►
From YouTube: Argo CD and Rollouts Community Meeting 2nd Feb 2022
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
All
right
good
morning,
everyone
and
welcome
to
the
february
2022
argo,
cd
and
rollouts
community
meeting.
I
am
your
host
for
today
jesse
soon,
and
I
am
a
maintainer
and
lead
of
the
argo
project
so
today
we'll
be
covering
the
argo
upcoming
argo
cd,
2.3
release,
as
well
as
the
upcoming
argo
rollouts
1.2
release,
and
just
a
reminder
that
the
argo
project
adheres
to
the
cncf
code
of
conduct.
So
please
be
courteous
and
respectful
to
each
other
and
also
a
reminder
that
this
meeting
is
being
recorded
and
is
uploaded
to
youtube.
After
all.
A
Right
and
if
you
haven't
already,
please
use
the
link
in
the
meeting
to
add
yourself
as
an
attendee,
so
we
can
keep
track
who's
joining
alex.
You
want
to
go
first,
like.
B
B
Can
you
please
confirm
jc,
it's
visible,
okay,
awesome
and
all
right,
so
my
name
is
alex.
I'm
a
software
engineer
at
intuit
and
I'm
also
a
maintainer
of
argo
and
my
biggest
focus
is
argo
cd
currently
and
today,
I'm
going
to
give
you
a
short
update
about
upcoming
2.3,
rpg
release
and
yeah,
so
some
statistics
for
for
that
release.
B
Basically,
we've
been
working
on
it
for
last
three
months
and
during
the
time
71
contributor
pushed
200
plus
commits
into
argo
cd
repository,
which
is
quite
impressive,
so,
starting,
I
think,
mid
of
last
year
we
are
following
three
months
still
is
cadence,
so
basically
the
conclusion
is
that
we
are
releasing
a
new
set
of
features,
no
matter
what
every
three
months
and
in
that
release,
we've
got
kind
of
contributed
and
polished,
32
enhancements
and
67
bug
fixes,
and
we
didn't
just
work
on
minor
improvements.
B
We
also
have
four
major
features
and
I'm
going
to
cover
those
four
major
features
shortly,
and
the
current
status
is
right.
Now
we
have
a
first
release,
candidate,
ready
and
next
day
list
candidate
is
coming
shortly,
because
we
do
have
some
bugs
so
upgrade
this.
You
know
carefully
if
you
want
to
give
it
a
try,
but
we
pretty
much
done
with
most
of
bug,
fixes
and
a
high
chance
that
next,
the
risk
and
debate
will
be
3,
and
so
it's
not
the
hard
process,
but
very
likely
by
the
end
of
next
week.
B
2.3
will
be
ready
for
general
use,
okay
and
then,
let's
start
with
the
features
that
are
coming
in
the
new
release,
and
so
first
I
want
to
talk
about
a
new
thinking
strategy
that
was
that's
available
in
2.3
release,
and
I
want
to
start
from
the
problem
statement
first.
B
So
actually,
on
that
screen,
you
are
seeing
a
typical
a
very
common
example
that
happens
when
people
migrate
from
fixed
replica
set
in
your
git
to
hpa
and
hpa.
You
probably
know
it's
a
horizontal
autoscaler,
that's
supposed
to
manage
replicas
based
on
on
load,
and
so,
if
you
use
argo
cd
and
if
you're
in
that
situation,
where
you
have
your
replicas
in
git
plus
you
have
hp.
At
the
same
time,
rvcd
will
be
complaining
that
your
application
is
in
outer
sync
state,
because
number
of
replicas
in
git
no
longer
matches
to
what's
in.
B
B
It
will
just
report
that
application
is
in
sync
process
in
is
in
the
synced
status,
but
the
problem
is:
if
you
try
to
sync
that
application
arcgi
would
still
apply
the
whole
file
from
git
and
basically
it
would
temporarily
reset
number
of
replicas
back
to
what
it
was
in
git
and
then
hpa
would
revert
it
back
to
the
required
value
so
and
you
might
have
a
downtime
in
production
and
yeah.
This
is
the
so
next
screenshot
is
kind
of
demonstrating
how
you
can
improve
the
giving
and
in
2.3
release.
B
B
If
you
sync
application,
which,
which
is
in
outer
sync
state
in
this
case,
you
know
replicas
in
outer
sync
state,
argo
cd,
just
won't
touch
replicas
at
all
and
yeah,
and
you
won't
get
you
will
not
get
any
downtime
so,
and
that
is
we
are
very
excited
about
that
feature
because
it's
been
requested
for
like
a
year
at
least-
and
I
think
it
was
the
most
second
top
most
popular
feature
request
in
arbor
city
and
it's
finally
available
in
2.3,
and
we
have
a
whole
blog
kind
of
dedicated
just
for
for
this
feature
and
for
one
feature
which
which
I'm
going
to
cover
shortly
so
use
that
link
to
learn
about
how
it's
implemented
and
how
it
works
in
details
all
right.
B
And
so,
while
we
were
working
on
on
this
new
thinking
strategy,
we
also
realized
that
we
can
make
ignore
custom,
differing
functionality
even
more
convenient.
And
so
let
me
switch
quickly
back
to
my
previous
slide.
So
in
in
this
example,
you
are
seeing
that
the
user
kind
of
to
use
that
feature
you
would
have
to
configure
all
the
fields
you
want
to
ignore
one
by
one
and
so
in
real
life.
You
might
want
to
ignore
more
than
just
replicas
of
deployment.
B
B
Other
managed
fields
would
have
information
about
which
controllers
are
responsible
for
which
changes
and,
and
so
what
it
let
us
do.
Is
you
can
tell
to
argo
cd
that
you
just
trust
one
particular
controller,
and
you
want
to
ignore
all
the
changes
that
was
that
were
introduced
by
that
control
controller
so
and
to
to
make
it
a
little
bit
more
clear.
The
real
life
example.
Is
you
probably
want
to
trust
a
reload
and
deployment
controller
and
no
matter
what
changes
they
do?
B
You
just
want
to
ignore
them,
because
those
controllers,
you
can
consider
them
kind
of
a
system,
level
controllers,
and
so,
instead
of
listing
all
the
fields
that
are
touched
by
those
controllers,
you
can
simply
use
the
new
configuration
available
under
ignore
difference
setting
it's
called.
B
B
B
One
of
the
features
we
introduced
is
called
compact
resource
tree
and
the
idea
is
that
often
your
application
has
resources
of
the
same
time,
type
that
has
no
children.
So
the
example
here
is
replica
sets
so,
for
example,
replica
sets
and
pods,
and
usually
you
kind
of
don't
want
to
it's.
Usually
all
of
them
have
pretty
much
the
same
information
and
you
want
to
work
with
them
as
as
a
single
group
and
compact
resource
three
allows
you
to
do
exactly
that.
B
So,
let's
let
the
animation
play
a
little
bit
more
so
before
the
feature
is
enabled
this
application
doesn't
even
fit
into
a
screen
once
feature
is
enabled
you
kind
of
see
everything
on
on
prezi
small
screen
right
and
in
addition
to
that,
you
can
also
see
that
we
introduced
a
zoom
feature.
So
you
can
zoom
in
and
zoom
out,
and
it's
not
listed
here,
but
there
is
even
one
more
interesting
feature.
B
Okay,
let's
move
on
so
next
improvement.
Is
it's
not
really
a
feature,
but
it's
again
very.
B
B
B
So
you
probably
heard
about
two
projects
that
were
created
at
argo
labs.
One
of
them
is
argus
identifications
and
second,
is
application
set.
So
those
projects
were
available
as
an
extension
of
argo
cd
and
but
the
key
hat
was
that
you
had
to
install
them
separately
and
during
last
year,
both
of
those
projects
matured
enough.
B
So
we
feel
it's
now
appropriate
to
just
include
them
into
default,
targo,
cd
installation,
and
so
what
that
means
is,
if
you
never
use
those
projects
before
you
will
just
get
new
set
of
features
and
if
you're
using
them
already,
you
need
to
make
some
action.
Basically,
you
you
just.
I
won't
explain
exactly
what
you
should
do,
but
long
story
short.
You
don't
have
to
do
much.
You
just
need
to
remove
to
remote
customize
bases
if
you
use
customize
to
manage
argo
cd
or
you
just
need
to
cuba.
B
City
will
apply
argo
cd
installation
in
first
ones
when
you
upgrade
to
new
release,
and
if
you
want
to
learn
more,
please
check
upgrading
instructions
in
august
documentation
and
that's
it
that's
end
of
my
update
and
thank
you.
Thank
you
for
listening.
A
A
A
So,
yes,
all
right,
yeah
we
were
supposed
to
have
a
release
candidate
by
today.
That
probably
is
going
to
happen
later
today
or
tomorrow.
There's
still
one
pr
that
needs
to
get
in
before
we
cut
the
release,
but
I'll
go
ahead
and
start
showing,
what's
actually
in
the
release.
A
The
first
feature
we
actually
demoed
this.
I
wrote
it
from
databricks
demo
this
in
december,
but
really
quick.
If
you
use
analysis
with
rollouts.
A
So
this
is
the
analysis
driver.
A
The
second
feature
that
is
coming
in
is
we
call
weighted
experiment,
steps
and,
if
you're
unfamiliar
with
rollouts,
there's
a
feature
in
rollouts
that
allow
you
to
kick
off
experiments
as
part
of
the
canary
steps.
The
reason
that
the
primary
use
case
for
this
is
when
you
want
to
do
something
that
we
call
baseline
versus
canary
experimentation.
A
What
that
means
is,
if
you're
doing
a
metric
analysis
on
two
set
of
pods
a
lot
of
times,
it's
unfair
to
compare
the
canary
pod
against
the
production
pods,
which
have
been
running
for
weeks
on
end,
because
the
prod
pods,
the
their
cache,
is
all
warmed
up,
and
so
you
have
different
performance
characteristics
than
the
canary
which
had
just
started.
You
know
five
minutes
ago,
and
so
one
common
practice
with
canary
deployments
is
that
you
start
something
called
a
baseline
in
addition
to
the
carry
and
they
both
start
at
the
same
time.
A
And
that
way,
when
you're
doing
your
metric
analysis,
you
have
apples
to
apples
comparison
of
two
pods
that
actually
started
at
the
same
time,
and
previously
this
feature
was
only
available
if
you
were
using
or
only
useful.
I
should
say
if
you
were
using
the
basic
canary,
but
now
what
this
feature
allows
you
to
do.
A
Is
you
can
start
the
experiment
like
you
would
with
a
canary
but
then
specify
some
canary
weights
to
those
two
templates,
and
that
way
you
can
leverage
your
traffic
routing
capabilities
like
istio
or
link
or
d,
or
something
to
split
the
weight
to
those
baseline
and
canary
pods.
And
so
you
could
do
something
like
five
percent
go
to
the
baseline.
A
Another
five
go
to
the
canary
and
the
rest,
which
would
be
90
go
to
the
your
your
production,
pods,
and
so
I
think
I
think
we
have
initial
support
for
istio,
with
this
release
and
for
the
other
traffic
routers,
we'll
have
to
add
support
to
those.
A
Okay,
all
right,
so
this
this
is
a
little
bit
of
a
a
com,
complex
problem,
but
I'll
try
to
explain
the
issue.
A
If
you
use
alb
ingress,
load,
bouncer
controller,
you
may
have
noticed
the
behavior
of
alb
that
they,
they
inject
readiness
gates
on
pods,
depending
if
that
pod
is
reachable
by
an
in
alb
ingress,
and
those
readiness
gates
are
important
because
those
reading
skates
block
or
they
don't
allow
pod
to
become
ready
until
they
are
registered
to
the
target
group.
The
underlying
target
group
in
aws.
A
Here's,
like
the
caveat,
is
that
those
readiness
case
can
only
get
injected
if
there
is
a
path
from
the
ingress
to
the
service
to
the
pod
and
what,
in
practice
with
that?
What
that
means?
Is
they
don't
like
it?
If
you
change
this
selectors
of
a
service
like
after
your
pod
starts
up,
because
readiness
gates
can't
be
modified
on
a
pod
once
it's
already
running
okay.
So
now
I
explain
the
problem.
A
The
the
solution
which
under
guidance
and
feedback
from
the
aws
team,
is
that
they
they
suggest
that
we
essentially
never
modify
service
selector
labels
or
at
least
don't
modify
them
at
a
point
where
it's
your
pod
is
already
running,
because
otherwise
they
won't
get
those
readiness
gates
and
to
in
order
to
do
that
with
rollouts,
we
created
a
new
way
of
modifying
service
selectors,
which
we
called
ping
pong
service
management.
An
idea
is
that
you
have
two
services
and
you're
just
kind
of
going
back
and
forth
back
and
forth.
A
You
know
you
know
100
to
ping
and
then
on
the
next
release,
100
the
pong
and
then
so,
there's
no
necessary
concept
of
like
a
stable
service
and
a
canary
surface,
because
you
know
every
odd
release
is
the
ping
and
every
even
service
is
the
palm.
A
But
what
this
allows
us
to
do
is
it
allows
us
to
modify
surfaces
at
a
time
where
it's
before
the
pods
are
created,
so
the
ordering
would
be
update
the
ping
service
to
the
new
replica
set
hash
and
then
scale
up
the
replica
sets,
and
by
doing
so,
we
now
will
get
those
readiness
gates
that
alb
injects
and
then
those
are
important
because
it
helps
with
the
whole
zero
downtime
functionality
that
readiness
gates
helps
prevent.
A
There
is
a
link
to
the
diagram.
I
I
think
I
covered
this
in
another
meeting,
I'm
describing
the
problem
and
we
you
now
have
two
solutions
to
this.
We
in
the
previous
release.
We
had
something
called
target
group
verification.
It
was
aws
specific
feature,
but
now
we
have
this
other
second
feature.
If
you
want
to
use,
you
want
to
ping
and
pong
back
and
forth
between
two
services.
A
A
Next,
we
have
a
a
cool
feature
to
retain
a
customizable
number
of
metrics,
and
before
this
you
were
stuck
at
the
hardwired
version
of
10
metrics.
But
you
know
if
you're,
depending
on
how
frequently
you're
taking
metrics
10,
might
be
too
small
and
you
need
to
like,
if
you're
taking
it
every
minute.
That
means
it's
only
like
10
minutes
of
of
measurements,
so
this
is
now
configurable
that
you
can
set
at
an
analysis
template
level.
I
believe
it's
you
can
also
set
it.
A
And
then,
lastly,
we
have
a
bunch
of
other
features,
but
it's
now
possible
to
use
multiple
traffic
routers
at
the
same
time,
so
maybe
you're
transitioning
between
istio
or
and
some
other
or
nginx
to
issue
or
you
or
you
just
have
a
use
case
where
you
actually
use
both
at
the
same
time,
you
can
rollout
will
update.
A
You
know
the
both
the
seo
virtual
service,
as
well
as
the
nginx
ingress,
and
you
can
have
multiple
if
you're
using
the
web
metric,
it
was
previously
only
possible
to
do
get
requests,
but
if
you
have
a
need
to
post
something
with
a
body
or
put
something
with
the
body,
you
can
now
do
that
with
the
web
metric
some
metric
providers.
I
think
that
prometheus
one
has
been
improved
to
add
more
metadata
about,
like
the.
A
I
think
this
is
like
what
is
actually
returned
or
or
the
query
I
think
that
is
actually
formulated.
It
will
be
recorded
so
that
you
understand
the
exact
query
filled
out
with
the
template
parameters,
so
that
if
you
need
to
debug
the
query
that
caused
your
rollout
to
fail,
we
have
support
now.
So
this
means
you
can
scale
up
the
rollout
controller,
it's
active
passive,
so
it's
a
standby
in
case
the
the
leader
is
not
responding
and
updating
the
kubernetes
release.
A
You
do
need
if
you're
using
currencies
1.23.
You
will
need
this
release,
because
this
is
the
first
kubernetes
release
that
dropped,
support
for
the
extensions
ingress
group
and
so
1.2
will
now
exclusively
or
actually
not
exclusive.
It
will
start
using
the
networking,
v1
api
group
and,
lastly,
we
have
some
performance
improvements.
A
This
was
actually
covered
in
the
argo
con
presentation
that
I
did
with
salesforce,
but
essentially
you
can
control
some
queries
per
second
and
burst
settings
in
case
your
controller,
you
notice
it
needs
some
tuning
depending
on
your
level
scale,
and
we
do
have
app
mesh,
which
I
would
like
to
get
into
this
release.
It's
currently
in
review
right
now,
but
I
hopeful
that
we
will
get
that
one
in
to
this
release
as
well.
A
Okay,
so
that
is
the
argo
rollouts
1.2
release
and
I'll
stop
for
any
questions
that
people
have.
B
Thank
you.
I
had
one
question
that
not
sure
if
you
covered
it
in
in
presentation,
but
do
you
know
what
is
the
approximate
release
date?
Do
you
plan
one
or
two
in
next
week.
A
So
we
were
help
we
should
have
had
the
not
to
release
candidate
like
today,
but
that
got
caught
up
on.
I
said
we're
we're
still
trying
to
get
important
pr
into
that.
Typically,
we
let
the
rc
run,
as
you
know
like
intel,
will
want
to
run
that
internally
see
if
there's
it
causes
any
issues
in
some
of
the
pre-pride
environments,
but
also
I'd
like
for
the
community
to
at
least
get
their
hands
on
this
as
well.
A
All
right
did
we
have
any
any
other
business
or
after
topics
or
agenda
items
that
people
wanted
to
add
to
today's
agenda.
A
Okay,
if
there's
nothing
else,
I
think
we'll
end
a
bit
early
today.
Do
if
you
have
any
other
discussion
topics,
please
feel
free
to
add
them
to
the
agenda
for
next
month
or,
if
you
have
actually
any
features,
you've
been
working
on
and
want
to
present
or
integrations
with
any
of
the
cd
or
rollouts
products.
Those
are
also
very
welcome
to
present
in
this
form
thanks
jesse
great
work,
great
presentation,
yeah.
Thank
you.
Thank
you.