are
new
to
anyone.
I
hope
not.
A
So
if
we
can
use
bundles
to
update
components
outside
of
cups,
binary
binary
releases.
So,
for
example,
when
a
helpful
person
contributes
to
fix
at
CD
manager
right
now,
we
would
have
to
release
a
new
version
of
cups
to
include
that
bundles
are
a
format
proposed
by
actually
a
Googler
called
Josh
oak.
It's
an
open
source
project.
They
basically
wrap
manifests
into
a
Ciardi
called
a
component
and
they
have
lots
of
nice
features,
but
we
can
use
them
fairly.
Basically,
as
a
sort
of
me
llamo
format
and
we
can
specify
the
at
CD
manager.
A
There's
a
bit
of
nuance
there
in
terms
of
like
how
does
updating
work,
you
want
to
sort
of
automatically
update
and
how
do
we
make
sure
you
don't
need
internet
access
all
the
time,
but
as
long
as
we
get
it,
I
think
the
goal
should
be
that
we
get
it
to
be
at
the
same
place
as
channels
are
where
you
don't
where
you
you
don't
need
it
to
run
your
cluster,
but
you
might
need
it
to
update
your
cluster.
For
example,
yeah-
and
it's
also
very
easy
to
mirror.
A
A
So
if
we
get
that
I
think
that
would
be
great,
and
then
we
have
had
this
effort
called
cop
server
for
a
while
that
uses
aggregated
API,
aggregated
API
server,
which
is
sort
of
not
the
recommended
way
to
do
things.
Unless
you
really
need
a
Pio,
irrigation
and
now
see
are
these
are
basically,
if
not
at
parity,
very
close
to
parity
and
being
there's
a
lot
of
efforts
to
make
them
get
them
to
parity.
So
we
can
essentially
take
that
most
of
the
same
code,
very
few
code
changes
but
switch
to
CR
DS.
A
That
will
let
you
run
in
theory
on
anywhere,
like
I,
think
you
could
put
cup
server
on
gke
or
on
e
KS,
but
basically
you're
able
to
run
your
store.
Your
clusters,
and
instance,
groups
on
energy
in
a
kubernetes
cluster
instead
of
s3,
will
still
support
this
three.
But
this
is
another
way
of
doing
things
and
what
what
it
really
enables
is
the
next
thing,
but
I
did
want
to
highlight
one
challenge
which
is
I.
A
We
use
an
API
group
of
cops
and
the
rules
on
CR
DS
are
that
it
has
to
be
have
a
dot
in
it.
So
we
have
to
be
like
cops
our
case
that
I
am
I,
have
a
PR
that
basically
makes
this
transparent
for
writing.
So,
if
you
have
manifests
that
use
the
old
API
group
cops
they'll
continue
to
work,
the
the
user,
visible
change
will
be
when
you
cops
get
cluster.
Oh
yeah,
mole,
for
example,
you
will
get
back,
something
which
says
cop
stop
gates
to
IO
/v
1.
Alpha
2
is
the
current
version.
A
So,
if
you're,
comparing,
if
you're
doing
like
a
sort
of
string
comparison,
you
will
likely
have
to
change
your
string
comparison,
your
source
strings,
but
if
you're
doing
cops,
apply
or
something
we
basically
right
on
the
fly
or
cops,
replace
or
cops,
create
or
anything
any
of
those
ones.
You
should
still
be
able
to
use
the
either
the
old
form
or
the
new
form
without
any
problems.
I,
don't
have
anyone
sees
any
obvious
breakage
to
their
workflow
because
of
that.
A
The
I
think
we'll
still
need
the
instance
group
in
the
cluster
and
you're.
Basically,
the
machine
tooling
were
sort
of
reference,
the
instance
group,
but
other
than
that,
the
the
same
controllers,
more
or
less,
should
just
work.
This
should
enable
this
should
make
it
easier
to
port.
Other
clouds
should
make
rolling
updates
easier,
should
integrate
better
with
the
cluster
autoscaler.
All
of
the
great
things
that
are
coming
with
cluster
api.
A
This
is
only
for
nodes,
not
for
masters,
at
least
to
start
with.
Just
in
terms
of
scoping
masters
are
much
trickier
and
cluster
api
is
itself
v1,
alpha
1,
so
it
is
still
very
alpha.
So
we
are
not
going
to
make
this
required
or
even
necessary
recommended
because
it's
an
alpha
api,
but
as
that
goes
to
beta,
we
can
certainly
make
it
a
recommended
option.
I
don't
know
whether
we
would
ever
make
it
the
only
option
but
or
at
least
not
make
it
the
only
option
on
the
clouds
which
we
currently
support,
but
maybe
like.
A
If
a
new
cloud
came
along
that
wanted
to
only
support
cluster
api,
that
would
be
that
would
be
a
viable
path
and
then
finally,
sort
of
in
conjunction
with
bundles.
We
can
start
to
look
at
some
more
operators
to
deploy
things
a
little
bit
more
safely.
Like
we
have
this
script
script,
we
have
this
tool
called
channels
right
now,
which
is
the
thing
which
actually
applies
your
add-ons
and
it's
not
perfect,
I
think
we
had
a
bug
recently,
where
you
know
when
you
start
to
change
options
on
an
add-on.
A
Your
whole
cops
cluster
just
because
you
tweet
some
small
setting
like
if
you
change
the
verbosity
of
a
CD
manager,
you
shouldn't
necessarily
have
to
restart
your
cops
cluster
or
it's
probably
today
you
do
and
that
net
so
that
that's
that
sort
of
my
goals
for
1:15
I
think
if
we
can
get
all
those
into
1:15,
I
think
I'll
be
pretty
happy
but
yeah.
That's
sort
of
why
we're
doing
all
these
feature?
Branch
branches.
A
And
he
and
want
to
add
anything
in
there
and
want
to
tell
me
not
to
do
something
comments.
Questions
concerns
cool
all
right,
Ryan,
you
have
a
question:
I'm
gonna,
swap
your
two
items:
Ryan
yeah
over
least
time
line;
first,
maybe
yep.
Just
what's
the
plan
I
think
we're
looking
pretty
good
I
think
they
only
hold
up.
A
Is
you
have
an
issue
you
found
around
STD
manager,
so
I
need
to
dig
more
into
that
I
think
they've
also
been
some
other
reports
which
I
need
to
dig
into
to
make
sure
that
they
are
known
and
I.
Think
they're,
mostly
around
the
docks
and
around
the
edge
TV
version,
locking
like
I
think
last
two
weeks
ago,
I
think
we
had
a
big
list
of
things
that
needed
better
documentation.
A
I
haven't
had
time
to
do
that,
but
I
think
one
of
the
real
things
that
that
came
up
as
a
like
code
change,
that's
needed
is
the
fact
that,
if
you're,
using
like
a
an
unusual
version
of
STD,
which
isn't
even
necessary
that
unusual,
like
it
just,
doesn't
work
with
that
CD
manager.
Right
now,
because
I
see
my
dress
as
a
support.
A
small
set
of
supportive
it
CD
versions,
so
we
need
to
expand
that
to
set
of
supported,
STD
options
and
or
at
least
warn
people
if
we're
gonna,
like
totally
turn
off
at
CD.
A
E
A
E
D
A
F
A
A
A
A
A
A
A
A
I
will
I
will
I
will
make
that
link
public
or
replace
the
link
with
a
public
link
when
I'm,
not
under
pressure
of
trying
to
figure
out
that
Mary.
My
log
works
that
catch
Google,
employee,
joke.
A
Exactly
yes,
I
can
do
this,
okay,
I
think
anyway,
I
will
I
will
figure
this
out
worst
case.
I
will
copy
it
to
another
account
where
I
definitely
can't
your
doctor
publicly
real.
C
Quick,
so
I
was
just
looking
at
that
issued
the
Travis
issue,
and
so
does
it
go
format,
issue
or
not
I
basil,
because
I
was
exactly
what
you
expected
Justin,
so
they
need
to
run
basil
gazelle.
So
it
looks
like
in
Travis
right
now.
It's
not
outputting
that
error
that
we
do
for
the
mare.
You
know
when
verify
basil
fails:
it's
not
outputting.
So
two
questions
really.
First,
should
we
just
you,
you
stripped
basil,
some
of
the
basil
stuff
out
of
there.
C
A
Yeah,
if
we
have
a
if
we
have
a
if
we
have
a
proud
job
that
is
verifying
at
which
all
of
these
things
were
done
in
prowl,
then,
okay,
general
and
I
think
that
one
was
okay,
because
it's
like
for
some
weird
linking
gazelle,
the
Basel
generator
doesn't
actually
need
basil.
So
that's
why
it's
sort
of
weird
like
we
get
away
with
it.
But
yes,
if
we
have
the
similar
job,
there's
no
reason
to
there's
no
reason
to
do.
That's
have
it
in
Travis
as
well.
Okay,
our
all
together,
perfect!