►
From YouTube: Lightning Talk: Introduction to Kubernetes Event-driven Autoscaler - Muhammad Aushafy Setyawan
Description
Kubernetes Event-driven Autoscaler provides us another option to do autoscale kubernetes object based on external metrics resources such as Apache Kafka, AWS CloudWatch, Pub/Sub, etc
A
A
Hello:
everyone
Introducing,
my
name
is
Muhammad
SAW,
my
road
status
as
an
infrastructure
engineer
at
a
sirclo
and
smart
company,
Today
I
will
discuss
about
the
Keller
event
driven
out
community.
So
the
agenda
is,
we
will
discuss
about
the
time
available
in
kubernete,
then
we
will
find
out
how
kubernetes
autoscaling
it
works.
Then
the
weakness
of
the
autoscaling
community
is
by
default.
What
the
picture
looks
like
and
also
we
will
discuss
alternative
solutions.
What
it
looks
like,
and
the
alternative
solution
is
to
use
a
Gadget.
A
Then
there
is
an
example
of
what
the
use
of
gede
itself
is
like
Okay
we
go
to
Sragen.
First
for
the
booklet
itself
say
kubernetes.
There
are
several
kinds
and
several
juices,
depending
on
what
our
application
is
like
Hi
and
first
there
is
a
pot.
Then
air
deployment
there
is
a
skateboard.
There
is
a
Diamond
set.
There
is
a
job,
and
also
there
is
a
concept
for
a
detailed
explanation.
I
will
explain.
Maybe
you
can
still
read
it
here.
A
Then,
in
the
next
section
we
will
discuss
about
the
way
I'm
schelling
works.
Autoscaling
there
are
three
ways,,
namely
horizontal
or
commonly
called
the
set
out.
Method.
This
method
usually
increases
the
number
of
hosts.
We
want
like
that,
and
if
we
use
a
horizontal
photo
skewer,
then
the
increase
or
decrease
in
the
number
of
pots
will
automatically
be
adjusted.,
handle
by
horizontal
Tote
scanner.
A
Then
there's
vertical
boost
power.
This
way
is
a
way
to
add,
or
you
could
say,
schelling
up
full
christos
use
of
risoles
from
pots
or
containers
like
that,
and
both
horizontal
types
or
Hi.
You
understand
if
he
uses
him
once
based
on
memory
usage
and
also
video
like
that,
then
the
3rd
point
is:
there
are
Skeletal
Muscle
Clusters.
The
autoscale
market
is
usually
identical,
more
towards
POM
Kubernetes
in
clubs
and
classmates.
This
room
works
based
on
the
number
of
Walcotts
in
the
Kubernetes
market.
A
So
when
the
white
note
is
almost
full,
then
this
Cluster
autoscaler
will
increase
the
number
of
workers
not
and
after
that
the
Kubernetes
Cluster
will
reball
what
is
in
it
to
spread
to
some
new
notes
were
combined
like
that,
then
there
is
a
disadvantage
of
using
kubernetes
autoscaling
by
default
feature.
So
this
is
already
a
comment
on
the
issue
from
kubernetes.
The
first
is
bonnets:
autoscaling
He
can
only
use
based
on
memory
and
also
Cut
Indonesian.
Then
the
next
point
is
kubernetes
autoscaling
by
default.
A
This
feature
can't
do
autoscaling
when
there
is
very
drastic
use
of
memory
and
also
CPU
So.
If,
for
example,
the
CPU
and
memory
increase
dramatically,
then
it's
not
will
run
autoscaling
like
that,
then
most
recently,
autoscaling
kubernetes
can
be
a
description
of
a
workstation
where,
when
our
application
has
Sam
how
Lex
memory,
then
this
will
have
an
impact
on
my
friend.
Benefits
autoscaling
will
increase
the
number
of
pots,
but
in
terms
of
performance
it
won't
have
much
impact
and
won't
help
at
all.
Then
exactly
what
happens
is
that
infrastructure-wise?
A
Hi,,
then
an
alternative
solution
is
to
use
a
Seven
driven,
auto
sealer
shirt,
so
event-driven,
auto
event,
driven
auto
sealer
is
more
towards
adding
Thompson
alytic
previously
only
based
on
memory
and
also
see
you
Now
with
Eza.
We
can
add
an
external
matrix
like
from
Mr.
Sharp
Apache
Kafka
buy
a
lot
more
ss&c.
A
Then
One
of
the
reasons
why
we
use
CDs
is
open
source.
The
second
is
web
installers,
so
there
are
lots
of
external
ms
The
matrix
adapter.
That
we
can
use
then
supports
multiple
workloads,
such
as
the
deployment
step
full
set
job
and
also
the
love
of
Christ,
and
the
last
one
is
vendor
agnostic.
So
this
is
very
helpful.
If
friends
use
Kubernetes
from
various
factors
like
that,
then
for
the
video
itself
it
has
components.
The
ABS
texture
is
like
this,.
A
You
have
an
object
so
for
each
of
these
objects
we
can
define
what
object
we
want
to
do
a
scale
scan
and
like
that.
Suppose.
Another
deployment
means
that
later,
if
you
trigger
an
event,
it
will
blacken
even
from
adding
the
number
of
foxes
like
that.
Then
there
is
an
external
external
triggers
sleep
here
more
towards
him
monitoring
from
the
external
Matrix.
Then,
if
an
event
occurs,
then
he
will
clear
the
Predator
potato
peeler
then
horizontally
on
this
skeletal
muscle.
He
will
add
the
amount
of
his
workload
or
the
amount
of
Fox
Sport.
A
Then
one
example
is
the
use
of
the
CD
itself.
This
is
if,
for
example,.
If
we
use
HP
Sharp
Apache
Kafka,,
if
you
don't
get
a
message
received
by
the
topic,,
then
it
can
automatically
go,
even
though
it
reaches
the
pot
bureau
like
that,
then,
if
you
Messenger
tank,
Pa
Dede,
it
will
detect
an
event,
then
represented.
It
will
do
autoscaling
and
add
the
number
of
pots.
If
then,
if
there
are.
A
A
lot
of
messages
received
by
caffeine,
then
this
PD
will
do
autoscaling,
adding
even
more
number
of
posts
like
that.
Then
the
last
one
is
each
object,
an
example
of
implementing
LTE
using
fresh
from
what
Jakarta.
So
we
make
the
name.
This
generation's
keltan
is
the
default
name.
Then
there
is
the
famous
island
How
long
after
that
there
is
the
last
Seriously
this
form
of
Arai.
We
can
define
a
lot
of
readers,
not
just
one,
but
in
this
time
we
only
use
the
word
for
what
it
is.
A
The
metadata
like,
for
example,
the
recipe
Sis
fast
Sis,
ask
How:
was
it
originally
held
or
DNS
Hi?
Then
there
was
the
same
review
group.
There
was
a
topic
in
this
case,
for
example
Ancol
topic
and
also
for
Ensemble.
The
group
was
corrupt,
so
it's
like
that
using
a
general
schedule
or
the
basic
way
more
or
less.
This
CD
can
add
researchers
from
Kubernetes
autoscaling.
That
Andi
is
basically
OK,
more
or
less,
I'm.
Sorry,
if
there
are
many
mistakes.