►
From YouTube: Webinar: Deployment Strategies on Kubernetes
Description
A practical look at the different strategies to deploy an application to Kubernetes. We list the pros and cons of each strategy and define which one to adopt depending on real world examples and use cases.
This webinar will cover:
- What are the different strategies and what can be the impact of change on the consumer
- How to choose the right strategy depending on business requirements
- What strategy Kubernetes offer out of the box and which tool can be used to cover the missing part
- A live demo of each strategy using Kubernetes
The webinar will conclude with a summary diagram designed to simplify the advice and help attendees choosing the right deployment strategy.
Join us for KubeCon + CloudNativeCon in Barcelona May 20 - 23, Shanghai June 24 - 26, and San Diego November 18 - 21! Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.
A
B
B
So
recently,
I've
been
working
while
I've
been
working
on
many
different
type
of
infrastructure.
On
recently
I've
been
focusing
on
communities.
So
today
we
are
going
to
talk
about
different
strategy
on
communities
on
here.
It
is
the
agenda,
so
a
quick
overview
of
what
commit
humanities
is
then
I'm
going
to
list
six
different
strategies
for
each
strategies,
I'm
going
to
to
show
you?
Oh,
oh,
it's
working!
On
give
a
little
example,
these
two
pull
on
cons
of
each
strategy
and
then
do
not
go
for
a
summit.
B
B
So,
as
you
can
see
on
the
screen
on
the
right
part
of
the
screen,
final
pearl,
you
are
two
different
cap
kind
of
cryptic
asset.
The
one
in
blue
is
the
version
one
for
example,
and
then
you
have
the
one
in
green,
which
is
going
to
be
version
two,
and
when
you
deploy
an
application,
the
pods
are
going
to
be
created
in
a
different
way,
and
this
is
what
we
call
deployment
strategy.
So
it
can
be
like
recreate
one
after
the
other
one
exit
routes,
I'm
going
to
list
them
later
on.
B
B
The
deployments,
then
you
have
the
replicas
set,
so
replica
set
is
managed
by
the
deployment
and
then
you
are
the
border
in
the
board.
You
have
the
containers.
So
just
a
quick
reminder.
A
replica
set
is
some
kind
of
controller
that
are
you
that
that
is
going
to
control.
How
many
pods
are
going
to
run
and
trying
to
make
sure
that
there
is
at
least
the
minimum
amount
of
pod
that
you
have
on
the
screen
where
it
said
repeat
three,
so
it
would
take
care
of
keeping
it
up
on.
B
If
one
fail,
it
will
restore
the
pod
Cuban.
It
is
everything
works
with
labels,
so,
as
you
can
see,
for
example,
replica
set
is
matching
pods
that
have
the
label
app
engine
aids.
The
services
on
the
middle
of
the
screen
also
target
ordered
put
that
label
up
equal
nginx
on
the
same
thing
for
another
ingress,
but
at
the
IRL.
So
you
can
have
your
service
time.
So
if,
for
example,
someone
access
a
domain,
the
applications
with
domain
a
food
at
Ballard
calm
then
want
to
access
/foo.
B
So
now
that
we
have
a
little
bit
of
more
understanding
about
all
configuration
work,
which
is
basic
for
deployment
strategy,
I'm
going
to
present
each
of
these
strategies,
there's
six
of
them
that
have
been
listing
two
of
them
the
recreate
and
run
a
two
strategy
that
are
native
to
prepare.
It
is
so
you
don't
need
any
step.
Any
extra
step
or
extra
Tulear
to
like
is
202
more
features
around
it.
The
blue-green
deployment
on
the
Canary.
B
B
So
the
first
one
is
recreate,
so
they
recreate
deployment
a
user
is
going
to
access
the
service
in
your
cluster.
The
the
request
is
going
to
one
pull
of
instances
or
in
this
case
V
one
direction.
One
of
your
application.
Then
we
are
going
to
shut
down
the
instance.
The
instance
start
it
again
on
when
it's
ready.
We
are
going
to
transfer
the
traffic
to
the
vision.
So
here
it's
just
a
very
dumb
way
of
doing
a
deployment,
so
you
shut
down,
shut,
shut
down,
start
the
instance
again
on
with
an
aviation
on
area.
B
B
B
You
can
define
it
this
way.
So
if
you
want
to
choose
to
recreate
deployment,
then
you
go
to
write
type
recreate
and
then
to
apply
the
new
version
of
your
deployment.
You're
gonna.
Do
a
chip
CTL
apply
under
new
manifest,
so
of
course
I'm
not
I'm,
not
mentioning
it
yeah,
but
I
assume
that
everybody
knows
that
when
you
have
a
deployment,
a
first
application,
it's
I
inside
your
deployment,
you
define
the
version
of
your
container.
B
So
if
you
want
to
deploy
a
secondary
version
of
your
application,
the
next
version,
then
you
define
a
new
tag
and
the
next
time
you
are
doing
my
humanities
is
going
to
make
sure
to
recreate
all
the
birds
on
replica
set.
Well,
the
repeater
said
that
is
going
to
create
the
new
birds
one
after
the
other
come
with.
B
So
if
we
look
at
the
pattern
of
the
traffic
during
the
release,
you
can
see
here
that
mission
one
represented
in
green,
get
some
traffic
and
then
because
we
shut
down
the
machines,
there's
no
traffic
anymore
for
a
little
amount
of
time,
so
the
service
is
unavailable
until
a
fashion
to
is
started
again
on
is
ready
to
access
exit
except
traffic
and
that's
what
we
get.
So
this
is
this
panel,
so
the
problem
with
this
deployment.
B
Well,
let's
start
with
the
poll,
it's
very
easy
to
set
up.
As
you
can
see,
there
is
only
one
field
to
change
in
your
deployment
on
said
type.
We
create
the
counts
of
this.
One
is
that
there
is
a
very
high
impact
on
the
user
on.
Of
course,
we
have
done
time
on
the
downtime
depends
on
both
the
shutdown
on
put
Eurasian
on
the
application.
So
if
your
vision
is
very
fast
goods
and
it's
gonna
be
very
short-
the
downtime
it's
a
bit.
B
B
If
so
or
that
works
you
have
a
user
accessing
version
1
through
a
service,
then
we
start
spinning
up
one
instance
of
the
new
version
way
that
it
is
ready
and
then
we
can
some
traffic
and
then
we
kill
the
previous
instance
operation,
one
on
the
other
two
lovebirds
start
a
new
one
of
the
fashion
to
sound
traffic
when
it's
ready,
remove
another
one
from
v1
etc.
So
this
is
so
you
would
deploy
your
application
one
after
the
other
one,
one
point
after
the
other
one.
B
So
it's
like
you're
a
slow
one
out
of
your
application
on
the
way
you
would
configure
it
it's
a
little
bit
different
than
the
other
one.
So
you
change
the
type
rolling
update,
but
you
have
some
extra
parameters
that
allow
you
to
define.
How
many
put
you
can
add
an
extra
to
the
amount
of
odd
that
you
have
already
in
your
application
and
what
is
the
maximum
maximum
amount
of
point
that
can
be
unavailable
so
most
of
the
time,
you
will
do
something
like
this
maximum
available
zero.
B
So
we
want
to
keep
the
amount
of
POD
minimum
and
we
only
want
to
have.
We
want
to
have
to
put
in
maximum
every
time
to
dress
during
the
running
update
for
these
two
values
so
yeah,
it's
it's
integrals
right,
it's
2
or
0.
You
can
also
define
a
percentage.
So
if
you
have
auto
scaling
in
place,
then
it
can
be
quite
ND
the
run
deployment,
so
the
warning
update
is
the
one
that
is
mostly
unwisely
used
everywhere,
very
simple
tool
to
set
up.
B
B
B
For
these
deployments
it
is
also
very
easy
to
use
you
just
have
to
set
the
field
type
equal
running
update.
The
good
thing
is
that
is
doing.
A
variation
is
slowly
release
across
instances,
and
then
it's
also
coming
and
hostage
for
application
like
another
ongoing
rearranging
of
the
data,
so
I
put
it
there,
but
what
I
mean
by
this
is
when
you
have,
for
example,
some
kind
of
state
in
your
application.
B
While
you,
your
other
application,
is
while
the
new
release
is
spinning
up
on
the
side,
you
want
to
have
a
little
bit
of
time
to
synchronize
synchronize,
all
the
data,
and
then
you
can
tell
me
that
the
previous
one
right
problem
with
this
one
is
that
the
whole
out
can
take
some
time
on.
It
depends
on
the
amount
of
product
amount
of
replicas
on
boards
that
you
are
going
to
have
for
this
application
that
you
want
to
update
all.
There
is
absolutely
no
traffic
of
a
control
over
the
traffic.
B
B
B
So
when
you
do
this
on
the
side,
you
are
going
to
apply
your
changes,
so,
as
you
can
see
in
in
the
Box
highlighted
box
on
the
bottom
right
corner
of
the
screen,
I'm
doing
a
cube,
CTL
apply
f
off
the
manifest
version,
2,
so
I'm
raising
to
your
application
on
the
side
of
government
one.
So
what
I'm
azeez
is
that
it's
going
to
use
a
lot
of
resources,
because
you
need
double
capacity
of
your
infrastructure
and
then
you
do
when
everything
is
ready.
When
your
application
is
ready
to
accept
traffic,
you
make
a
switch.
B
So
you
do
a
cutie
tail
patch
on
the
service
on
you
say
a
I
want
version,
2
magic
action
instead
of
mission
1.
Then,
if
you
think
that
everything
is
fine,
you
can
safely
do
that.
The
previous
deployment,
or
you
can
keep
it
up
and
running
over
there
and
undo
a
whole
back
in
case
a
some
issue
detected.
B
The
blue-green
deployment
is
mostly
used,
while
one
of
the
reason
that
we'll
use
this
one
is
called
fronting
application,
so
fronting
application
is
you
are
going
to
save
some
HTML
in
the
HTML
and
say
some
assets,
so
CSS
or
JavaScript
the
time
you
are
going
to
make
a
request
to
your
application,
you're
going
to
have
while
doing
an
update
right.
If
you
make
a
request
to
an
application,
the
time
that
you
you
receive,
the
content
on
your
browser
pass
your
HTML
on
after
to
load
the
assets
on
the
server.
B
If,
at
the
meantime,
you
have
a
change
in
a
version,
then
it's
not
going
to
work
correctly
right.
So
in
such
scenario,
you
would
use
a
blue-green
deployment
for
this
if
you
want
to
do
it
with
multiple
services,
so
you
can
do
it,
of
course.
So,
in
this,
in
this
screen,
I'm
showing
like
you,
have
an
application
called
Logan
and
on
the
side
an
application
called
cat.
B
In
this
scenario,
I'm
appending
the
vision
in
the
name
on
what
we
do.
We
use
ingress
on
in
the
ingress.
You
just
apply
the
new
version
there.
So
the
way
you
will
do
it,
you
apply.
You
spin
up
your
application
on
the
side
like
you
would
do
with
a
single
service
right.
So
all
your
services
with
the
invasion
on
then,
when
everything
is
ready,
you
apply
the
change
at
the
ingress
revenge.
So
you,
your
map
or
the
service
name,
looking
v2
dot,
v2
to
me
to
match
the
new
version.
B
B
So
the
good
thing
with
the
blue
team
deployment
is
the
whole
out
is
instant
on
the
whole
back
as
well.
So
it's
very
quick
if
you
are
using
ingress
with
multiple
services.
All
the
services
are
going
to
be
updated
at
once,
so,
as
I
was
mentioning
before
it's
very
good,
fit
for
and
front-end
applications
that
blurred
vision
assets
from
the
same
server.
B
It's
also
not
a
way
to
fix
application
dependency
air.
So
if
you
have
a
few
applications
running
in
one
cluster,
each
application
depends
on
each
other.
This
is
this
is
a
way
that
you
could
fix
the
deployment.
So
instead
of
releasing
one
applications
and
another
one,
and
then
you
don't
really
know
how
to
do
it,
it
could
be
aware,
but
it's
not
suggested
right,
then
cons
of
this
one.
It's
quite
expensive,
because
you
require
double
the
resources
and
then
you
need
to
have
copper
testing
in
place.
B
So
you
need
to
test
your
platform
with
unit
testing
integration,
testing
etc,
and
this
should
be
done
before
raising
to
prediction,
but
you
can
still
roll
back
of
course,
but
there
is
some
impact
on
the
consumer
on
you.
You
want
to
be
able
to
have
something
that
is
working
out
of
the
box
yeah
next
one
line
of
the
shadow
Department,
which
is
the
latest
one
in
this
list.
It's
also
called
mural
or
dark,
that's
a
funny
name,
but
what
does
that
mean?
So
you
are
going
to
have
user
accessing
information
one
on
the
application.
B
Then
you
are
going
to
release
version
2
of
your
application
on
the
side
like
this,
you
wait
at
observer,
see
everything
is
ready
and
then
you
can
sound
traffic
to
it.
Oh
sorry,
let
me
stay
here.
It's
not
the
shadow
implement.
This
is
a
canary
Department,
okay!
Well,
let's
go
with
a
canary
explanation.
Yeah
so
can
I
raise
to
be
different.
You
can
re
is
I'm,
not
change
that,
so
we
have
a
user
accessing
a
savage
balancer
on
what
it
does.
Is
you
speed
up
your
application
on
the
side?
B
Then
you
can
define
okay
I
want
90%
of
the
traffic
to
go
to
version
1
with
a
figure
information
too.
So
this
is
a
very
safe
way
to
deploy
an
application.
That
means
you.
Do
you
test
your
application?
Well,
you
could
test
your
application
with
live
traffic
on
look
for
failure
and
then
the
good
thing
with
is
that
you
are
doing
it
on
a
very
little
subset
of
users.
B
So
if
there
is
a
failure,
you
can
quickly
rollback
on
it's
only
a
little
a
little
bit
didn't
part
of
the
user
that
impacted
with
a
failure
and
then,
when
everything
is
fine,
you
can
roll
out
everything
with
under
percent
of
the
traffic.
So
the
problem
is
a
canary
department.
Is
that
you're
gonna
need
to
extra
tooling?
If
you
want
to
have
it
precisely
define
and
lock
so
I
would
explain
it
in
a
bit.
B
B
Not
one
on
scalar
version
whenever
you
think
it
is
right,
so
in
this
way,
so
that's
what
I
was
explaining
a
little
bit
before
part
of
the
traffic
is
going
to
own
in
version
2
on
most
of
the
parts
going
to
underpin
one.
So
the
way
that
I'm
describing
here
that
you
are
seeing
it's
the
poor
man's
way
of
doing
a
canary
deployment,
because
the
spread
of
the
traffic
is
going
to
be
defined
by
the
amount
of
pot
that
you
have
in
your
application.
B
If
you
want
something
a
bit
more
precise
in
terms
of
traffic
distribution,
you
could
use
East,
you
sto,
which
is
a
service
mesh,
and
that
is
very
popular.
These
days,
still
not
polish,
already
I
think
but
I'm
not
sure
what
I
never
used
it
in
position.
Maybe
some
other
people
can
tell
you
more
about
this.
So
the
way
you
would
this
describe
it
is,
you
have
Reus
was
called
root
wool
and
then
you
define
how
much
weight
you
want
on
specification
on
the
weight
on
the
second
irrigation.
B
B
B
B
You
just
reduce
the
application
to
a
subset
of
users,
so
in
this
case
it
is
10%.
You
leave
the
way
from
there
for
some
time
on.
If
you
think
that
there
is
no
failure
in
detected
or
if
you
think
that
this
is
good
to
go
into
or
after
everything
then
later
on,
you,
you
add
100
percent
of
traffic
on
version
2
and
you
can
safely
delete.
B
B
B
The
problem
is,
is,
is
it
can
be
a
bit
slow
because
you
need
to
wait
some
time
right
if
you
want
to
detect
failure
on
if
you
want
to
test,
if
the
new
version
of
the
new
release
has
been
successful
in
a
way,
if
yeah
you
might
require
to
have
sticky
sessions.
So
why
sticky
session?
So
it's
something
that's
yeah.
A
B
Not
forgiving
as
well
none,
but
if,
if
someone
a
user
access,
your
application,
one
version
or
the
other
one,
you
want
to
keep
your
user
on
this
specific
version,
so
you
don't
want
to
other
user
going.
Oh
I,
one
of
the
requests
went
on
mission,
one
of
the
application
on
the
second
request
that
the
user
asked
it's
going
to
version
two.
Sometimes
you
can
add
some
mismatch.
So
to
fix
this,
you
would
do
sticky
sessions
but
yeah.
This
is
something
that
is
not
suggested
to
do.
B
B
So
a/b
testing
is
it's
mostly
used
by
by
the
marketing.
So
if
you
want
to
release
a
new
version,
you
want
to
test
the
convention
if
there
is
an
improvement
in
the
combination.
But
if
you
are
running
a
store,
you
want
to
test
on
a
specific
set
of
user,
that
the
new
version
is
working
better
than
another
one
and
if
it
is-
and
you
can
reduce
the
nucleation
tail,
but
if
it's
not-
and
you
don't
want
to
impact
many
people
were
so
you
put
it
on
sign,
so
others
at
work.
B
So
you
have
a
user
ear
represented
with
a
desktop
or
laptop
application.
That
goes
to
mission
one
and
then
on
the
other
side,
you
want
to
match
another
type
of
user.
The
person
that
is
using
a
mobile,
for
example,
you
want
to
match
them
to
access
version
2
of
the
application.
So
there
is
a
few
thing
that
you
can
filter
on
your
users,
so
you
can
use
geolocalization
language,
cookie
user
agent.
B
So
all
these
parameters
that
come
from
the
request
on
the
header,
then
you
can
figure
out
some
people,
bring
them
and
drag
them
on
the
mission
that
you
want
them
to
access.
So,
in
this
scenario
here
that
I'm
presenting
a
user
that
are
using
a
mobile
phone
are
going
to
access
version,
2
or
the
version.
The
mobile
version
of
your
website
on
the
other
user
is
going
to
access
the
desktop
version
of
your
website.
B
B
B
B
A
B
We
see
if
you
have
complete
control
over
the
traffic
distribution,
so
you
can
vary
like
yeah.
You
can
do
almost
everything
you
would
say
on.
It
is
a
great
tool
that
can
be
used
as
with
dice
business
to
improve
commercial.
So
if
you
are
running
a
store
website,
then
this
is
this:
is
this
is
pretty
nice
to
use?
The
problem
you
see
is
that
it
requires
an
initial
load.
Balancer,
like
is
200
link
Adi,
so
by
using
in
traditional
balancer,
you
need
to
take
care
of
putting
it
in
place,
updating
it,
etc.
B
So
it's
a
little
bit
more
work,
but
it
can
be
beneficial
depending
on
your
requirement.
One
of
the
thing
that
I
noticed
is
when
you
use
such
a
tool.
It
can
be
very
hard
to
troubleshoot.
Why?
Because
you
have
some
requests
coming
to
your
crystal
on
sometime,
it's
going
to
fail.
You
don't
really
know
why,
so
you
have
to
to
bring
some
extra
tooling
around
this
on
strings,
need
to
be
able
to
tell
to
help
you
to
troubleshoot,
so
something
that
you
could
use
is
tracing.
B
B
On
the
next
one,
the
shadow,
so
previous
light
goes
wrong
and
the
shadow
diplomat
is
the
last
one
that
I'm
going
to
present
here.
It's
also
called
a
mirror
on
dark
on
the
way
it
works,
so
you
spin
up
your
application
mission
one
and
then
on
the
side.
You
are
going
to
recession,
but
you
are
going
to
some
traffic
to
it
without
impacting
the
user.
So
what
it
does
it's.
B
We
split
the
traffic
from
this
user
on
use
on
the
same
traffic
on
mission,
one
edition,
but
only
mission,
one
is
received
by
you
by
your
user
on.
By
doing
so,
you
can
test
on
scale
your
application
on
the
on
see
if
it,
if
it's
performant,
without
impacting
anybody
right,
it's
very
nice.
The
problem
with
this
is
that
you
need
the
better
size
of
your
infrastructure,
because
you
are
going
to
have
two
versions
running
together
and
you're.
B
Gonna
need
to
have
some
looking
system,
because,
if
you
application
rely
on
dependencies
like
database,
you
don't
want
to
have
any
insert
update,
delete
on
your
database
from
the
original
running
measure.
Most
wise,
you
can
double
the
data,
so
you
need
mocking
not
only
when
everything
is
fine,
you
out,
we
should
do.
B
How
would
you
do
this
so
again?
Yeah
people
will
say
that
I'm
an
East,
you
fanboy,
which
I
am
maybe
but
this
is
this-
is
a
very
easy
way.
So
they
have
a
very
nice
way
to
describe
all
you
want
to
to
shake
a
request.
When
you
want
very
much
requests
to
go
on.
If
you
do
shadow
deployment,
you
would
use
feed
the
gold
of
mijo
on
its
going
to
to
speed
your
traffic
from
one
measurement
to
hundreds
of
one.
B
The
only
thing
that
you
have
to
be
careful
is
you
need
to
add
labels
so
mission.
One
is
going
to
receive
one
of
the
percent
of
the
traffic
combination
to
only
zero,
so
all
the
things
that
I've
been
describing
in
this
slice
available
on
github,
so
you
can
come
check
it
out.
You
have
all
this
example
home.
So
that's
why
I'm
not
expending
too
indepe
and
if
you
look
at
the
pattern,
so
it
would
look
like
this,
so
you
have
the
traffic
going
in
which
is
going
to
vision.
B
B
You
do
a
powerful
testing
of
your
application
with
pollution
traffic,
which
is
very
nice.
There
is
no
impact
on
the
user
on
you
decide
whenever
you
want
to
run
out
and
then,
if
you
think
that
the
performance
of
the
applications
he's
good
enhance,
and
you
can
run
it
out
or
not
so
to
you
in
a
way
you
will
use
a
shadow
Department
if
performance
is
very
important
and
you
don't
want
to
from
one
Vidya
stone
as
one
to
lose
after
months.
B
So
so,
if
you
have
an
increase
in
latency
or
something
exists,
then
you
you
just
like
say:
no,
we
don't
want
to
reduce
application.
Then
someone
can
walk.
Insist
to
improve
promise.
Is
that
it's
quite
complex
to
set
up
as
I
was
mentioning
before
it's
very
expensive,
since
you
need
double
capacity,
it's
not
a
true
user
test
and
can
be
misleading
in
some
case,
because
because
you
need
mocking
right
so
because
you
need
mocking
is
the
data
that
is
going
to
be
received
is
going
to
be
instantly.
B
Maybe,
except
if
you
had
some
latency
random
methods
in
your
request,
so
yeah,
it
depends
of
course
of
your
requirement,
but
this
one
I
so
in
my
case
I
never
used
it
in
prediction,
but
yeah
I
think
it's
too
complex
in
some
ways.
So
it
really
depends.
If
you
have
a
lot
of
time,
then
go
for
this
one,
because
this
is
yeah.
This
is
live
traffic
testing,
but
on
the
side.
So
this
is
quite
amazing.
B
To
sum
up,
if
you
use
a
recreate
well,
you
will
use
a
Ric
wait.
If
don't
I'm
is
not
a
program,
that's
already
I've
never
been
using
it,
but
yeah
I
would
see
some
scenario
feel
empowered.
If
you
need
to
keep
I,
don't
know,
maybe
a
host
name
to
be
the
same
host
name,
then
you
will
use
this
one,
because
it's
recreating
after
the
other
one
shut
down.
I
start
with
a
previous
body
shut
down,
recreate
and
work
runs.
Deployment
are
both
native
to
communities.
B
Blue
green
or
shadow,
it's
quite
expensive,
so
yeah.
If
you
have
good
Reds,
why
not
write
but
yeah?
My
favorite
tia
is
still
around
appointment
or
cannery.
If,
if
you
are
not
too
sure
so,
yeah
a
canary
on
a
be
testing
is
many
used
when
you
have
not
too
much
confidence
over
the
quality
of
the
release.
So
if,
if
you're,
not
yeah,
if
you
didn't
test
correctly
or
if
you
just
some
people,
just
writing
prediction
right,
so
you
can
do
it!
B
Then
you
would
use
cannery
that,
like
this,
you
you
do
it
at
the
rate
subset
of
users
and
then
cannery.
A
busy
singing
shadow
require
some
extra
components.
In
your
crystal
to
finish,
we
created
at
container
solutions
this
little
diagram,
so
it's
kind
of
a
decision
table
which
you
could
use
if
you
are
not
sure
which
technology
which
diplomat
you
could
adopt
for
your
radius
pipeline.
B
This
one
is
available
on
the
github
repository
as
well.
So
if
you
want
to
use
it
just
go
there
share
it's
it's
it's!
It's
very
useful.
It
can
be
useful
on
next
on
the
line
if
you
want
to
play
around
so
that
was
my
chain
before
we
created
a
repository
with
each
of
this
deployment
strategy
listed
there
on
the
step-by-step
tutorial
on
how
to
use
it.
We
have
been
using
mini
cuba,
so
you
can
do
it
on
your
own
laptop
you
don't
in
the
cluster
running
somewhere,
which
is
quite
convenient
and
yeah.
A
A
A
A
B
B
Was
not
made
for
easing
in
cuba
letís
and
what
they
did.
Is
they
just
adapted
spinnaker
to
make
it
work
on
communities
and
then
travel
strategies
that
they
are
using
there,
but
I
guess
all
the
native
one
on
the
blue,
green
diamond
deployment
should
be
available,
I
guess
but
yeah,
it's
pretty
difficult
for
me
too,
to
talk
about
it
since
I
have
been
I
haven't
been
using
it
I
guess.
If
you're
gonna
spin
up
your
website,
you
should
be
able
to
see
it.
B
B
So
yes,
so
what
I
planted
in
this
presentation
is
you
work
with
a
cute
CTL
come
on
right,
so
keep
CTO
come
on
can
be
used
everywhere
in
your
CI
system
at
any
moment.
So
if
you
have
any
bit
of
scripting
around
it
and
you
can
just
do
it,
you
still
apply
on
your
deployment
on
it's
going
to
be
released
to
your
cluster,
but
still
I
will
suggest
using
maybe
helm
helm,
which
is
a
package
manager
from
Cuba
richy's.
B
It's
doing
exactly
the
same
thing,
but
it's
that
you
will
do
with
communities,
but
it
has
some
very
nice
feature
and
you
could
do
all
the
deployment
that
have
been
these
things
there
with
helm,
but
yeah.
You
still
need
some
scripting
around
it,
so
it's
not
straight
for
one
against,
but
if
you,
if
you
follow
the
steps
that
are
described
in
the
github
repositories-
and
you
will
see
it's
not
that
if
you
can
tell.
A
B
B
B
A
B
B
A
Perfect,
just
from
our
just
from
our
perspective,
I'd
like
to
mention
that
history's
made
sure
enough
and
enough
customers
and
enough
companies
data
currently
using
aid.
At
the
same
time,
the
current
milestone
sto
0.35.
As
far
as
remembers
it's
not
it's
not
one
at
zero
yet
but
target
in
this
milestone
real
soon.
So
please,
stay
tuned
for
the
first
announcement
for
mister
community
can
I
do
the
next
questions,
even
as
an
example
for
sure
the
deployments.
B
So
I'm
not
sure
what
kind
of
example
before
I
guess
shadow
repayment-
it's
mainly
for
performance
reason
that
you
will
use
it
so
in
your
CIA
CD
system
for
in
a
burn.
You
your
reason,
application,
and
you
want
to
make
sure
that
you
are
not
decreasing
the
performance
of
your
application.
Then
you
will
do
a
shadow
repayment
on
monitor
all
the
performance
are
going
on
this
application
of.
A
B
You
have
many
way
to
do
it
I
mentioned
before
that,
when,
depending
on
of
course,
the
database
that
you
are
using
the
best
scenario
that
you
could
get
is
having
some
kind
of
rebalancing
systems.
So
when
you
release
a
new
database
version,
you
need
to
make
sure
that
when
a
new
part
is
spin
up,
so
I
will
use
run
to
deployment
which
is
going
to
do
it
one
after
the
other
one.
B
A
A
B
So
hem
is
its
kind
of
an
abstraction
right.
So
since
it
is
just
that
well,
it's
not
just
to
make
a
jingles
course.
You
can
see.
Is
it's
very
powerful,
but
I
have
seen
something
somewhere
on
guitar
by
guess
that
they
are
going
to
improve
hem
on
make
all
this
well
at
least
the
canary
development
possible,
but
I'm
still
not
sure.
B
Yeah
and
it'll
be
different
right,
so
it's
you
could
do
exactly
the
same
thing
that
I've
been
explaining
on,
with
with
a
hand
command
in
a
way,
but
yes,
I,
guess
it's
very
good
question
I
actually
started
working
a
few
weeks
ago
on
creating
this
example
with
Helmer.
So
since
you're
asking
I,
guess:
I'm
gonna
just
put
it
there
in
the
repository
at
some
point
this
week
or
next
week,
love
this
we'll
be
able
to
play
around.
B
Yes,
we
are
talking
about
him.
I've
been
developing
a
plugin
recently,
which
is
called
helm.
Monetary
Nia
had
monitor
on
what
does
is
during
your
deployment.
You
can
query
some
some
instance
like
vomitus
or
Kiba
now
for
logging
yeah,
that's
extreme
I'm,
based
on
on
the
query.
So
if
there
is
a
detection
of
the
videos
and
it's
going
to
take
care
of
holding
back
automatically
your
hand
release
so
I
suggest
you
to
give
it
a
shot.
A
A
A
B
B
B
B
So
I'm
not
sure
to
understand
the
question
yeah.
So
you,
the
thing
is
ingress
is
part
of
communities
on
helm
is
just
a
wrapper
around
active
humanities
management,
Nexus
Q,
CGI
CLI
in
a
wave,
it's
just
a
superset
of
this
yeah
I-
would
suggest
just
just
use
hem
to
release
all
these
components
that
you
need
in
your
in
your
crystal,
since
it's
quite
powerful
on
your
Quincy
behind.
It
is
quite
active.
B
And
I
have
this
feeling
that
hem
is
becoming
the
the
tool
that
is
used
by
most
of
the
people
to
to
deploy
applications
on
communities,
but
yeah
we
have
to
understand
our
communities
is
walking
and
then
you
can
maybe
move
to
him
even
it
if
helm
is
to
pay
around
community
ISM
and
you
can
define
it
as
default
for
right.
You
have
the
abilities
operator
that
is
going
to
take
care
of
all
the
configuration,
and
then
you
have
the
helm
user.
That
is
just
like
someone.