►
From YouTube: CDF SIG Interoperability Meeting 2020-04-30
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
So
why
is
they
wait
for
others
to
join?
I
can
talk
about
the
sack
in
the
migration.
So,
as
we
discussed
during
the
previous
meeting
two
weeks
ago,
we
said
we
will
maybe
track
him
the
because
it
looks
easy
to
use
and
it
reduces
the
fault
we
spent
to
keep
the
notes
from
Google
Docs
and
then
sync
them
back
to
github.
So
we
tried
that
and
it
works
fine
as
far
as
I
can
tell,
and
because
of
that
I
took
this.
A
The
previous
work
or
Google
bot,
the
quant
to
read-only
mode
and
I
will
keep
it
there
as
historical
reference.
But
from
this
weekend
onwards
we
will
switch
the
MD
at
I/o.
So,
as
usual,
you
can
add
your
names
under
participants
for
today's
meeting.
We
can
verify
that
it
works
for
everyone.
You
just
click
this
time,
Seoul
pen,
whatever
sign,
and
then
you
should
be
able
to
edit
the
document.
A
So,
apart
from
that,
we
have
the
agenda
available
already,
starting
with
action.
Item
evolve,
followed
by
a
quick
update
about
the
newsletter
and
reason
why
I
have
newsletter
endogenous
this
once
my
team
is
interoperability
and
I
believe
it
will
go
out
either
today
or
tomorrow,
followed
by
seek
update
talk.
We
have
a
scheduled
have
time
in
next
week's
talk
meeting
which
I
want
to
go
through
slice,
quick,
be
and
then
I
was
a
pull
request
proposal
for
name
sake,
working
group
on
CI,
robots
and
I
reached
out
to
max
and
invited
him
to
our
meeting.
A
We
will
have
a
discussion
on
that
and
for
the
second
half
of
the
meeting,
we'll
have
a
presentation
from
Stefan
on
flagger
and
as
usual,
if
you
have
any
additional
topic
you
want
to
discuss.
Please
add
that
to
agenda
and
we'd
see
if
we
have
time
to
discuss
that.
So,
let's
start
I
think
we
have
more
people
now
and
I.
Don't
remember
any
action
point
from
the
previous
meeting.
I
don't
see
and
it
I'll
accept
hacking.
The
migration
which
we
completed.
So
we
can
move
to
the
next
one,
which
is
CDF
names
lot
of
Tracy.
B
Hey
everybody,
yes,
I,
believe
the
CDF
newsletter
with
the
interoperability
theme
is
all
set
to
go
out.
All
the
articles
were
submitted
last
week
and
I
believe
it's
slightly
been
held
up.
I
think
there's
a
backlog
on
the
technology
from
the
linux
foundation,
but
if
my
understanding
is
it's
going
out
today
or
definitely
latest
tomorrow,
so
yeah
many
thanks
to
all
the
folks
who
wrote
articles
for
that.
B
Additionally,
well
actually,
first
on
the
newsletter,
if
you
haven't
subscribed
to
it,
please
do
go
ahead
and
I'll
pop
the
link
in
there
and
please
share
it
and
if
we're
looking
to
get
some
of
these
articles
posted
on
other
sites
afterwards,
if
you
know,
if
anyone
would
be
interested,
let
us
know
and
then
one
thing
following
off
the
back
of
that
there
is
going
to
be
a
CDF
podcast.
So
if
you
haven't
seen
CDF
is
kicked
off
some
podcast
now
and
so
we've
got.
Some
of
the
authors
of
the
posts
are
gonna.
B
Do
a
little
panel
on
interoperability,
so
I
believe
Fatih
will
be
there
and
Cara
as
well,
and
we've
got
some
folks
from
puppet
as
well
who'll
be
joining
that
podcast.
So
watch
out
for
that
and
I'll
share
some
links
in
the
slack
once
once.
We
know
when
that's
going
to
go,
live
right,
I
think
that's
it
I
mess
any
questions
or
comments.
A
Yeah,
thank
you
so
the
next
topic
on
the
agenda.
There
was
a
discussion
during
last
week's
technical
site
committee
meeting
Tuesday
last
week,
and
one
of
the
discussion
topics
was
to
ask
six
to
come
and
present
to
talk,
so
it
can
be
aware
of
what's
happening
and
what
kind
of
work
the
six
are
doing,
and
we
volunteered
to
present
that
as
the
first
week
and
I
have
few
slides,
I
want
to
go
through
quickly
and
the
slice
are
already
linked
in
the
meeting
minutes
and
please
feel
free
to
update
slice
directive.
A
You
see
anything
missing
there
or
ask
questions
at
commands,
and
so
on.
It's
like
six
slides.
The
first
slide
basically
talks
about
what
this
week
is
about
when
it
was
approved
and
what
are
the?
What
are
our
aim
aims
and
so
on,
followed
by
what
we
actually
aim
to
do
with
the
sick.
This
is
not
the
complete
list,
because
we
have
lots
of
things
to
do.
A
We
have
ambitions,
but
we
have
to
start
from
somewhere
and
the
four
things
listed
only
slide
is
basically
we
want
to
follow
CDF
mantra
and
we
want
to
provide
neutral
forum
for
Aaron
from
different
communities,
companies
projects,
including
users,
to
have
discussion
and
have
that
discussion,
neutral
forum
enabling
the
dialogue,
and
we
also
need
to
ensure
the
shared
knowledge
between
different
projects
and
from
projects
between
projects
and
users.
And
finally,
the
main
item
that
we
have
been
working
on
last
couple
of
weeks
is
the
setting
to
roadmap.
A
So
it's
basically
the
updates
I
want
these
bullet
points
and
again,
if
something
is
missing
here,
just
add
that
and
try
adding
neutral
forum
and
basically
I
think,
given
the
trend,
we
have
been
seeing
within
six
participation
where
representatives
from
different
projects
like
Tecton,
Jenkins,
Eiffel,
open
de
jewel
and
so
on.
I
think
we
are,
we
can
say
we
are
successful
because
people
from
different
foundations
joining
to
our
meetings
and
presenting
the
tools
technologies.
A
The
other
thing
is
basically
enabling
to
dialogue.
You
know
we
have
been
having
regular
presentations
from
projects
and
users
to
ensure
that
we
have
representation
from
project
sign
users.
People
can
come
and
talk
about
what
they
are
doing
within
their
project,
the
latest
developments
and
how
it
relates
to
the
interoperability
area
and
as
part
of
that
activity.
A
We
realized
that
it
would
be
good
to
have
this
thing
called
vocabulary,
which
was
at
a
strong
describes.
It
pretty
well
thanks
Eric
that
so
that's
what
we
basically
started,
working
on
and
I
think
they
have
10
tools,
technologies
listed
in
that
document
and
hopefully
more
to
come.
So
we
can
really
make
this
a
rosetta
stone
for
CIC
the
ecosystem.
A
So
and
I
said
he
says
one
of
the
very
first
contributions
we
are
making
to
domain
because
I
don't
think
there
is
a
document
like
this
available
that
has
terms
of
more
than
10
projects
in
one
place
and
I'm
knowledge-sharing.
This
is
also
related
to
previous
one.
We
have
in
having
presentations
from
users
and
projects
and
up
to
now
we
had
six
presentations.
A
So
they
can
be
part
of
our
community
and
we
are
collecting
those
presentations
in
our
repository
available
for
anyone
to
look
offline
and
the
meeting
recordings
should
be
on
youtube,
cdf
youtube
channel,
but
only
the
first
one
is
available.
There,
I
open
the
ticket
or
an
issue
on
CDF
foundation's
report
to
get
the
remaining
recordings
available
on
YouTube
channel,
so
people
can
go
back.
Bojong
gede
hear
the
conversations
discussions
echoing
because
they
are
also
valuable
to
hear
and
the
last
slide
is
about.
Where
are
we
heading
and
thanks
to
Tracy?
We
start
organizing
roadmap
document.
A
A
So
that
was
the
update
to
talk
and
the
next
topic
on
the
agenda
is
discussion
on
proposed
special
task
of
working
group,
CI,
robots,
I'm,
not
sure
if
I
mean
of
you
have
noticed
this
to
request
opened
by
marks
to
talk
repository
and
as
you
will
notice
there.
Some
conversation
was
happening
because
some
of
the
key
words
used
in
the
proposal
cut
our
interest.
It
talks
like
integration
within
tools
and
so
on,
and
then
I
thought,
let's
reach
out
to
max
and
inviting
to
our
meeting
and
hear
from
him.
A
C
C
There
was
also
a
lot
of
motivation
to
implement
similar
functionalities,
which
most
of
the
time
end
up
somehow
in
well
really
weird
scripts
or
some
kind
of
yeah
pipelines
or
automated
partitions
in
it
which
yeah
interfere
with.
This
is
the
ICT
pipelines,
which
was
weird
to
be
honest
and
I.
Think
it's
missed
a
little
bit
the
the
the
idea
of
what
needs
to
be
done
so
I
thought
about
it.
C
Were
who
get
addressed
or
notified,
for
example,
if
there
was
a
request
or
if
you
want
to
qualify
something?
This
is
actually
everything.
What
happens
before
my
CIC
tea
party
can
start.
Of
course
you
can
integrate
directly
into
the
CIC
tea
also,
but
then
it's
I
think
on
the
Technic
level,
a
little
bit
too
much
down
to
be
visible
for
everyone.
C
C
Said
I
think
this
is
a
really
really
important
component.
That's
why
I'm
not
sure
if
she
are
robots
or,
however,
wants
to
call
it
to
say
it's
right
right,
waiting
for
it,
but
when
thinking
about
it,
it's
really
about
the
yeah
bridging
this
dis
gap
between
the
the
human
and
the
technical
part
of
the
whole
deployment.
D
Absolutely
I
think
we
can't
be
calling
a
chat
up,
so
you
mentioned
at
the
end
of
the
the
issue
of
the
idea
and
I
think
that
it's
kind
of
nice,
because
you
you
want
to
have
it's
not
just
about
receiving
notifications,
I,
think
having
that
greater
control
and
being
able
to
extend
it
as
well.
So
you
can
have
your
own
add
in
your
own
custom
and
alternations
as
well.
I
think
I
think
it's
really
really
good
I
think
it's
sort
of
creates
a
you
mentioned
process
this
on
the
court.
D
There
I
think
that
kind
of
developer
experience
as
well.
It
really
really
helps
aid
that
helps,
helps
developers
having
working
patterns
know
what
there's
a
matter.
What
project
they're
working
on,
but
having
a
consistent
way
of
using
they're,
interacting
with
their
CI
pipeline
to
treating
bills
or
or
approving
or
anything
in
test
results,
for
example,
I
think,
is
really
in
this
microservices
world
is
kind
of
more
and
more
important
to
have
this
calm
or
level
of
consistency
across.
B
Yeah,
a
couple
of
comments,
yeah
I
think
it's
really
interesting
again,
the
the
same
kind
of
things
we're
looking
to
tackle
really
ahead
again
and
just
you
know
the
different
types
of
terminology.
So
you
sort
of
use
the
robot
seminar
Lachie
Jenkins
X
is
using
the
chat,
ops,
one
and
again
it
sort
of
points
to
that
need
for
clarity
for
the
users.
So
we
can
move
that
forward.
My
battery
is
running
low.
I
need
to
go
deal
with
this,
but
yeah
also
I
be
interested
in
just
ways.
E
C
C
What
exists?
How
does
it
work,
then?
Maybe
we
have
also
the
first
steps
to
get
a
little
bit
clarification
on
the
wedding
which
I
think
it
needs
the
nice
output
of
it
and
I
think
through
this
process,
then
we
will
identify
okay,
this
is
workgroup.
Does
it
has
really
any
value
and
what
will
be
the
target
of
it
or
do
we
say
like
okay?
Maybe
well,
we
need
to
wait
whatever
one
two
years
until
on
the
marker
there's
a
couple
of
solutions:
bridging
this
gap.
A
This
one
clever,
because
this
proposal
is
on
CDF
top
level
and
we
are
having
this
discussion
just
because
out
of
curiosity,
we
are
not
discussing
this
because
we
are
asked
to
approve
this,
or
anything
is
just
to.
You
know,
have
a
dialogue
and
it's
it's
not
proposed
to
be
under
seeking
therapy
or
anything.
It's
just
a
proposal
like
how
signatory
parent
was.
E
E
E
It
already
has
what
accommodations
for
libraries
and
things
like
that
sure
what
I'm
just
having
a
hard
time,
understanding
what
the
work
would
be
if
it's,
if
we
started
for
a
cig
or
a
working
group
to
do
I
guess
the
best
thing
I
can
think
of
is
to
come
up
with
in
a
standard
interface
that
tools
like
robot
and
prowl,
which
is
no
kubernetes,
paste
and
robot
framework.
Could
it
here
would
that
be
correct?.
C
C
But
you
have
a
hard
time
to
adopt
it
to
your
own
tools
to
your
own
pipelines,
which
you
have
already,
because
you
have
a
huge
amount
of
patterns.
So
there's
for
some
of
the
question
to
answer
how
this
should
integrate
into
a
pipeline.
Is
it
just
a
simple
one
of
the
one
of
the
actions
which
needs
to
be
take,
or
can
it
somehow
intercept
the
pipelines
or
I?
Even
the
pipeline's
trigger
this
one
also
I
think
already.
C
This
is
like
I
said:
that's,
not
a
big
huge
field
with
tons
of
tools
around
yet
to
be
really
wanna
present
clear
and
precise.
How
this
can
looks
like
that's:
why
I
try
to
formulate
just
the
big
this
yeah
humans,
the
icd
gap,
into
a
way
we
can
say:
okay,
maybe
make
sense
to
investigate
further
into
it.
I.
B
I
was
gonna
jump
in
another
I
figured
out
technology.
Yes,
so
one
way
I
think
we
could
approach
this
like
with
the
road
map.
Undress
grew
mad
made
the
great
suggestion
we
break
things
down
by
tasks.
You
know
what
our
folks
trying
to
do
or
what
are
the
tools
trying
to
do
and
he
proposed
you
know
an
initial
whole
bunch
of
different
tasks
like
pre-deployment
and
post-deployment.
B
But
for
me
it
would
seem
like
okay,
let's
because
I
think
everybody
understands
one
set
of
tools,
but
maybe
not
across
the
different
ones,
but
if
we
could
have
ticket
task
by
task
and
then
sort
of
say,
okay,
these
are
the
things
Probot
are
doing,
and
this
is
how
it's
doing
it.
These
are
the
things
say,
chat,
officer,
Jenkins
X
are
doing,
and
then
we
identify
the
common
areas
and
then
we
again
just
go
through
that
process
of
saying
how
could
they
interoperate?
What
are
the
common
interfaces?
D
B
B
So
if
you
scroll
down
to
I,
think
the
next
section
so
I've
got
no
next
okay,
so
this
bit.
So
this
is
based
on
the
comment
and
dress
put
in
and
we're
mapping
out
all
sorts
of
different
tasks.
So
this
was
these
initial
proposals
allows
folks
to
take
a
look
at
that
see
if
it's
a
good
division
and
also
with
this
specific
kind
of
robot,
robot
use
case
I'll,
see
if
it's
I'll
take
a
look
and
see
if
it's
covered
well
and
how
we
can
talk
about
the
different
tasks
that
they
get
done.
A
Okay,
so
we
are
approaching
to
flag
your
presentation,
but
the
PR
max
opens
is
available.
Own
talk
repo,
so
the
discussion
is
not
limited
to
this
meeting.
Only
so
you
can
go
and
comment
directly
under
the
PR
and
keep
conversation
discussion
happening
there
and
then
the
usual
talk
process
will
probably
happen
coming
weeks
and
this
proposal
is
probably
revealed
and
they
will
look
at
the
discussion
under
the
PA
laser.
So
any
comments
just
put
under
the
PR
directly-
and
we
can
point
to
this
meeting-
means
as
well
as
a
reference.
F
My
screen
everything
good
yeah
cool,
hi
folks.
Thank
you
very
much
for
inviting
me.
Thank
you,
Tracy
for
reaching
out
I'm
very
happy
to
talk
about
tracker
here.
I
know:
Jenkins
X
is
using
it,
but
I'm
going
to
do
like
a
general
presentation
of
what
it
does
and
what's
on
the
roadmap,
and
please
interrupt
me
during
the
presentation.
I
don't
want
to
make
this
boring.
So
if
you
have
any
questions,
please
just
ask.
F
So
a
little
bit
about
progressive
delivery,
I
think
James
from
Redmond
came
up
with
the
with
the
term
is
just
an
umbrella
term
for
a
couple
of
techniques
that
you
can
apply
to
your
at
the
end
of
your
CD
pipeline
and
to
carry
religious,
80,
testing
blue,
clearly,
traffic
mirroring
and
all
these
sorts
of
techniques.
The
goal
is
to
make
production
release
news
more
stable
and
less
frightening
to
developers,
and
this
is
one
of
flagyl
scope.
F
F
Another
thing
that,
from
my
perspective,
is
the
requirement
is
that
the
pipeline
should
be
should
have
a
reconciliation
shouldn't,
just
be
something
that
reacts
to
things.
I'm
I
mean
mostly
I
systems
are
using
hooks,
so
you
push
something
triggers
a
pipeline
and
that's
it.
It
stops
I
think
the
CD
pipeline
should
reconcile
the
state
and
how
this
can
be
done
is
with
something
that
runs
on
your
cluster
detects
unwanted
mutations
alerts
about
them.
There
is
a
drift
in
configuration
and
then
tries
to
reapply
the
continuation.
F
F
The
project
started
two
years
ago.
We
set
out
some
goals
and
I
think
we
are.
We
check
them
all
this
year.
Of
course,
the
the
primary
goal
is
to
is
to
give
developers
confidence
in
automating
production,
which
it's
like
the
idea
is
you
can
deploy
on
Friday,
that's
the
I,
don't
be
afraid
to
deploy
and
go
home,
something
some
automation
should
roll
back
around
doing
something
goes
wrong.
F
F
F
Flagler
start
with
as
canary
release
operator
and
during
the
last
two
years
have
evolved
a
lot.
Now
it
can.
You
can
implement
a
couple
of
stretch,
deployment
strategies
with
flagger
general
issues.
We've
progressed
the
traffic
shifting
a
bit
testing
if
you
are
doing
things
for
front-end
apps
that
need
some
kind
of
special
affinity,
Bluegreen
deployments
with
traffic
mirroring
what
that
means.
Let's
say
you
have
a
she
learning
workload.
F
You
want
to
deploy
a
new
version
of
it
machine
learning
workload
is
idempotent
by
by
default,
so
you
can
just
mirror
all
the
traffic
flow
from
the
production,
one
to
the
one
that
you
want
to
deploy
and
based
on
some
metrics.
You
can
decide
if
the
new
model
fits
the
purpose
or
not.
So
the
blue,
green,
with
traffic
metering
has
is
specialized
for
only
idempotent
api.
F
Last
but
not
least,
is
the
classic
way
of
doing
Bluegreen
for
legacy,
apps
and
all
kinds
of
stateful
applications
that
write
things
on
on
disks
inside
the
container
and
so
on.
For
those
you
cannot
apply
these
modern
strategies,
so
those
people
ask
hey.
Can
we
do
Bluegreen
classic
way
with
right?
Now
it's
it's
implemented
and
you
don't
need
any
kind
of
service
natural
ingress
controller
to
do
Bluegreen
deployments.
It
works
with
kubernetes,
any
kubernetes
eni.
F
So
how
a
Kennedy
deployment
looks
like
you
have
version
one
running
in
your
cluster.
You
over.
You
want
to
overwrite
version,
one
reversion
to
flag
that
detects
that,
and
instead
of
just
doing
kubernetes
rolling
deployment,
it
stings
up
to
v2,
slowly
routes,
traffic
towards
it
it
measures
metrics.
It
can
alert
you
and
if
everything
is
ok,
all
the
metrics
are
good.
All
your
tests
are
passing,
then
it
will
replace.
V
1
will
be
2
in.
F
Not
at
once,
but
during
doing
the
rolling
deployment
over
t1,
so
the
couple
of
stages
in
the
canary
release
to
make
it
safe
to
make
to
wait
for
four
connections
to
drain
and
all
the
stuff
they'll
be
testing.
You
can
turn
flag
to
route
segment
of
your
users
to
the
new
version.
For
example,
you
can
say
route
all
my
insiders
users
that
have
let's
say
on
inside
the
cookie
or
depending
on
what
headers
your
English
solution
of
words.
We
can
say
I
want
to
route.
F
My
internal
users
or
I
want
to
route
some
users
from
some
region.
If
you
have
a
region
header
and
all
these
things.
So
here
it's
about
how
you
want
to
do
the
baby
test
and
what
information
you
have
inside
your
HTTP
requests
to
make
this
type
of
deployment
and
blue
green
in
the
blue
green
strategy
flag
around
conformance
test.
F
It
can
also
run
low
tests
and
if,
if
all
these
tests
are
good,
if
the
matrix
could,
without
routing
any
kind
of
online
live
traffic
to
v2,
it
does
the
switch
from
you
want
me
to
the
request
way.
This
is
the
most
wasteful
way
of
doing
a
deployment,
because
you
need
twice
as
many
as
much
compute
resources
doing
this.
F
How
do
you
actually
define
a
Canary's?
Well,
flagger
is
a
is
an
operator,
so
it
comes
with
the
Shirdi.
You
have
a
llamo
that
you
have
to
write
and
a
couple
of
things
here,
but
cannery
release
references,
a
deployment,
so
you
don't
have
to
define
in
here
in
this
resource.
Your
actual
diplomat,
it's
just
a
pointer
to
where
your
deployment
is
and
flagger
takes
over
that
that
deployment
and
does
a
bunch
of
things
for
you.
For
example,
you
can
specify
how
your
application
should
be
exposed
inside
the
cluster
and
outside
the
cluster.
F
Your
ports,
you
can
tell
it
retries,
depending
on
which
ingress
or
which
service
makes
you
use.
You
have
more
advanced
settings
that
you
can
do
the
service
and
the
most
important
thing
you
can
tell,
rather
how
to
run
the
analysis
for
your
app
that
what
interval,
what
kind
of
thresholds
and
what
metrics
flag
look
to
validate
the
analogies.
F
B
F
So
if
you
want
to
do
this
thing
without
flagger,
let's
say
use
something
like
this:
do
you
have
to
write
a
bunch
of
your
mods,
for
example,
you
have
to
duplicate
your
deployments,
your
deployment
Hammel
all
deployment,
the
new
deployment.
If
you
have
horizontal
pod
autoscaler,
do
you
have
to
duplicate
those?
If
you
have
config
maps
or
secrets,
you
also
have
to
duplicate
those,
because
changing
up
on
figma
for
change
occur.
Secret
could
bring
your
whole
application
down
like
most
down
panels
that
we
had
last
year
from
the
cloud.
F
Vendors
were
not
due
to
a
new
deployment
of
the
app,
but
due
to
a
Mis
configuration
of
that
deployment,
so
duplicating
all
your
configs
and
making
sure
you
ran
an
analysis
for
those
is
also
important.
So
you
also
have
to
create
service
mesh
objects,
neutral
service
routes,
traffic
policies
and
all
this
stuff
will
flag
girl.
You
have
to
create
your
deployment,
spec
kubernetes
declare
inspect
optionally.
F
You
could
create
a
horizontal
photo
scaler
and
that's
the
recommended
way
of
doing
it,
because
traffic
shift
with
only
one
instance,
a
single
pole
is
kind
of
silly,
so
you
should
be
able
to
scale
based
on
on
the
traffic
or
on
the
CPUC.
And
so
let's
say
you
have
these
two
communities
obviously
deployment
the
horizontal
photo
scaler
and
you
create
a
cannery
object
that
reference
that
deployment
and
all
the
things
on
the
deployment
application.
F
The
cluster
IP
services,
all
the
service
mesh
objects
are
created
for
you,
and
this
gives
you
an
advantage
to
move
from
one
service
nation
to
another.
For
example,
you
can
migrate
from
vco2
up
mesh
from
rubbish
to
link
ID
without
changing
anything,
because
flagger
will
detect
okay,
now
I'm
running
on
link
ID,
and
it
will
create
a
semi
objects,
move
to
up
mesh.
It
will
create
a
black
object
and
so
on.
So
it
makes
it
easy
to
it's
portable
across
different
implementations,.
F
What
are
the
supported
traffic
management
solutions
reported
East
your
link,
ID
and
app
mesh,
and
if
you
don't
want
to
use
a
service
nice
solution,
flagger
also
works
with
ingress
controllers,
with
contour
glue
arranged
annexed
and
the
can
respects
the
same.
You
can
just
swap
all
these
technologies
and
to
get
work
the
same.
F
How
does
Flagler
validate
a
deployment
it
uses?
Of
course,
the
deployment
health
status
from
the
kubernetes
api
software
is
the
kubernetes
api.
Are
the
pods
ready?
Are
the
pods
healthy?
Are
things
in
restarting
and
all
these
all
these
classic
kubernetes
information
that
you
have
are
over
a
deployment.
It
comes
with
built-in
matrix.
So
all
these
providers,
the
ingress
controllers
or
cell
or
service
mesh
offer
through
pramukh,
use
two
things.
F
You
can
look
at
the
request,
success
rate,
so
you
can
detect
if
your
application
is
returning,
500,
taels,
404
errors
and
all
these
things,
and
you
can
also
look
at
the
request.
Latency
average
duration,
so
you
can
say
if
my
new
version,
the
average
request
duration,
is
over
500
milliseconds.
Then
it
breaks
my
SLA
and
I
want
to
roll
it
back.
So
you
get
these
two
metrics.
F
A
F
Ok,
let's
talk
about
metric
templates,
so
this
is
a
new
feature
made
a
couple
of
months
ago.
At
the
beginning,
tracker
was
only
about
primitives,
so
ok,
Easter
comes
with
Prometheus
at
mesh,
comes
with
prometheus
Lincoln
takara
tomy.
So
it
was
the
natural
choice
to
look
at
the
Prometheus
server
and
get
all
the
metrics
from
there.
F
But
when
we
talk
about
business,
metrics
or
other
things
like
metrics
that
come
from
databases
and
another
another
sources,
then
people
use
data
dog
cloud
watch
and
there
are
so
many
other
metric
providers
out
there
for
now
flaggers
for
its
data
dog
and
cloud
watch
as
metric
providers,
and
it
has
it
offers
a
custom
resource
were
called
metric
template.
Where
you
can.
You
can
write
your
query
and
that
query
can
have
different
parameters,
for
example,
was
the
target
name,
the
main
specs
and
so
on?
F
F
Another
thing
that
flagger
allows
you
to
define
alerts
and
at
the
beginning,
when
install
flagger,
you
could
tell
it
hey.
Here
is
my
slack
channel
or
my
Microsoft
team
channel
or
a
discord
channel
and
all
the
canneries,
all
the
diplomas
that
are
running
on
your
cluster.
All
the
events
were
forwarded
to
that
channel.
Well,
that
doesn't
quite
scale,
because
if
you
have
like
100
tabs
ten
things
they
don't
not.
Everyone
wants
to
look
at
a
particular
channel.
F
Maybe
they
use
different
other
providers
and
so
on
so
to
make
this
configurable
there
is
another
custom
resource
called
other
providers,
so
you
can
define
slack
discord
rockets
at
Microsoft
things
as
the
provide
for
others
and
then
inside
the
can.
Redeployment
can
say
for
this
deployment.
I
want
to
rout
the
others.
These
providers
and.
F
I
want
to
set
the
verbosity.
Maybe
the
on-call
team
wants
to
be
alerted
only
when
can
early
deployment
fails
and
is
rolled
back,
but
maybe
the
QA
team
wants
to
know
when
when
a
deployment
starts
and
when
it
ends
so
there
are
different
levels
of
sending
others.
It's
also
about
the
webhook
future.
As
I
said
you
can
before
flagger
starts
to
route
any
kind
of
traffic.
Maybe
you
want
to
run
your
integration
test
directly
in
your
production
system
because
well
staging
should
be
perfect
copy
of
production,
but
usually
it
isn't.
F
So
maybe
you
want
to
have
like
a
final-round
integration
testing
running
directly
on
on
production,
or
maybe
you
want
to
define
some
smoke
test
or
some
conformance
test
that
you
can
wrap
with
in
help
and
another
tools,
so
the
web
hooks
are
are
different
types
of
web
hooks.
This
is
the
pre
rollout
type,
so
before
routing
any
kind
of
traffic,
you
can
run
this
kind
of
conformance
test
and
during
the
the
Kennedy
analysis
you
can
start
load
tests.
Why?
Because
maybe
you
don't
have
traffic
at
the
moment
that
you
are
doing
the
deployment?
F
So
if
there
is
no
traffic,
flagger
Gazan
cannot
query
the
primitive
server.
For
example,
it
will
return
like
no
no
data
points
and
your
your
Kennedy
will
fail.
So
you
can
use
load
tests
to
supplement
the
lack
of
of
live
traffic,
and
the
load
test
component
also
allows
you
to
bring
your
own
testing
framework.
There
are
couple
of
integration
integrations
there
and
you
can
easily
swap
healthcast
with
any
other
CLI
or
any
other
tool
that
you
may
use,
and
the
last
feature
is
about
manual
gating,
so
flagger
was
built
to
automate
everything.
F
People
that
want
to
press
buttons
when,
when
something
happens,
and
they
want
to
be
in
charge
of
the
declarant
release,
so
how
this
works
you
can
through
webhooks,
you
can
have
configuration
rollout
confirmation,
promotion
and
roll
back
and
you
can
easily
build
something
like
chatot
on
top
of
it.
So
let's
say
when,
when
flagger
detects
a
new
version,
it
will
tell
you
on
slack
head
detected,
a
new
version
am
I
allowed
to
start
the
cannery
deployment,
so
your
slack
mode
can
do
that
:
command.
F
F
So
there
is
another
type
of
web
hook
called
rollback
and
Flagler
will
will
will
pull
on
that
on
that
web
book
and
if
the
gate
is
opened
any
to
automatically
do
the
rollback,
so
you
can
implement
it
to
I,
don't
know
some
slack
column
and
a
third
thing
is
confirmed
promotion
so
after
it
does
all
the
analysis,
maybe
you
want
to
do
the
analysis
only
for
10%
of
your
traffic
and
you
want
to
leave
the
90
percent
10%.
You
want
to
leave
this
state
for
a
longer
time
than
the
analysis
time.
F
So
Flagler
will
ask
you,
can
I
do
the
final
rollout
and
you
have
to
confirm
it
and
only
then
would
happen
so
you
can
prolong,
for
example,
an
a/b
test
for
hours
or
days,
if
you
wish
to
do
so
with
such
a
worker
and
of
course,
flagyl
works
in
the
tops
pipelines
because
it
doesn't
mutates
the
object
that
you
are
applying
on
the
cluster.
It
creates
this
separate
thing
called
primary
and
it
also
detects
config
maps
and
secret
and
duplicates
all
of
that.
F
So
if
you
continuously
run
to
cattle
applied
from
idle
flux,
changing
the
X
are
low.
Cg
and
so
on,
Flagler
will
not
spin
up
new
new
deployments
all
the
time
because
it
had
it
has
checksum
mechanism
and
detects
if
the
real
change
or
if
it's
a
no
operation
changing
just
ignores
it.
So
it
talks
quite
well
with
good
hops,
one
more
minute.
Okay,
what's
on
the
roadmap,
I
want
to
improve
the
conformance
testing
feature
right
now,
it's
a
demon
that
runs
all
these
conformance
tests.
F
So
if
this
demon
for
some
app
needs,
some
are
bug
or
some
service
account
or
missed,
call
something
god
you
have
to
bundle
up
all
your
role
based
access
on
this
single
demon,
and
this
doesn't
quite
scale.
So
my
idea
is
to
use
kubernetes
jobs
or,
let's
say,
spin
up
a
Tecton
workflow
and
leave
that
workflow.
Do
the
conformance,
testing
and
just
report
back
so
imagine
your
conformance
test
must
call
some
database.
It's
an
I
am
role,
but
other
confirming
test
needs
to
be
just
a
ping.
B
F
Now
the
manual
gating,
where
I
go
I've
talked
before
it.
It's
all
these
gates
are
in
memory.
So
if
you
kill
the
pod,
all
the
gates
closed
are
closed.
So,
for
example,
even
if
you
say
over
the
weekend,
I
want
to
do
automatically
promotions.
If
something
restarts,
then
automatic
promotions
will
not
happen.
So
I'm
planning
to
come
up
with
a
proposal
how
to
do
persistence
of
the
manual
dating,
maybe
through
BCD
or
some
other
storage
like
this
tree
and
something
interesting
is
the
kubernetes
English
with
two
proposal.
F
Maybe
you've
seen
it
in
the
so
it's
a
new
proposal
for
ingress
ingress
right
now
is
very
limited.
If
you
want
to
change
things
specific
to
a
different
proxy,
you
have
to
use
all
these
crazy
applications.
Don't
have
validation
and
everything.
So
you
can
come
up
with
like
a
bunch
of
annotations.
For
example,
flagger
does
carry
diplomas
with
engine
X
by
setting
all
these
things.
F
So,
instead
of
having
all
these
annotations
in
Greece
v2
comes
with
some
well-defined
types
for
traffic
shifting
from
for
a/b
testing
that
I'm
so
happy
about
it
because
it
fits
so
well
with
with
what
plugger
does
so.
Even
if
no
one,
and
until
today
implemented
in
Greece,
we
I
plan
to
play
with
it
and
when,
when
some
English
control
switches
over
it,
flagger
should
be
able
to
to
work
with
it.
So
that's
the
end
of
my
presentation.
F
D
D
D
B
F
So
the
idea
is
that
you
have
to
have
two
kubernetes
deployments
running
in
parallel.
So
if
you
do
this
manually,
you
have
to
copy
paste
your
deployments
every
time
you
do.
A
new
deployment.
Also
flagger
does
dependency
detection.
So
it
looks
at
the
deployment
and
sees
what
kind
of
things
are
there
referenced
secrets
and
config
maps.
So
when
you
change
a
secret,
flagger
creates
a
copy
of
that
secret
transactional
analysis
for
your
change
and
afterwards
promotes
that
version
of
the
secret
to
the
new
version.
F
So
if
you
are
doing
this
manually-
and
you
want
to
defend
yourself
from
convince
ages,
then
you
have
to
have
for
each
secret
yeah
moles
in
your
repository
for
each
config
map,
yeah,
moles
and
so
on.
Right,
so
he
trims
the
hound
in
half
and
it
also
generates
all
the
service
mesh
objects.
So
if
you
look
at
what
what
do
you
need
to
do
to
make
this
traffic
routing
with
this?
Do
you
at
least
five
Yama's,
the
beautiful
service
routes
and
all
this
stuff?
F
B
A
A
F
F
A
Pdf,
whatever
format
PDF
loads
yeah,
so
they
are
all
at
all
time
so
to
reminders.
I
think
we
should
all
join
to
talk
meetings,
because
they
are
important
conversations
happening
and
it
happens
every
second
week.
The
next
talk
meeting
will
be
Tuesday
next
week
and
it
happens
so
4
p.m.
UTC.
So
it's
important
that
we
as
committed
are
and
take
part
in
the
conversations
that
happen
in
the
technical
side
committee
level,
and
the
second
is
reminder
is
about
the
next
meeting.