►
A
And
we're
live
hello
again
after
we
explored
abstract
last
week
and
did
a
little
bit
of
observability
and
monitoring.
I
asked
max
and
everyone
else.
A
If
we
might,
we
want
to
continue
with
our
kubernetes
setup
and
we
had
some
quirks
with
deploying
things,
and
hopefully
this
week
is
another
week
where
things
work
again
and
just
as
a
reminder,
we
have
been
working
to
to
use
terraform
to
provision
the
cloud
resources
in
headstart
cloud
and
there
was,
I
think
there
was
a
race
condition
or
something
which
did
not
work
with
the
terraform
module
talking
to
the
api
and
at
some
point
we
left
in
the
unsatisfying
state,
I
would
say,
but
yeah
that's
about
it.
A
So
we
have
the
gitlab
cicd
yammer
around
and
we
can
basically
start
redeploying.
Maybe
what
do
what
do
you
think
max.
B
Yeah
we
as
an
attempt
to
fix
it.
We
added
this
parallelism
one
I
would
just
I
don't
know
just
remove
this
and
commit
it
and
oh
no,
we
don't
even
have
to
commit
anything.
We
just
can
go
to
gitlab
because
we
only
triggered
the
weapon
by
the
web
interface
anyway.
A
A
Only
that
terraform
0
16,
I
think,
was
released
or
0
15
and
it
was
like
introduced
as
the
way
to
the
first
1.0
release
if
everything
works
as
planned.
A
Works,
the
probably
yeah,
I
think
I
I
wanted
to
create
a
new
xkcd
303
excuse
not
just
like
something
is
down,
but
the
pipeline
is
running
like
what
you
are
doing
with
your
time.
Yeah,
I'm
slacking
off,
because
the
pipeline
is
running,
then
you,
then
you
choke
from
10
years
ago
or
something
it.
C
A
That,
indeed,
sometimes
the
problem
is
so
complicated
that
it's
really
hard
to
find
a
spot
where
you
can
like
look
into
it.
But.
D
C
D
I
think
we
should
increase
the
speed
when
we
would
use
our
own
runners.
A
C
A
I'm
open
of
doing
that.
The
thing
is,
if
it's
not
needed
urgently,
it
puts
probably
another
another
one
hour
or
two
hours
of
installing
it
and
updating
it
and
maintaining
it.
So
could
totally
do
it.
On
the
other
side,
I
was
thinking
of
another
thing:
we've
do
at
gitlab,
for
example,
for
like
mono
wrappers,
where
you
have
different
use.
Cases
to
fulfill
have
some
sort
of
detection
in
your
pipeline,
whether
you're
deploying,
for
example,
only
the
block
on
the
website
or
to
a
full
deployment
or
just.
A
I
think
there
are
checks
for
the
for
gitlab
as
a
project,
whether
the
frontend
components
are
actually
affected
so
like
doing
a
git
diff
on
on
the
current
ref
and
and
then
deciding
whether
the
the
front-end
tests
need
to
be
run
or
if
the
test
is
only
affecting
the
back-end
and
things
like
that,
because
otherwise
I
have
no
idea
how
many
tests
we
would
be
running,
but
I
think
it
should
be
like
500k
or
1
million,
and
this
takes
quite
a
while.
B
So
good
thing
the
deployment
just
worked,
so
I
would
stay
with
the
yeah
assumption
that
something
was
wrong
with
head
snack
cloud
last
week
because
we
didn't
change
the
thing
now
and
it
just
worked
flawlessly.
So
I
would
remove
this
parallelism
one
again
just
make
it
a
little
bit
faster
again.
B
So
I'm
using
a
similar
setup
for
my
personal
kubernetes
cluster
and
there
I
also
use
ansible
and
one
thing
with
ansible-
is
that
it's
a
python
tool
and
you
need
to
install
all
the
python
dependencies
to
run
it,
which
takes
a
little
bit
of
time.
So,
for
example,
in
our
ansible
directory,
we
have
this
requirements
yeah
the
requirements
txt,
where
we
tell
which
ansible
version
we
want
to
use
and.
B
Know
100
dependencies
or
something
that
are
all
installed
each
time
or
if
we
run.
If
we
run
this
in
a
gitlab
runner,
we
will
install
those
dependencies
every
time
which
is
not
nice.
Looking
at
the
runtime
of
the
ci
drop,
and
it's
also
not
so
nice
for
the
package
registry
pi
pi
of
python,
because
we
pull
those
packages
all
the
time
again
and
again,
which
is
a
little
bit
inefficient.
B
B
B
B
B
B
Yeah
by
the
gitlab
environment
already
so
we
just
add
a
name
and
attack
and
for
now
we
just
go
with
latest.
So
each.
A
B
We
can
probably
iterate
over
this
later
and
add
versioning
to
it.
What
can.
B
A
B
So
in
our
before
script,
we
have
to
do
a
little
bit
of
a
work
around
because
the
back
here,
the
ssh
key,
is
used
as
a
file
and
ssh
doesn't
like
it.
If
your
private
key
is
public,
readable,
and
so
it
will
not
use
this,
then
so,
let's
just
take
all
away
all
the
permissions
for
the
group
and
other
accounts
from
this
file.
So
we
can
actually
use
this
key
and
also,
let's
switch
into
the
subdirectory
ansible,
because
it's
the
directory
we
have
or
our
stuff
in
front.
B
B
So
I'm
curious:
if
this
will
even
work
with
the
shared
runners.
I
guess
we
have
to
look
it
up
a
little
bit.
What
we
can
do
if
we
can
use
some
text
to
use
a
shared
runner
that
allows
us
to
to
use
docker
and
docker.
We
will
see-
because
I
usually
run
this
in
those
headset
autoscaling
runners
we
do
a
while
ago
and
those
just
work
with
docker,
and
I
suspect
that
this
will
not
work
out
of
the
box
with
gitlab
runners.
If
anyone
knows
how
to
use
docker
in
shared
gitlab.
A
Is
there
are
pr
pre-built
templates
available?
If
you
want
to
use
docker
and
docker?
Let
me
see,
because
I've
been
creating
a
pull
request
around
this
as
well
for
updating
it
for
the
main
branch,
and
I
might
have
an
example
in
a
test
playground
repository
which
you
can
just
copy
paste
or
somehow
copy
paste.
A
A
A
I
don't
know
if
the
stage
is
correct,
but
I
think
we
don't
have
a
build
stage
and
the
all
the
entire
logic
of
building
the
base
using
the
base
image
of
an
antibody
should
be
done
in
in
a
docker
file.
Yeah,
you
don't
need
to
modify
the
script.
A
This
is
basically
wrapping
or
detecting
which
branch
it
is
on
and
then
doing
the
build
stuff
and
taking
because
previously
this
has
been
two
jobs
and
we
thought
about
making
it
a
compatible
with
either
main
or
master
branch
so
making
it
like
using
the
default
branch
name
and
also
combining
it
into
one
scripture
or
on
the
job,
because,
okay.
A
A
Yeah,
you
should
be
using
late
test.
I
think.
B
D
D
A
A
All
right
little
brain
sorry
for
the
foot.
If,
but
it's
it's
easier,
if
it's
hidden
away
somewhere,
you
just
build
it
and
get
attacked,
and
the
rules
also
take
into
account
that
you're
in
a
commit
branch
and
the
dockerfile
exists.
A
B
B
And
then
we
just
want
to
want
to
run
pip,
which
is
already
available
in
the
python
image,
because
we
don't
have
any
cache
available
in
our
runner.
We
can
also
just
disable
it
for
pip
and
install
everything
from
here
like
ansible
and
the
h
cloud
package,
and
if
you
remember
what
we
did,
I
don't
know,
maybe
months
ago
was
also
use
ensemble
galaxy
to
install
those
dependencies.
B
B
I
think
we
can
add
here
right
changes,
because
we
only
want
to
do
this
for
some
specific
files,
because
we
do
it
each
time,
then
the
joke
is
on
us,
because
then
we
don't
reduce
our
amounts
of
yeah
installing
python
dependencies,
but
if
we
do
it
only
if
there's
stuff
changed
in
yeah,
we
can
also
remove
this
one,
because
we
don't
have
this,
but
only
if
we
change
the
docker
file,
the
requirements
txt
or
the
requirements
amplify
only
then
we
want
to
build
this
new
image.
Otherwise
we
just
want
to
use
the
existing
one.
D
D
You
can
say
doing
data
build,
you
can
say
docker
a
bit
dash
dash
cash
from
and
before
you
pull
in
the
latest
image
and
use
this
latest
image
as
cache.
So
then
the
the
transaction
will
also
hit.
D
B
B
Trigger
the
ci
run
over
the
web
interface,
then
I
think
the
changes
will
not
be
evaluated.
It
will
just
trigger
all
the
time
if
I
remember
correctly,
there's
something
in
the
docs
about
this-
that
it's
yeah,
that
this
only
works.
If
you
have
a
merge
request
or
yeah
only
if
you
refer
to
a
commit
like
if
it's
a
push
event
or
match
request
event,
I
think.
B
We
should
have
just
installed
it
out
of
the
each
time
from
scratch
that
would
have
been
so
much
easier.
Okay,
so
the
job
is
definitely
not
triggered
over
here.
D
D
Currently
in
our
company,
we
doing
for
one
applications
that
we're
doing
currently
a
monorail
approach
and
also
with
child
pipelines
and
all
the
stuff,
and
it
works
quite
flawlessly.
D
But
for
that
I
need
also
to
read
about
how
all
this
real
stuff
is
working,
because
before
that
I
use
mostly
only
I
used
only
only
an
except
yeah,
okay.
A
If
it
still
works
with
rules,
it's
more
like
a
programming
language
approach,
I
would
say,
but
I
think
there
are
some
patterns
you
cannot
really
catch,
which
you
can
with
only
and
accept.
So,
for
now
both
will
stay.
It's
not,
I
would
say
the
best
way,
because
it's
two
two
ways
to
do
it
right,
but
still
over
time
we
will
find
a
way.
A
The
current
approach
for
making
main
the
default
git
branch
is
to
use
only
except
with
variables,
which
is
also
a
possible
thing,
so
you
can
do
easy
comparisons
with
ci
variables
in
in
only
and
accept
see.
I
need
to
find
the
merch
request
for
my
docker
build
template
which
references
to
epic,
where
this
is
being.
A
Let's
I
see
python.
B
And
as
we
can
already
see
it,
or
it
takes
almost
three
minutes
and
isn't
finished,
that's
why
I
usually
prepare
to
prefer
to
do
this
once
in
the
base
image
and
then
only
update
it
for
it
for
changes,
because
it
speeds
up
the.
A
A
You
can
I've
posted
a
link
in
the
zoom
chat,
which
is
the
birth
request
for
the
docker
ci
template
build,
and
there
is
we
discuss
that.
Just
looking
for
the
issue
which
I
will
be
solving.
B
B
A
Me
it's
like
I'm
on
I'm
on
fire,
which
means
I'm
super
busy,
and
I
don't
know
when,
when
when
work
stops
yeah
I
know
when
I
moved
to
the
couch
and
continue
watching
westworld,
which
I
kind
of
missed
in
the
past
five
years.
But
yeah
I've
been
I've
been
quite
busy
and
I'm
toying
you
can't
you
can
set
the
emoji
as
a
busy
state
or
as
a
state
in
your
profile,
so
yeah,
basically
modified
by
state
yeah
on
on
the
upper
right
hand,
corner.
A
A
The
issue
which
says
how
to
use
the
div,
let
me
see
if
I
have
another
match
request
around
I've,
been
working
on
some
contributions.
Lately.
Do
I
have
a
csc
playground?
A
B
B
B
Apparently
I
tested
something
for
this
two
weeks
ago.
Already,
that's
interesting,
no
memory
of
that
okay.
Anyway,
we
have
this
image
here
now,
which
means
we
can
use
it
and
I'm
pretty
sure
that
this
job
will
trigger
each
time
we
hit
the
run
pipeline
button
now,
so
I
propose
that
for
now
we
just
disable
it
because
it
will
cost
us
five
and
a
half
minutes
each
time
we
click
on
run
and
then
in
a
later
stage,
when
we
change
our
pipeline
a
little
bit
to
run
for
each
push.
B
A
B
A
Means
it's
a
template
import
me.
I
would
yeah
in
my
world.
I
would
rename
extends
to
import
but
different
story
kind
of
used
to
what
we've
designed
with
our
singer,
2
in
the
dsl
so
yeah.
If,
if
the,
if
the
configuration
doesn't
tell
you
by
the
keyword,
what
it
does,
I
tend
to
name
objects
and
shop
names
in
a
way
that
I
can
immediately
see
what
it
does.
B
Yeah,
that
makes
sense,
so
we
want
to
run
our
ansible
playbook
and
we
want
to
use
our
existing
ssh
key,
because
we
know
that
this
key
has
access
to
all
our
machines,
because
this
is
the
same
key
we
deploy
with
terraform
and,
I
think
yeah.
It's
called
main
yammer,
it's
our
playbook
and
yeah
in
the
diff.
We
can
completely
skip
the
cube
conflict
jumps
and
we
run
it
in
check
mode.
B
And
I
think
it's
also
like
this-
I
don't
think
we
need
the
div,
because
if
I
remember
correctly,
we
added
this
in
our
answer-
configuration
but
yeah
over
here
we
already
said
always
display
a
div,
that's
nice,
and
why
we
skipped
the
cubeconfig
tag
is
because
in
this
row
we
pull
the
cubeconfig
to
our
localhost,
fix
the
hostnames
in
it,
and
this
way
have
a
working
cubecontract
to
talk
to
our
deployed
cluster.
But
in
this
stage
we
don't.
We
neither
need
this
cube
config
yet,
nor
do
we
have
it
because
there
is
no
cluster
deployed.
B
B
B
B
B
B
B
B
Interesting,
we
need
to
decide
where
we
store
this
cube
config,
because
what
we
want
to
do
is
to
store
this
cube
contract
as
an
artifact.
So,
as
a
user,
we
can
download
it
from
our
gitlab
interface
and
connect
to
this
created
cluster,
but
I'm
not
entirely
sure
how
I
thought
we
would
do
that
anyway.
What
we
need
is
an
additional
in
additional
variable
where
we
overwrite.
B
B
C
B
A
Tell
the
audience
I
was
just
thinking
around
like
predefined,
ci
cd
variables
and
having
the
k8s
prefix.
If
there
is
something
in
there,
it's
just
I'm
not
sure,
and
I
wanted
to
google
myself.
C
B
B
So
we
can't
just
like
download
this
cube
config
file
in
the
home
directory
of
the
current
user
or
something.
Instead,
we
have
to
put
it
somewhere
where
it's
in
the
scope
of
our
repository
and
only
then
we
can
use
it
as
an
artifact
and
we
can
use
the
same
variable.
We
used
a
few
times
before
ci
project
directory,
because
that's
already
like
the
directory
of
our
current
git
repository,
so
everything
in
there
is
available
if
we
want
to
as
an
artifact,
and
so
we
just
want
to
put
our
q
config
file
in
there.
B
Let's
call
it
dot,
cube
config
and
what
happens
down
here
in
our
keep
conflict
drop,
which
I
probably
like
confu,
which
is
a
little
bit
confusing,
because
I
was
thinking
about
it
while
talking
so
that
might
not
have
been
obvious.
But
what
we
tell
ansible
here
is
to
override
this
variable
with
the
contact
content
of
this
variable.
B
A
A
B
B
A
B
C
A
Tried
my
googly
pho
is
good.
I've
seen
it
in
the
release
kickoff
for
13.12
on
monday
and
if
I
can
paste
it
in
the
zoom
chat,
so
you
can
subscribe
and
follow
along.
B
B
On
the
other
hand,
yeah
those
stages
show
us
the
order
in
which
it
will
be
executed,
so
it
I
mean
that
makes
the
graph
a
little
bit
white.
I
think
otherwise,.
A
A
The
only
thing
is
sometimes
I
keep
using
that
because
the
code
quality
scans
take
like
10
minutes
or
so,
but
meanwhile
you
could,
for
example,
do
some
up
until
building
something
this
could
be
still
could
be
still
speed
up.
Speed
up
be
improved.
Let's
move
different
different
english
terms,.
A
B
Primary
you
mean
because
we,
you
know,
we
set
it
in
our
inventory
over
here
right
yeah.
I
think
you
are
right.
D
Aeroplane,
I
I
forgot
to
yeah,
I'm
currently
sitting
near
the
airport.
B
A
The
problem
is,
I
would
love
to
use
more
emojis,
but
some
are
not
export
supported.
A
You're
playing
around,
I
think,
yeah
you
can.
There
is
also
a
new
business
state
since
a
few
versions
where
you
can
just
say,
hey,
I'm
busy
and
never
respond
to
anything.
Oh
just
kidding.
A
We
can
talk
it
over.
The
reason
I
was
late
today
is
I'm
running
a
blog
post
around
using
check
you
and
jason,
and
I've
also
found
a
way
to
use
this
d8
api
endpoint
for
ci,
linking
or
for
linting
a
gamma
configuration
from
the
cli,
and
I
wrote
a
github
for
it
on
monday
and
the
blog
post
is
about
to
be
published.
Then
I
need
to
do
a
recording
for
the
gitpod
community
day
next
week,
then
I
have
a
dockercon
talk.
I
have
an
slo
conf
talk.
A
I
have
some
talks
coming
up
and
maze
kind
of
busy,
but
I
was
thinking
of
maybe
trying
up
stress
ourselves
or
maybe
on
on
a
different
side.
Continue
with
the
promises
operator.
We've
been
talking
about
for
a
while
or
use
exploring
tanka
and
chase,
not
jason
net.
B
A
B
Yeah,
I
use
it
a
little
bit
to
create
the
namespaces
and
network
policies
in
my
kubernetes
cluster,
but
I
wouldn't
say
I
have
the
qualification
to
like
give
an
introduction
to
it.
I
mean
we
can
figure
it
out
together,
but
it's
not
I'm
just
tried
it
out
a
while
ago
and
since
then
I'm
using
it-
and
I
don't
really
know
if
what
I'm
doing
is
the
way
you
are
supposed
to
use
it
and
everything.
A
Yeah,
maybe
if
we
cannot
find
someone,
we
could
do
it
the
other
way
around,
like
we
kept
learning
rust
together,
not
the
two-hour
edition,
but
like
reading
the
tutorial
and
trying
things
out
and
then
and
then
like
see
how
things
are
going,
because
I
I
would
love
to
encourage
everyone
to
like
contribute
to
the
session
and
not
just
staring
at
the
screen.
Saying
press
this
button,
but
maybe
changing
roles
and
saying
hey
one.
One
does
decoding
and
four
others
look.
A
Okay,
that's
a
it's
an
austrian
building
place,
but
a
construction
place
construction.
Yet
whatever
it's
super
hard
to
translate
in
english,
but
yeah,
maybe
like
try
something
else
out
of
the
sudden
and
you
have
little
experience
and
if
we
have
a
success
together
it
it
weighs
more
than
it
weighs
it
weighs
a
lot,
not
comparison,
not
comparing
anything.
A
A
A
Maybe
like
first
come
first
serve,
propose
a
topic
for
one
week
and
then
we
do
it
and
have
it
more
in
a
visual
way.
We
are
doing
that
in
our
developer.
Evangelism
handbook
as
well
just
to
share
it
really
quickly.
A
Just
cannot
type
today
we
have
a
team
calendar
in
our
hand.
B
I
don't
think
so,
then
the
client
usually
says
like
private
key,
is
not
allowed
to
be
public,
readable,
ignoring
it
or
something
like
this.
B
D
A
Can
you
just
quickly
open
the
the
real
url
essentially
soon
chat,
because
I
also
figured
that
tomorrow
is
release
day,
so
it's
release
evangelism
day.
A
Team
calendar
yeah,
we
kind
of
populated
that-
and
I
think
the
reminder-
oh
yeah,
and
on
friday
there's
a
session
around
gitlab
commit.
So
if
you
want,
if
you
want
to
provide
your
story,
I
would
encourage
you
to
just
talk
about
this
setup,
for
example,
or
what
we've
been
learning
in
the
past
or
some
other
experiences
you
might
want
to
share
and
on
friday,
there's
a
session
hosted
by
tron
on
creating
a
good
talk,
help
with
the
storyline
coaching
and
other
things
which
I
would
recommend
to
attend.
A
Yep,
and
if
you
speak
german
next
week,
is
the
german
virtual
connect
day
at
gitlab,
and
I
will
be
talking
about
pipeline
efficiency.
I
still
need
to
update
the
slide
deck
but
yeah.
Thank
you.
Con
is
upcoming
in
europe.
C
A
That's
about
it
and
I'm
thinking
about
using
the
same,
not
the
same
calendar
but
similar
approach
of
creating
our
own
team
calendar.
I
just
need
to
figure
out
how
this
works
with
permissions
and
such
because
I
cannot
share
the
zoom
url
in
public,
so
this
needs
to
be
a
separate
invite,
but
for
the
for
the
topics
and
for
the
ideas
collection,
we
can
totally
use
the
public
calendar.
B
Yeah,
I
think,
looking
at
the
clock,
we
should
also
cut
it
here
and
figure
out
why
ssh
isn't
working
off
camera
and
then
maybe
talk
about
it
next
time,
because
this
will
be
some
try
and
error
and
it
works.
A
B
B
A
In
some
way,
we
can
also
change
change
the
role
of
you're
doing
you're
doing
basically
nothing
and
we
instrument
you.
You
know
it
sounds.
B
Like
we
can
do
this,
like,
like
everyone,
tell
me
what
you
place,
play
sketlap
ci,
everyone
is
typing
and
who,
who
types
it
the
most
wins.
D
A
Yeah,
typically
pair
programming
session,
one
is
one
is
like
playing
the
the
type
robot
to
be
honest
and
the
other
one.
Instructs
with
maybe
do
this
so
do
this
and
trying
to
learn
together
even
more
just
thinking
out
loud
if
it
doesn't
make
sense,
ignore
me.
B
A
I
think
I
think
we
are
crazy
enough
to
figure
it
out
and
hopefully
see
each
other
next
week.
I
would
just
say
we
wrap
it
up
like
that.
Unless
someone
else
has
something
more
to
say,
then
I
would
say
we
say
bye,
bye
till
next
week.