►
From YouTube: NATS v2.2: The JetStream Release - Waldemar Quevedo Salinas, Synadia Communications, Inc
Description
Don’t miss out! Join us at our upcoming event: KubeCon + CloudNativeCon North America 2021 in Los Angeles, CA from October 12-15. 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.
NATS v2.2: The JetStream Release - Waldemar Quevedo Salinas, Synadia Communications, Inc
NATS v2.2 is the largest feature release since version v2.0. The NATS v2.2 release provides highly scalable, highly performant, secure and easy-to-use next generation streaming in the form of JetStream. It allows remote access via websockets, has simplified NATS account management with an embedded account server, extended monitoring metrics, subject mapping & traffic shaping, built-in headers, which together further enables NATS toward our goal of securely democratizing streams and services for the hyperconnected world we live in.
A
So
from
the
beginning,
nats
is
a
already
a
10
year
old
project,
one
of
the
earliest
cncf
projects.
Since
the
beginning,
some
of
its
rates
was
to
be
performed,
have
be
very
simple,
have
been
very
secure
and
always
be
available,
like
an
always
available.
Dial
tone
has
multiple
qualities
of
service.
Multiple
communication
patterns,
as
I'm
going
to
cover
in
a
bit
and
thanks
to
the
simple
protocol,
has
a
multiple
client
implementations.
A
A
You
know
as
a
client,
you
only
did
the
url
and
the
credentials
there
is
gossiping
and
a
lot
of
things
that
make
not
deployments,
be
very
flexible
and
have
a
very
a
long.
So
I
have
a
straightforward
api
and
from
the
nets
v2
series
it
has
a
security
and
multi-tenant
accounts,
isolations
of
the
subjects
namespaces
and
also
it
supports
multiple
arbitrarily
network
topologies,
which
is
a
very
advanced
powerful
pattern
that
we
call
the
net
adaptive
edge
architecture.
A
So
you
can
have
a
silo
of
nuts
clusters
and
those
be
able
to
be
connected
to
another
remote
cluster,
another
another
data
center
and
expand
into
a
super
cluster
types
of
setups.
And
then
you
can
have
what
we
call
the
leaf
nodes
connections
that
can
be
outside
of
the
super
cluster
or
create
a
home
spoke
topology,
so
that
you
can
have
like
a
centralized,
nats
cluster
and
then,
through
the
leaf
nodes,
they
can
all
communicate
all
the
clients
can
still
communicate.
So
it
can
all
of
these
three
building
blocks.
A
A
Handle
that
message,
request
response
is
sending
a
message
inspecting
a
single
one
response,
publish
you
can
broadcast
into
a
number
of
members
from
a
group
and
two
subscribers
is
the
load
balancer,
where
only
one
of
them
will
receive
it.
This
is
a
very
classic
nets,
since
the
v1
series,
which
is
an
at
most
ones,
delivery.
So
in
terms
of
delivery
guarantees,
that's
with
ensures
that
you,
you
have
to
be
connected
into
nets,
to
able
to
receive
a
message.
A
A
You
you
have
to
avoid
becoming
a
slow
consumer,
so
if
in
case
there
is
no
one
when
and
we
make
a
request
when
it's
in
nets
very
similarly,
you
can
make
a
request
and
then
just
retry,
making
the
requests,
something
that
we
do
with
http
services,
something
that
is
new
in
the
v2.2
is
a
instead
of
what
used
to
be
a
nasty
mouse
when
making
a
request.
Now
you
get
a
very
fast
error,
called
the
no
responders
error
because
kind
of
like
an
analog
of
the
503
http
status.
A
Now
you
can
have
your
client,
your
requesters
in
nets
very
quickly,
detect
whether
there's
interest
in
the
subject
and
then
try
another
logic
in
terms
of
at
least
once
delivering
nets.
What
initial
support
for
for
it
came
in
it's
not
streaming
in
the
next
stream
project,
also
known
as
stan
as
the
being
the
opposite
of
nets.
A
A
It
was
in
technical
preview
previewed
in
december
2019,
the
previous
kubecon
and
the
clustering
was
finalized.
How
the
first
version
of
its
of
its
of
jet
stream
clustering
got
released
in
march
2021
on
lightning
streaming.
It
is,
it
is,
built
internet,
so
it
is
in
a
subsystem
that
is
part
of
nets,
and
you
don't
have
to
use
protocol
buffers
and
also
another
request
response
protocol
to
able
to
persist
the
messages
which
was
one
of
the
pain
points
from
nasa
streaming,
and
also
you
don't
you're.
A
A
A
You
can
also
have
consumers
and
the
consumers
that
they
can
be
either
push
or
pool
based,
and
you
decide
from
where
and
how
you
want
to
con
be
delivered
those
messages
or
in
case
of
pulling
how
what
is
the
reliability
time,
for
example,
for
each
one
of
those
messages
and
unlike
nets,
you
also
you
do
have
different
types
of
acknowledgements.
You
can
have
like
an
act
of
negative
act,
also
terminate
messages
and
I'm
going
to
cover
those
in
a
bit
as
well.
A
But
first
I
want
to
introduce
something
was
fundamental
into
in
the
development
of
jet
stream
that
we
now
recognize
like
a
key
component
for
doing
operations
with
next,
which
is
a
nat
cli,
much
like
in
kubernetes.
You
have
something
like
the
cube
ctl
of
keep
cuddle
that
you,
where
you
can
point
everyone
to.
A
Okay,
so
give
me
the
the
output
from
this
cube
color
command,
and
we
have
adopted
this
nut
cli,
this
developed
by
one
of
the
core
maintainers
from
nasa,
and
this
gives
you
a
lot
of
utilities
to
be
able
to
establish
a
common
tool
set
for
the
next
community
for
doing
operations
on
nets.
So
you
can
do
very
from
the
various
things
basic
like
pop
subbing
to
doing
some
of
the
more
advanced
monitoring
and
jet
stream
management
all
through
the
net
cli.
A
You
can
also
leverage
the
system
account
functionality
to
have
more
advanced
reports
of
the
status
of
the
system.
For
example,
you
can
now
query
the
number
of
servers
and
connections
that
are
connected
do
similar
things
that
you
can
also
do
with
that,
not
stop,
but
all
through
nets
which
and
that's
about
just
not
http
protocol.
A
A
The
server
not
server
hype
with
hyphen,
js
and
you'll
get
a
cool
banner
saying
that
the
jet
stream
got
started
and
by
default
has
dynamic
limits,
but
for
the
production
recommended
setups
we,
you
should
be
using
the
allocating
how
many,
how
many
memory,
how
much
memory
and
storage
you
want
to
use
for
for
jet
stream
setting
up
jet
stream
is
quite
straightforward.
It
just
takes
explicitly
in
case
of
a
production
setup.
How
much?
Maybe
you
want
to
use
what
is
going
to
be
the
storage
directory?
A
In
this
case,
we're
already
setting
up
a
system
account
name
assist,
and
something
that
has
to
be
noted
is
that
you
cannot
make
the
system
account
use
a
jet
stream.
So
at
least
you
can
enable
it
and
gesturing
globally
for
all
the
accounts,
or
you
can
opt
in
certain
number
of
accounts
so
that
they
can
only
a
certain
number
of
accounts
have
the
jet
stream
functionality
enabled
and
in
case
of
a
clustering.
A
It
is
very
similar
to
how
you
you.
You
would
set
up
a
regular
float
of
snap
servers.
In
this
case,
the
membership
is
determined
by
the
cluster
routes,
each
one
of
the
name.
The
servers
that
are
part
of
a
jet
stream
cluster
requires
to
have
a
server
name.
In
this
case,
we
have
the.
There
is
a
configuration
for
the
server
name
a,
and
we
are
part
of
the
cluster
that
is
a
b
c
d
e
in.
In
this
case,
we
have
a
five
nodes
as
part
of
the
jet
stream
club
cluster.
A
A
In
this
case,
we
are
monitoring
what
is
the
status
of
messages
that
are
being
persisted
into
jet
stream.
In
this
case
the
leader,
it
is
the
server
a
and
again
all
of
these
you
can
all
of
this
info.
You
can
get
as
a
nat
server
report
or
raft
gesturing
report
using
the
nat
cli,
and
in
this
case
we
know
that
the
node
a
is
the
leader
of
of
the
jet
stream.
What
is
they
called?
A
A
Name
b
has
a
number
of
other,
and
nazi
li
has
another
set
of
functions
to
able
to
manage
the
membership.
So
you
can
ask
any
of
those
members
to
step
down.
So
there's
like
a
new
election
and
there
is
a
another
member
takes
over
the
leadership.
So
you
can
replace
those
notes
in
case
there's
a
failure,
for
example,
so
again
all
of
through
their
nat
cli
again
through
that
naturalize.
How
you
can
also
make
streams,
in
this
case
we're
making
a
stream
named
foo
a
lot
of
the
defaults
and
these.
A
So
if
now
we
just
yesterday,
we
just
regular
nets,
so
regular
nets
protocol
is
all
it
takes
to
be
able
to
have
those
messages
be
persistent,
so
we
send
10
messages
into
put
into
foo,
and
then
we
can
create
a
consumer
that
in
this
case
it
was
a
pool
based
consumer
that
is
going
to
have.
Those
will
be
fetching
each
one
of
those
messages
and
in
terms
of
applications,
what
you
would
use
is
what
you
call
the
nats
client
jet
stream
context.
A
First,
I'm
gonna
show
you,
I'm
gonna,
show
you
how
it
works
in
go
so
you
as
part
of
the
nats,
go
library,
client
library.
You
will
first
start
a
connection
with
your
credentials
in
this
case,
I'm
just
using
a
user
password
and
then
call
the
jet
stream
function
as
part
of
the
the
connection,
and
that
will
give
you
an
enhanced
type
that
you
can
use
to
persist
messages
again.
A
It's
just
regular
nets,
so
classic
nets
on
through
will
also
persist
message,
but
they
we
didn't
change
the
interface
of
classic
nets
and
in
order
to
get
the
publish
with
the
act
responses,
we
have
a
new.
We
use
the
producer
setup
that
you
can
get
from
the
gesture
in
context.
So
now
you
can
publish
a
message
on
hello.js
and
receive
an
ack
that
acknowledgement
that
your
message
got
persisted
as
part
of
the
metadata
from
the
act
response.
A
The
push
based
consumers
use
the
not
suggestion.
Client
context
are
very
similar
to
regular
net
subscribers
they're,
essentially
the
same
you're,
just
reducing
the
just
in
context
to
create
on
the
fly
and
ephemeral
push
based
consumer.
This
is
ephemeral.
So
if
you
stop
this
application
as
it
is
in
this
example,
it
will
replay
all
of
the
messages
from
the
beginning,
and
you
can
see
that
you
can
mix
and
match
both
jet
stream
based
subscriptions
and
classic
nets.
A
In
this
case,
the
gesture,
the
the
jet
stream
one,
is
being
redelivered
all
the
previous
messages
and
the
classic
one
is
only
getting
the
new
messages
all
everyone
subscribing
to
through
push
in
order
to
make
those
applications
resist
the
replay
on
from
where
they
were
when
on
restart.
You
need
to
make
give
it
an
a
durable
name.
A
In
this
case
my
app
becomes
a
durable
name
from
this
subscriber
and
also
you
can
change
from
there
auto
hacking
that
the
previous
example
was
doing
so
that
you
can
manually
act
each
one
of
the
messages.
In
this
case
it
is
a
simple
act
that
you're
done
with
that
message
so
that
you
can
go
to
the
next
consumer
sequence.
There
are
multiple
types
of
acts
in
this
case,
we're
making
a
synchronous
subscriber
push
by
subscriber.
A
A
A
So
in
this
case
you
create
a
pulse
of
subscriber
on
foo,
using
the
durable
name
pool
and
it's
a
simple
for
loop.
That
is
going
to
be
fetching.
Five
messages
at
the
time
it
will
quickly
get
a
set
of
five
messages
messages
if
they
are,
there
are
any
ready.
Otherwise
it
will
wait
for
a
bit
before
retrying
next
part
of
the
loop
in
this
case,
giving
up
also
new
recommended
to
check
the
changelog
is
that?
A
Well,
we
also
released
the
initial
mqtt
support,
websockets
part
of
the
release,
jet
stream
adjacent
clustering.
Of
course,
there's
an
embedded
nas
account
server.
You
can
use
to
avoid
having
the
other
using
the
http
based
one.
A
gesturing
uses
heavily
the
message
headers,
but
regular
nas
descriptions
can
also
use
those
headers.
That's
why
there
were
two
new
protocols
hpop
and
each
message
and
other
features
such
as
traffic
shaping.
I
recommend
you
to
check
the
change
log
from
what
is
new
in
the
nas
documentation
for
the
roadmap.