►
Description
Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io
Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects
CI/CD in Light Speed with K8s and Argo cd - Billy Yuen & Parin Shah, Intuit
Enterprises have benefited greatly from the elastic scalability and multi-region availability by moving to AWS, but the fundamental deployment model remains the same.At Intuit, we have adopted k8s as our new saas platform and re-invented our CI/CD pipeline to take full advantage of k8s. In this presentation, we will discuss our journey from Spinnaker to Argo CD.1. Reduce CI/CD time from 60 minutes to 10 minutes.2. Reduce production release (or rollback) from 10 minutes to 2 minutes.3. Enable concurrent deployment using spinnaker and argo cd as HA/DR to safely adopt the new platform with no downtime.4. Be compatible with the existing application monitoring toolset.
To Learn More: https://sched.co/GrRs
A
Okay,
good,
okay,
hear
me
all
right
so
good
afternoon.
I
know
it's
kind
of
late
in
the
day,
but
you're
gonna
jump
around
a
little
bit.
That's
fine
too!
You
know
so
my
name
is
Billy
and
that's
her
in
they
form
into
it
with
a
platform
engineer
at
their
payment.
So
what
we're
gonna
talk
to
you
guys
about
today
is
how
we
kind
of
revamp
a
whole
deployment
system
with
kubernetes
and
experience,
and
the
speed
that
we
learn
up
came
from
the
whole
experience,
so
the
Jennifers
I
want
to
just
kind
of
go
through.
A
A
They
want
and
I
was
a
lot
of
that
I'll
talk
about
why
we
also
decided
to
build
a
CD,
continuous
deployment
to
go.
I'll
go
I'll,
go
CD,
and
after
that
pro
give
you
a
full
demo
end-to-end,
you
can
see
how
how
everything
works
together
sounds
good
before
I
start
does
anyone
know
about
you
see
spinnaker
that
type
of
couple
hands
at
least
some,
because
I'm,
basically
making
a
comparison?
Great,
ok,
good!
A
It's
not
I'm
trying
to
give
a
little
bit
more
detail,
but
it's
not
the
intent
for
this
talk
so
now
before
that,
of
course,
I'm
gonna
show
you
a
little
bit
talk
about
company
a
bit.
We
were
fully
into
it.
Just
make
sure
the
echo
company
mission
is
to
empowering
possibility
around
the
world.
What
we
mean
by
that
is.
A
We
want
to
provide
a
best
financial
tool
to
make
your
company
success,
successful,
whether
it
is
accounting,
payments,
payroll
or
tax
right,
of
course,
a
flagship
article,
turbo
tax
in
QuickBooks
I'm
sure
somebody
part
of
it
we're
actually
in
Silicon
Valley
in
Mondrian.
Next
to
Google
I've
companies
found
on
1983,
we
went
out
to
in
1993
and
we
have
about
9,000
employees
in
21
locations
globally.
We
did
about
six
billion
dollar
revenue
last
year
and
we
about
15
million
customer
worldwide.
So
just
give
you
some
context.
A
Our
company
size
now
for
DevOps
really
I
want
to
just
kind
of
give
you
kind
of
stuff
from
the
beginning
where
we're
at
you
know,
given
the
age
of
our
company,
we
started
from,
of
course,
data
center.
We
have
own
data
center
and
the
way
we
operate
before
it's
basically,
you
know
engineer
focusing
on
development
and
we
dedicate
an
ops
team
that
would
do
the
deployment
for
us
and
we
also
have
a
platform
infrastructure
to
maintain
our
data
center
infrastructure.
A
The
challenge
with
that
is
basically,
if
you
think
about,
let's
just
say:
we
go
to
excreme'
from
the
time.
I
develop
a
feature
by
the
time.
I
need
a
scheduled
deployment
time
that
will
take
about
three
weeks
with
all
my
weeks
to
get
features
out
and
then,
when
we
have
to
do
any
changes
for
infrastructure
mail,
I
need
two
orders:
new
server
or
network
changes.
Now
you
talk
about
months
right,
so
that's
kind
of
the
motivation
forced
to
move
to
AWS
right.
So
when
we
move
to
either
way
your
ass.
A
Essentially,
we
kind
of
make
engineer
that
responsible
for
both
develop,
of
course,
development
still
there
now
they
get
to
deploy
any
time
they
want,
but
they
also
need
to
own
the
infrastructure
it
turns
out.
Wasn't
quite
of
you
know,
for
my
learning
is
not
the
perfect.
It's
not
such
a
great
I
did
that
at
the
end,
because
what
we
really
want
is
allow
engineer
focusing
what
they
do
best,
either
Java
or
nodejs
or
JavaScript,
whatever
it
is
and
before
any
time
they
want,
but
we
don't
want
to
burden
them
with
all
the
ideas.
A
You
know
expertise
right
that
takes
away
time
for
them
to
focus
on
what
they
do
best
for
our
customers.
So
that's
why
we
decided
to
invest
in
a
new
Niska,
modern
sass,
so
that
we
can
solve
or
engineer
the
right
way.
Basically
let
them
you
know
code,
a
feature
as
much
as
that
software
customer
problem
and
deploy
anytime
they
want.
But
then
we
have
some
central
team
that
how
manages
infrastructure
and
also
give
a
common
way
to
deploy
code
so
and
I
just
want
to
give
you
a
kind
of
what
we
try
to
accomplish
here.
A
So
once
we
move
the
80s,
we
were
able
to
release
our
feature
under
under.
We
literally
from
you
know
some
coding
to
actually
get
it
now.
We
can
do
it
in
a
week,
but
of
course
we
like
to
get
food
shelter
customized
soon
as
we
can
right.
So
that's
what
we
decided
to
look
into
top
leveraging
dark
and
kubernetes
so
that
we
can
actually
literally
deploy
daily.
So
we
can
get
future
other
customer
as
soon
as
possible
and
also
the
other
side
benefit.
That
is
is
about
thing
about
recovering
right.
A
A
lot
of
time
when
you
have
a
deployment
problem,
the
first
thing
we're
gonna,
do
rollback
right
so
with
spinnaker,
because
it's
also
is
still
based
on
ec2
technology.
This
actually
takes
about
30
minutes.
It
will
roll
back
because
you
actually
have
spin
up.
You
know
all
your
ami
again
but
with
kubernetes
dr.,
because
it's
only
a
process
now
so
glitter
we
can
roll
back
in
three
minutes
right.
So
that
really
helps
with
our
recovery
too.
A
A
So,
if
you
think
about
what's
the
deployment
really
is
more
than
your
code
right,
you
write
the
code,
you
create
a
binary,
but
then
you
also
need
your
runtime
environment,
for
example
like
if
you're
running
Java,
you
need
your
jdk,
you
may
have
a
a
PMI
app
dynamics
to
monitor
your
JVM,
etc.
So
those
things
all
have
to
be
installed
in
your
environment.
So
then,
here
comes
the
problem
right,
so
we
need
to
deploy
that
every
time
into
your
am
I.
A
Let
this
the
left
size
model,
how
you
deploy
a
mine
spinnaker
in
your
build
time.
You
need
to
download
all
the
chef
script
that
actually
what
we
learn
is
actually
take
an
almost
extra
10
minutes
to
your
build
time,
because
you
have
to
download
a
lot
of
extra
things
and
pandora
into
your
and
package
and
I.
Don't
know
if
you
guys
know
about
ami
bakery,
there's
a
consecrate,
my
bakery
centuries.
You
can
actually
create
a
beautiful
image
with
all
your
dependencies.
A
What
it
really
does
is
actually
spin
up
a
PC
to
execute
all
your
chef
recipe
to
do
all
the
installation
and
then
snapshot
it
guess
what
that
takes.
Another
10
minutes
right
so
literally
the
whole
build
process
is
just
that
I'm
long.
We
take
20
minutes
right
so
moving
over
to
using
darker
and
kubernetes.
A
What
we
learn
is
really
remember,
darker,
just
a
virtual
I
file
system
right,
so
everything
we
needed
builder.
We
can
just
put
into
the
base
image
every
time
we
do.
We
just
overlay
with
the
app
code.
So
now
the
build
times
are
saying
between
the
apt,
oh,
but
the
rest
of
the
time.
Loader
takes
about
a
minute
or
less,
because
all
we
need
to
do
is
just
lay
the
co
over
and
then
create
a
docker
image
for
deployment
and
check
in
it.
A
There's
one
more
subtlety-
and
it's
gonna
mean
something
later
on
in
the
talk
is
when
you
need
to
deploy
so
that
just
your
binary,
but
you
also
need
to
get
application
configuration
to
change
for
your
vibe,
for
example,
your
cue
environment,
to
be
different
than
your
that
environment,
if
that'll
be
setting,
will
be
different,
etc.
So
you
need
to
application
external
to
your
apps.
On
top
of
that,
you
also
need
to
have
something
to
describe
your
infrastructure.
For
example,
your
pre
prod
environment
might
only
have
one
instant
of
two
instances
versus
your
problem.
A
I
have
five
to
ten
right,
so
you
need
some
way
to
describe
the
infrastructure
now
with
spinnaker.
The
spinnaker
basically
is
a
UI
to
the
database,
so
what
they
do
is
they
keep
track
of
the
easy
to
size
or
whatever,
no
buttons
in
the
database
in
kubernetes.
The
the
difference
is
now.
You
actually
have
true
info
structures
code
right,
you
describe
your
environment
using
gamble,
so
this
is
something
because
we
like
to
embrace
get
ops
for
basic
audibility
and
also
control.
That's
a
huge
win
for
us.
A
That's
and
I'll
show
it
in
a
minute
now
so
I
mentioned
about.
We
invest
in
this
big
infamous
whole
epic
initiative
call
into
modern
cess,
especially
as
we
try
to
get
a
you
know,
the
coupe
nad
serviced
along
with
all
the
monitoring
we
look
at
the
middle.
It's
like
eff
dynamics
way
fun
spoiling
all
those
things
integrated
into
basically
deployment
even
set
in
kubernetes,
along
with
what's
on.
If
you
look
at
the
get
help
that
line
right
there,
these
are
all
the
one
we
use
the
technology
we
use
for.
A
Csc
is
use
CI,
CD
pipeline
races,
so
we
have
consistent
panel
for
every
development.
It
doesn't
matter
is
a
small
job
versus
a
full-service,
and
we're
gonna
also
talk
about
something
little,
not
quite
that
regulated,
but
is
how
we
can
Tony
banish
all
the
cluster
member.
We
have
many
many
cluster
within
a
company.
We
also
have
a
homegrown
to
call
AI
KSM
that
will
help
us
managing
multiple
cluster,
so
of
course,
well.
I'll
talk
in
particular
we're
just
gonna
focus
on
a
CI
CD
pipeline
that
consists
of
using
github
flourish
and
drinkin
or
Jenkins
2.0.
A
Other
factories
are
talk
of
registry
and
continuous
deployment
called
Archos
TV,
which
we
open
source
to
the
public
and
I
believe
how
to
talking.
If
I
mentioned
that
already
now
I
know
the
first
question
some
people,
my
ass,
is
why
not
just
use
spinnaker,
because
spinner
character
is
a
cloud
driver
to
work
with
kubernetes,
so
we
actually
studied
a
problem
a
bit
and
I
would
kind
of
go
through.
Why
designed
not
to
use
Spinnaker's
spinnaker
the
good
thing?
A
Storing
infrastructure
data
in
database
like
I
mentioned
earlier,
is
a
good
and
bad
thing
as
a
separation
in
HTF
and
but
also
it
doesn't
allow
you
to
do
them
to
do
infrastructure
as
code
good
thing
support
immutable
image
right,
cloud-based
deployment,
you
want
immutable
image,
so
that
allows
for
easy
roll
back
and
you
see
we
deploy
to
any
environment,
but
on
the
flip
side
is
because
of
you
guys
built
is
very
ASG
easy
to
centric.
It
doesn't
support
native
Cooper
native
resources,
meaning
like
in
Cooper
nari.
A
You
have
something
called
ingress
servers,
deployment
parts,
it
doesn't
represent
those
concept
well
right
and
the
point
I
keep
coming
back
and
but
again
again
is
about
it
doesn't
embracing
projections
code
and
with
ami
the
bakery
process
takes
a
while,
it's
probably
not
playing
now,
but
sometime.
You
can
take
up
to
an
hour
to
complete
the
bakery
process
and
every
time
you
deploy
you
need
to
do
it.
You
see
to
startup
right.
That's
why
the
deployment
is
taking
about
20
minutes.
A
The
first
thing
we
found
Scott's
case
tonic
that
helps
with
generating
8y
ammo
for
specific
environment
basic
a
sonic.
Let
you
do
is
you
can
create
a
base
yamo
for
your
environment,
but
then
you
can
override
different
portion
rate
for
different
environment,
but
that's
only
one
piece
of
the
puzzle
right
we
like
to
embrace
get-ups
for
environment
changes.
So
once
again,
why
we
like?
That
is
because
this
provides
with
traceability
and
we
pull
request.
A
We
can
have
some
level
approval
process
the
track
changes
with
that
said,
then
we
need
some
way
actually
compare
the
difference
between
what
is
in
kubernetes
SCV
versus
enacting
a
deployment
repo.
So
that's
why
we
need
another
tools
that
give
us
visualization
of
the
difference.
So
we
know
wasn't
a
git
repo,
deponent
repos
actually
is
in
kubernetes.
A
Hence
we
also
that's
what
we
built
I'll
go
CD
in
addition
to
the
command
line.
We
have
a
UI
to
that
helped
you
do
roll
back
the
deployment
and
synchronization.
Now
next
program
is
gonna,
actually
walk.
You
through
I'll,
go
see
the
high
level
and
give
your
full
time
or
how
everything
ties
together.
Awesome.
B
Thanks
believed
for
stating
the
problem,
we
have
the
solution
right
in-house
argo
CD.
So
what
this
does?
It
gives
like
environment,
specifically
a
manova
rights,
because
it's
all
and
gate
you
can
mention
for
pre-burn
women.
You
want
to
run
two
pots
for
Prada.
You
wanna
run
three
pots.
You
can
have
yama
specific
over
eyes
based
on
your
environment.
So
it's
pretty
cool
that
ways.
It
has
an
excellent
UI
to
monitor
of
fleet
of
services
in
spinnaker.
B
Argo
CD
adoption
it's
well
received
in
the
community,
mainly
used
for
CI
CD,
and
a
lot
of
machine
learning
data
for
workflows.
A
lot
of
companies
have
started
using
it
as
well.
We
have
Adobe
a
Google
Blackrock
to
name
a
few,
so
yeah.
Let
me
write
like
jump
into
the
demo
whoo
you
suffered,
so
I
wanted
to
start
with
spinnaker.
So
this
is.
This.
Is
one
service
that
I
host
I'm
an
application
developer,
not
a
nob
sky?
B
So
when
this
came
to
me,
it
was
like
here
you
have
to
handle
this
and
make
sure
everything
is
fine
with
the
ec2
instances.
If
there's
an
issue,
you'll
have
to
debug
it.
So
the
first
problem
we
saw
is
every
time
you
check
in
code
into
master.
It
starts.
It
does
a
CI
in
a
separate
place.
If
we
use
classic
Jenkins,
it
pulls
in
all
the
chef
scripts.
It
takes
roughly
15
to
17
minutes
to
create
an
RPM.
It
gets
uploaded
into
s3,
and
then
it
gets
pulled
in
here.
B
B
One
cool
thing
I
really
liked
about
spinnaker
was
it
gave
how
many
ec2
instances
we
are
running
for
each
environment.
It
gave
a
lot
of
like
if
it's
healthy.
It
shows
green
if
it's
not
healthy
chose
red,
but
you
really
don't
know
what's
happening
inside
it.
You
always
had
to
SSH
inside
the
box.
Look
at
the
configurations.
If
there's
nginx,
you
have
to
go
to
nginx
conf
to
see
what
actually
is
in
each
of
those.
B
A
B
This
is
a
similar.
This
is
the
Argo
city
UI.
This
is
that
this
is
my
Dewey's
environment.
This
is
the
application,
and
this
is
a
parts.
Pods
are
similar
to
if
you're
in
the
spinnaker
world
ec2
instance,
where
the
application
is
actually
running.
So
currently,
in
my
III
service,
we
are
running
two
parts,
so
if
I
click
on
this
part,
it
actually
tells
me
every
single
thing
in
the
manifest
file,
and
this
is
from
guitars.
B
This
is
from
grid
everything's
written
in
grid,
and
we
have
parameter
overrides
that
gets
overridden,
and
if
there
is
an
issue
and
there's
not
a
sync,
you
will
clearly
see
like
something
in
red
over
here
and
you
can
easily
debug
it.
So
you
could
see
every
single
configuration
there's
like
Splunk
forwarding.
What
is
our
index
and
stuff
like
that?
Oh
pretty
cool.
Other
cool
thing
is
the
Springwood
application
actually
puts
in
logs
here.
For
some
reason,
it's
not
showing
up
one.
Second.
B
Sorry
sometimes
kicks
me
out,
but
hopefully
now
it's
yeah
here
you
go.
So
all
the
logs
are
here
as
soon
all
the
traffic
coming
here.
So
we
just
write
our
application
into
standard
out
log.
It
gets
just
pushed
right
here
and
for
monitoring
purposes.
We
push
this
to
splunk,
but
you
can
clearly
see
that
you.
If
there
is
an
issue,
you
can
see
it
right
here
we
have
some
sidecars
running
where
we
wanted
to
move
it
into
wave
front,
so
you
can
see
logs
for
those
as
well.
B
So
every
single
container,
that's
running
inside
your
pod.
You
can
see
the
logs
for
them
and
say
there's
something
wrong
with
this
board
and
I
figured
out
okay,
this
is
not
working.
Fine
I
just
go
here,
delete
it
a
new
one
will
cause
pin
up
instantly
once
it
figures
out
that
I
need
to
delete
it.
So
this
is
everything
that
spinnaker
also
did.
But
now,
let's
look
at
some
more
stuff
that
we
can
see
here.
You.
B
Also
shows
replica
sets,
so
if
I
click
on
any
one
of
the
replica
set
again
the
same
thing,
the
concept
is
same.
It
shows
the
manifest
of
the
replica
said
how
many
posit
needs
to
run
and
all
of
those
things
I
add
something
more
here.
You
can
see
the
deployment
that
is
trying
in
you
can
see
the
service.
You
can
see
endpoints,
so
you
can
basically
have
an
overview
of
everything
that
I
need
for
this
environment.
B
It
also
runs
a
conflict
map,
so
we
are
running
this
Telegraph
conflict
map
where
it
actually
even
tells
us
like
what
it
does
and
what
the
name
of
it
is
and
then
what
events
are
fired
and
if
you
look
over
here,
it
says
status
is
synced.
That
means
it's
synced
with
github
everything.
That's
in
github
is
already
saying.
If
there
is
it's
not
saying
that
will
show
weird
red
button,
they're
saying
that
it's
not
synced
and
maybe
unhealthy
when
it's
unhealthy
and
the
last
thing
that
I
really
love
is.
B
It
also
shows
all
the
ingress
stuff
that
we
need
for
the
service.
It
tells
you
what
the
HTTP
host
the
health
check
and
stuff
like
that
now,
as
an
application
developer,
I
released
my
code
in
a
couple
of
hours
or
even
in
minutes,
I
figured
out
something's
going
wrong.
I
can
easily
roll
back.
So
what
it
does
is
it
provides
history
for
the
last
five
versions
that
I've
checked
it
in
so
say,
for
example,
I
I
deployed
this
code
with
this
kit
revision
and
something
went
wrong.
B
I
can
instantly
go
back
to
the
previous
one
click
your
hit
rollback.
It
will
go
back
to
the
previous
version,
which
is
stable,
so
you
can
have
like
a
fast
release
and
a
rollback.
Now
you
would
think
that
we
are
talking
about
guitar
guitar,
and
this
rollback
is
not.
You
know
in
sync
with
the
github,
but
what
what
this
does
is
like.
It
will
immediately
row
back
then
I
go
in
and
update
the
same,
get
version.
B
It's
really
smart.
In
terms
of
saying
that,
okay,
this
image
is
already
running
I'm,
just
going
to
sync
it
and
not
redeploy
or
do
a
rolling
update
again.
So
in
that
ways
it's
pretty
smart
yep.
So
this
is
just
the
production
version
that
we
have
exact
same
thing,
so
every
specific,
every
single
environment
is
one
application,
and
you,
if
you
have
an
issue
in
one
any
one
of
those,
you
can
see
it
right
there.
So
now
this
is
the
CDP
surface.
We're
not.
B
We
haven't
spoken
about
the
CI,
like
when
I
check
in
my
code
in
to
master
what
happens
so
we
use
something
called
as
bluish
in
Jenkins
for
our
CI
C
D
pipeline.
This
is
like
the
end-to-end
solution,
so
what
happens
is
as
soon
as
I
check
my
code
in
in
to
master
it
comes
here,
runs
a
bunch
of
unit
tests
that
it
needs
to
and
creates
a
docker
image
in
the
package.
After
that
it
goes
into
the
dev
environment
and
run
some
regression
tests
against
it.
B
Then
it
goes
deploys
it
into
e
to
e
or
on
sanity
test.
Now
you
don't
see
the
production
environment
here,
because
we
are
still
conservative.
We
don't
want
every
single
check
in
that
we
make
to
go
into
production,
so
we
created
a
separate
pipeline
just
to
make
sure
that
when
we
check
that
in
it
goes
to
production,
but
our
target
state
is
to
have
a
broad
box
here
too,
so
that
we
can
release
fast.
If
there's
an
issue,
we
can
roll
back
first,
so
yeah.
B
This
gave
like
an
end-to-end
the
overview
of
like
from
the
time
to
check
in
to
production.
Another
cool
thing
that
this
is
internal.
This
is
not
open
source,
but
this
is
something
for
cluster
management
and
monitoring.
So
what
this
does
is
we
see
a
bunch
of
clusters
over
here?
Now
you
see
this
one's
unhealthy
so
for
the
cluster
admin
it
needs
to
actually
know
what
happened
so
for
that
instance,
he
would
just
click
right
on
sorry,
yeah,
that's
a
little
slow
it.
B
B
You
know
flaky
and
figure
out,
what's
happening
over
there,
so
it's
a
cool
tool
for
management
and
the
other
thing
is
I'm:
an
application
developer
for
validation
to
service,
so
I
don't
need
to
go
into
a
name
space
for
someone
else
or
account
service
or
whatever,
and
do
cube
CTO
get
secrets.
You
don't
want
to
do
that.
So
how
do
you
manage
that?
So
this
gives
us
right
access,
so
I
have
access
only
to
my
namespace
and
I
can
do
get
secrets
only
there.
So
that
is
something
that
we
saw
so
yeah.
B
That's
the
demo
I
had
to
talk
a
little
more
about
blueish
fringe.
It's.
So
how
did
that
pipeline
get
created?
It's
just
a
genkan
file.
It
remains
in
your
code.
It
remains
in
your
git
repo
itself
and
it's
customizable
so
say.
For
example,
validation
service
only
needed
three
people
of
pre
prod
environments.
Some
services
might
need
five
six,
so,
depending
on
that,
you
can
customize
it
around
regressions.
So
it's
not
like
you
have
to
like.
B
You
know
do
this
set
of
things
for
every
single
application,
so
that
saves
us
money
too,
because
some
of
our
applications
don't
even
need
so
many
pre,
prod
environments
gives
a
complete
CI
CD
view,
and
when
we
were
doing
this,
we
were
always
already
live
in
spinnaker.
We
wanted
to
make
sure
that
we
are
not.
You
know
making
sure
that,
like
we're,
making
sure
that
it's
highly
available
and
customers
are
not
getting
impacted,
so
what
we
did
is
that
we
enabled
parallel
deployment
where
it
was
running
with
classic
Jenkins
pipeline
with
spinnaker
in
u.s.
B
West
and
together
we
created
a
separate
pipeline
with
blue
ocean
Jenkins
with
Argus,
CD
and
East.
That
way,
we
created
an
HR
hid
our
side
and
flipped
traffic.
Once
we
were
comfortable
eventually,
we
will
get
rid
of
the
spinnaker
pipeline,
but
this
is
something
that,
when
you're
migrating
you
can
do
and
make
sure
that
your
application
is
always
available
here.
A
Is
basically
with
our
existing
spinnaker
deployment,
we
uses
the
classic
Jenkins
the
thingis
1.0,
which
uses
different
artifact
for
bill.
When
we
move
to
the
new
pipeline
with
with
kubernetes,
we
are
actually
using
pollution,
taking
a
two
point
which
only
look
at
the
chickens
file
right
so
essentially
think
about.
Basically,
we
have
a
same
repo,
but
we
can
actually
kick
off
two
different
pipeline,
because
each
bill
system
uses
different
artifacts
for
their
bill.
That's
pretty
much
how
we
kind
of
cheated
right.
B
So
yeah,
just
going
over
that
spinnaker
were
surrendered.
Northern
SAS
with
kubernetes
the
CI
build
from
14
minutes.
We
went
down
to
four
minutes
for
code
deployment
from
90
minutes
from
like
making
the
image
to
going
through.
All
the
pre
prod
moments
into
production
was
90
minutes.
We
could
achieve
it
in
almost
twenty
minutes.
Release.
Time
was
again
18
minutes
because
it
brings
up
three
instances
to
apply
the
code
regression
so
that
please
took
18
minutes
over
here.
It
was
three
minutes.
B
Trolling
deployment
and
ec2
is
already
up
you're,
just
getting
the
pods
up
and
running.
So
it's
just
the
time
spring.
Boot
application
takes
to
come
up.
The
exact
same
was
rolled
back
because
it
does
the
exact
same
thing.
It
just
brings
the
new
instances.
So
yeah
that's
about
it.
We
achieved
this
and
we
are
trying
to
reduce
the
code
deployment
time
more.
But
that's
about
it.
Thank
you.
If
you
have
any
questions,
yeah
sure.
B
So
the
question
was:
how
do
you
interact
secrets
in
your
CI
CD
pipeline?
So
we
use
cube
CT
I
get
secrets
based
on
namespace,
so
it's
not
pipeline
specific
and
we
have
an
into
a
specific
way
we
put
in
secrets.
So
that's
completely
separate.
So
that's
a
one-time
thing
and
in
Gate
we
actually
point
to
that
in
the
git
repo
in
the
deployment
repo.
We
actually
mentioned
that
hey,
look
at
this
place
and
get
the
secrets.
So
that's
part
of
kit.
It's
up.
B
So
yeah,
so
the
question
was:
how
do
you
trigger
from
the
Jenkins
like
from
Jenkins
2.0,
argo
CD
right
so
jenkins,
file
itself
has
a
combination
with
our
go
CD.
It
understands
we
do
like
our
go
start
like
create
and
then
do
it.
So
it's
like
in
the
get
jenkins
file
itself.
We
call
our
go
with
like
the
username
and
password,
and
so
it
connects
it.
The
kid
there's.
A
A
gate
of
the
personal
trigger
that
jenkins
chopped,
it
executes
right
within
a
jenkins
file.
We
actually
used-
and
I
I
don't
wanna-
be
at
the
source
code.
But
basically
is
we
actually
use
a
container
that
has
AA
go
tools
in
it
essentially,
and
then
we
inject
the
secret
from
the
Jenkins
folder
secrets
right
right.
That's
how
we
kick
it
off.
Yes,
yes,.
A
So
the
question
is:
do
we
have
a
comparison
with
the
spinnaker,
but
the
cloud
driver?
The
answer
is:
no.
We
haven't
gone
that
far
it
just
basically,
we've
reviewed
at
UI
plus,
like
I,
said.
That's.
Do
you
think
we
considered
the
UI
itself
is
kind
of
more
easy
to
native,
because
when
we've
both
through
the
screen,
basically
we'll
be
using
that
screen
for
more
than
a
year,
so
we
kind
of
know
it
doesn't
want
to
represent
all
the
Cuban
Air
Resources,
like
deployment
in
part,
is
like
ec2,
but
there's
also
concept.
A
A
A
A
So
so
the
question
is:
is
that
easy
way
to
plug
it
in
the
answer
to
that
is:
come
to
our
booth
out,
develop
oil
and
I'll
go
CDs
there,
so
talk
to
them.
Actually
we're
more
the
team
to
actually
help
drive
the
use
cases
and
put
it
together,
but
I
actually
developers
actually
at
a
booth
so
come
to
the
booth
and
top-1.
A
Okay,
the
question
is:
what
about
home
indication
so
I
think
I
can't
escape
that
on
when
we
first
look
at
how
the
way
I
thought
about
it,
as
at
least
a
lot
of
us
thought
about
small,
like
a
release
package
release
manage
likes.
For
example,
there's
a
home
chat
for
spinnaker
right
I
can
just
run
that
and
install
six
different
parts
and
run
my
spinnaker.
A
A
A
A
B
A
The
person
you
talk
to
right
there,
Danny
yeah,
he's
the
man
he's
doing
the
psu
right
now
he's
working
for
the
POC.
So
what
can
they
release?
We
have
experiment
with
blue
green,
but
for
the
most
part
we
try
to
stick
with
the
rolling
updates.
Unless
we
have
to
then
we'll
tow
back
the
blue,
green
right.