►
From YouTube: Istio User Experience Working Group April 30, 2019
Description
Istio User Experience Working Group session from April 30, 2019
A
So
we
have
a
few
things
to
talk
about
this
week.
One
thing
that
I
wanted
to
show
is
a
colleague
who
has
an
interesting
way
of
writing
controllers
of
a
script.
I
I
know
we
have
some
Callie
folks
here
to
show
us
some
key
features
and
also
a
few
work
items
that
have
been
marked
for
one
to
current
owners
and
I'm,
going
to
try
to
persuade
people
to
to
take
them
with
their
developers.
A
B
Good,
so
I
have
been
basically
working
on
a
JavaScript
library
for
h2o,
so
that
when
you
execute
the
code,
it
actually
generates
the
Hamel
so
that
you
don't
have
to
know
the
intricacies
of
the
experience
and
that
you
can.
You
know,
capture
some
patterns
there
that
are
common.
So,
first
of
all,
what
I
have
here
is
a
kubernetes
cluster,
where
I
just
installed
in
Co
and
I
installed
a
looking
for
application
without
anything
else.
So
the
first
thing
that
a
user
may
want
to
do
is
to
actually
expose
the
surface.
B
So
here
we
have
this
excellent
service
function
that
where
I
just
pass
it
some
parameters,
I
say
you
know,
I,
give
it
a
name.
I
say
this
with
the
external
court
that
I
want.
This
is
the
name.
Product
page
is
simple
of
the
service
that
I
want
exposed,
and
then
I
just
give
it
this
information
here
about
the
endpoints
and
now,
when
I,
if
I,
execute
this
and
actually
execute
it,
can
people
see
my
screen.
B
So
if
I
execute
that
it
basically
generate
the
two
llamo
that
I
need
to
the
Gateway
and
the
virtual
service,
and
by
using
that
I
actually
didn't
have
to
know
anything
about
what
a
gateway
or
a
service
see
are.
These
are
the
other
thing
that's
interesting
about
this.
Is
that
there's
some
things
here
that
have
been
automatically
filled
for
me
so,
for
example,
the
selector
here
is
this
name?
My
name
was
looked
up
to
so
the
kubernetes
cluster
was
queried
and
this
label
was
automatically
generated.
B
For
me,
the
other
piece
of
information
here
that
was
automatically
generated
was
it
was
this
internal
port
number
so
again
this
method
queried
kubernetes
and
obtained
that
information
and
filled
them
that
he
had
for
me.
What
I
could
do
with
this
is
that
I
can't
just
directly
send
it
to
you
know:
I
can
pipe
that
to
my
kubernetes
cluster
and
I.
Do
that
so
initially
you
know,
I
couldn't
see
anything
in
my
product
page
and
now,
if
I
look
at
it
again
now,
I
can
expose
that
service
and
now
I
can
see
the
book
info.
B
So
not
because
we,
the
hoping
for,
comes
with
three
versions
of
reviews,
which
is
not
realistic,
but
that's
what
happens
if
I
keep
refreshing.
It
goes
to
all
the
different
versions
of
reviews,
as
we
all
know,
so,
let's
suppose
that
I
want
what
I
wanted
to
do
was
to
actually
only
in
version
two.
What
I
can
do
is
to
write
the
rites
of
code
here
that
says
that
version
four
reviews
to
be
v2.
B
Incidentally,
what
I
also
have
is
this
piece
of
code
that
I
will
show
you
later,
that
is,
this
destination
rule
manager
that
actually
is
listening
for
deployment
we
created
and
deleted
and
automatically
generate
the
destination
rule
Hamel.
That
I
also
need.
So,
if
I
take
a
look
at
my
destination
rules
here
that
have
been
automatically
generated
for
me,
I.
B
I
see
that
in
automatically
has
generated
all
of
the
harmless
diversions,
based
on
the
way
that
these
things
were
labeled
and
I
will
go
over
the
code
for
how
that
was
written
in
a
little
bit.
So
now,
if
I,
if
I
generate
for
this
code,
that
stuff
everything
tv2
I
just
apply
that
and
then
you
know
it
creates
this
new
virtual
service
yeah
Moe,
but
as
a
user
I,
don't
necessarily
have
to
know
anything
about
that.
Now.
B
If
I
refresh,
we
see
that
it
worked,
and
it
all
goes
to
be
so
we
can
keep
on
going
now.
So,
for
example,
let's
say
that
I
wanted
to
have
a
route
rule
where
50%
of
the
traffic
goes
to
B,
1
and
50%
goes
to
V
3.
So
if
I
do
that
again,
when
I
refresh
it
only
goes
to
V
1
and
V
3,
so
then
actual
is
a
little
bit
more
complicated.
B
So
it's
basically
saying
that
if
the
user
is
a
name,
is
Jason
then
go
to
V,
2,
otherwise
go
to
V,
1
and
actually
the
set
way.
Since
this
is
it
more
a
general
abstraction
that
we
have
here
and
the
really
satellites
and
set
versions
are
actually
sugar
on
top
of
that,
so
this
actually
explains
a
little
bit
more
of
what
the
what
that
abstraction
is
where
you
really
have
a
source
and
a
destination,
and
you
can
you
can
give
attributes
to
that.
So
now,
if
I
apply
that.
B
A
B
B
So
I
create
that
what
happened
is
that
I
have
this
process
of
starting
on
the
background
that
actually
watches
for
the
creation
of
deployments
and
then
updates
that
piece
of
metadata
destination
reflected
data
automatically.
So
if
I
now
look
at
that,
so
this
is
the
reviews
destination
rule.
You
see
that
it
was
added
if
I
delete
it
again,
it's
going
to
be
removed
again
so.
A
B
What
this
does
is
that
if
the
watcher
is
watching
for
deployment
and
whenever
deployments
are
created
or
deleted,
it
automatically
updates
the
destination
rules
piece
of
metadata,
so
as
the
user
I
never
have
to
worry
about
that-
and
this
is
not
an
equity
before
as
a
container,
you
know,
together
with
East,
you
that's
running
in
the
background.
I
never
have
to
know
about
it.
That
pieces
may
take
it
metadata
it
just
to
maintain
for
me
without
me,
as
a
user
ever
having
to
worry
about
it,.
A
This
technique
seems
like
it
really
simplifies
people
who
have
custom
configuration
that
they
want
to
do
and
custom
reactive
things.
They
want
to
do
program.
These
DMS
wouldn't
changes
happen.
It's
also
something
a
great
way
to
prototype
new
source
of
futures.
That's
why
I
wanted
to
take
a
look
at
it
and
give
us
give
my
daughter
her
thoughts,
our
thoughts
on
it
I.
D
Think
that's
like
a
great
option
when
we
want
to
automate
some
santa
clauses,
because
actually,
for
example,
for
me
what
I'm
doing
I'm
just
having
some
placeholder
in
in
the
Yama's
file
and
then
try
to
change
this
when
deploying
depending
on
the
environment.
So
that's
like
give
us
like
a
more
or
more
obstructed
way
to
do
it.
Without
this
going
get
me
something
the
MFI
lunches,
change,
valuables
or
change
the
placeholder.
A
B
A
B
E
B
I
think
that
so
there
was
an
interesting
blog
about
templating,
yeah,
Moe
versus
you
know,
outside
of
the
context
of
ECU
versus
programming
languages
and
there's
definitely
people
who
believe
on
both
sides
of
that
discussion
and
and
we're
very
kind
of
set
in
there
under
way
of
thinking.
So
we
probably
want
both
when
what
happened
with
templating
languages
is
that
you
know
you
reach
the
limit
of
programmability
at
some
point,
and
you
don't
want
to
put
a
full-fledged
programming
language
in
your
templating.
B
So
so
then
people
always
find
something
wrong
with
it,
because
the
thing
that
they
wanted
to
write
is
an
expressible
etc
and
they
keep
adding.
You
know
like
people
throwing
up
floor
and
happy
with
helmet,
cetera,
so
I
think
what
was
nice
about
this?
The
JavaScript
is
actually
the
PL
approach
is
that
you
just
leave
it
to
the
user
to
be
able
to.
You
know,
write
whatever
they
want,
and
maybe
you
write
that
for
certain
ops
personas
who
aren't
comfortable
in
code
at
all
this
isn't
for
them.
B
But
on
the
other
hand,
this
could
be
for
a
developer,
who
wants
to
build
tools
for
adopts
persona?
That
would
be
easy
to
to
develop
so,
for
example,
you
could
imagine
that
you
say
under
certain
conditions.
You
know
this
is
how
I
want
to
change
my
routing
and
be
able
to
codify
that
in
a
programming
language
to
see
if
I
see
if
I
watch,
I'm
watching
and
I
see
something.
B
In
my
you
know
my
log
or
if
I'm
analyzing
this
traffic-
and
this
is
what
I
you
know
detected,
then
in
that
case
you
know
this
is
how,
to
you
know,
I
can
change
the
routing
or
do
whatever
against
conversion,
because
it's
gone
bad
and
then
that
this
is
a
new
tool
that
can
be
rapidly
developed
by
somebody
to
me
than
by
an
awful
person.
Yeah.
F
There's
a
there's:
a
company
called
Paloma
bul
uni.
That's
got
a
different
approach
to
do
to
doing
deploy
native
stuff
and
in
general
they've
created
a
strongly
set
of
wrappers
around,
for
example,
for
Google
platform.
They've
created
strongly
type
wrappers
around
all
types
of
Google
objects,
so
VMs
and
networks,
and
all
that
kind
of
stuff-
and
you
end
up
writing
a
little
program
that
creates
all
these
objects,
links
them
and
a
strongly
type
fashion
perform.
F
B
Yes,
absolutely
supremely
is
completely
related
to
what
we're
doing,
and
in
fact
we
were
somewhat
inspired
by
that
and
and
we're
you
know
in
our
programming
module,
we're
trying
to
build
stuff
for
IBM
assets
in
in
that
in
that
in
that
vein,
but
approach
that
we're
having
is
that
will
based
on
operators
we're
based
on
CRD.
So
we
have
this
strong
sort
of
kubernetes
rather
than
having
you
know
these
wrappers
that
just
do
rest.
B
Api
calls
and
they're
done
with
our
approaches
to
build
the
program
model
on
top
of
operators
that
are
doing
the
lifecycle
management
of
these
various
resources,
and
so
what
god
gives
us
is
that
there's
some
health
checks
that
the
operators
themselves
are
doing
to
maintain
the
health
of
the
resources
that
they're
managing
and
then
we
build
on
top
of
that.
So
that
would
be
that's
one
way
in
which
were
different
than
Cooley,
but
it's
different,
definitely
the
same
space.
Absolutely
that's
the
probably
the
most
related
thing.
That's
out
there
is
follow
me.
Yes,
I.
A
Heard
I
think
Lynn
asked
about
templating
approach
versus
a
programming
language
approach.
I
found
that
the
templates
were
great
for
writing,
but
for
some
of
the
defaults,
like
the
port
number
I
found
that
I
was
using
JSON
fat
to
try
to
kitten
get
the
defaults
out
of
all.
There
is
do
objects
and
those
Jake's
iPads
turn
out
to
be
much
more
complicated
than
you
would
think
to
write.
It's
nicer
to
be
able
to
write
them
in
a
programming
language.
B
Yes,
and
also
in
the
programming
language,
you
are
sort
of
reusing.
You
know
you
could
have
a
variable
name
for
something
and
then
just
reuse,
that
variable
name,
you
know
use
pasture
out.
You
know
for
something
that
repeats
itself
is
like
five
from
the
animals,
so
those
are
like
just
basic
things
that
appeal
language
gives
you
that
you
may
not
have
necessarily
notably
ready.
G
G
You
know:
Colleen
we're
going
with
things
with
that,
so
I'm
going
to
go
through
kind
of
give
a
brief
update
on
where
we
are
with
Kelly
and
then
I
can
have
at
this
slide
at
the
end,
talking
kind
of
about
our
future
plans
with
it.
So
this
is
kind
of
just
the
simple,
like
you
know,
product
page
on
place
to
get
more
information
about
Kelly
or
main
website
Kelly
that
I
Oh
ple
under
github.
We
have
right
now,
like
our
official
issue.
G
Tracker
is
under
JBoss
in
there
JIRA
instance,
but
we
found
for
some
users
it's
not
handy,
to
enter
issues
there.
So
we
also
cut
this
is
being
opened
by
our
through
the
github
issues,
and
then
we
also
the
list
here
some
of
our
social
things.
You
can
kind
of
click
on
to
find
more
information,
that
is
your
Twitter
or
medium
there's,
also
a
YouTube
tail
in
there.
G
So
I'll
start
off
just
kind
of
you
know
describing
what
ple
is
where,
if
do,
our
ability
console
ie
to
see
what's
happening
within
your
mesh,
find
out
information
about
that?
Okay,
it
means
how
much
traffic
is
flowing
between
your
different
components,
different
metrics
and
graphs,
and
such
in
there.
We
also
have
some
mesh
configuration
management
within
ple
itself.
There's
things
like
wizards.
If
you
want
to
set
up
routing
between
components
that
way,
you
can
also
view,
and
edit
UML
configuration
files
to
allow
you
to
set
up
and
figure
through
things
any
way
you
want.
G
We
are
a
open
source
project
under
the
Apache
2.0
license.
Currently,
most
our
contributions
are
coming
from
Red
Hat,
but
we're
looking
to
expand
and
grow
our
community,
so
people
want
to
help
out
with
Cali.
Please
let
us
know
if
you
have
any
thoughts
or
comments
of
things
that
you
want
done
in
there
please
reach
out
to
our
community.
G
What,
if
forgot
on
a
previous
slide,
actually
is
our
mailing
list.
They
do
have
been
mailing
lists,
so
you
see
people
to
find
that
from
ela
bio
and
if
you're,
using
something
like
the
sto
quick,
install
key
Ellie
is
installed
by
default.
So
if
you've
done
that,
you
might
have
already
played
around
key
ally
or
seem
it,
it
can
also
be
optionally
installed
using
the
helm,
installer
I,
don't
believe
this
installed
by
default,
you
get
past
a
parameter
for
key
ally
to
be
in
there.
G
G
I'm
assume
people
to
see
my
screen
again
so
I'm
on
a
backup
cluster
right
now,
which
does
not
have
things
configured
the
way
I
wanted
it
to
for
the
demo,
but
that's
okay,
so
for
people
who
have
not
seen
ple.
This
is
what
our
current
version
of
Kali
looks
like.
This
is
based,
I,
think
on
either
master
or
close
to
master.
So
if
there's
any
kind
of
bugs
that
pop
up,
it
might
just
be
because
of
that.
G
Normally,
when
you
access,
kale,
you'll,
be
granted
access
to
or
TV
show,
on
access
to
the
overview
page,
which
is
a
brief
overview
of
what's
happening
within
your
cluster.
Here
we
can
see
that
there's
five
different
namespaces.
We
have
information,
but
what's
running
in
the
namespaces,
this
is
booking
applications
a
bit
of
traffic
information
and
other
things
in
there.
So
from
here,
if
there's
any
problems,
you
will
get
a
kind
of
a
red
or
orange
indication
that
something's
wrong
in
your
cluster
and
then
quick
links
to
go
to
something's
in
there
so
something's
on
fire.
G
You
should
have
access
to
go
quickly
to
that
I'm,
just
trying
to
see
if
my
clusters
up
and
running
sis
and
how
that
my
other
one
but
I
think
it's
instead.
Okay,
so
we'll
continue
on
instead
that
this
one
doesn't
have
an
error
setup
in
it.
So
here
is
our
graph
view
that
we
have-
and
this
is
showing
you
what's
happening
between
your
services
or
your
micro
services
in
here
your
different
workloads
and
applications
right
now.
This
is
looking
at
the
info
demo
that
I'm
sure
most
people
are
very
well
aware
of.
G
You
have
things
like
the
side
panel,
there's
various
different
ways.
You
can
render
and
view
your
your
graph
right
now,
it's
a
version
tap
graph,
so
it
kind
of
groups
things
together
by
application
inversion.
But
if
you
don't
want
that,
you
can
get
things
like
a
workload
graph
which
will
separate
things
by
workload.
G
I'll
just
go
back
to
the
version
tap
graph,
as
that
seems
to
be
the
nicest
one
that
we
have
in
here
and
my
other
example
was
going
to
show
I
had
a
fault
injection
setup,
and
it
was
quite
nice
as
it
would
show
you.
The
edge
being
read
from
a
fault
injection
and
then,
if
you
clicked
on
the
side
panel
under
response
codes,
it
would
show
you
a
500
response
and
then
a
flag
value.
G
So
if
you
had
something
like
false
injection
or
a
circuit,
breaker
or
something
else,
it
was
activated,
you
will
actually
know
what
was
happening
in
your
graph.
But
it's
not
very
interesting
here
as
it's
not
in
able
to
give
this
environment
there's
a
few
different
things
that
you
can
do
as
I
said,
there's
different
ways
you
can
set
up
your
graph
to
be
shown.
You
can
show
different
things
on
the
edges.
G
So
if
you
want
to
see
the
requests
per
second,
you
can
see
that
if
you
want
to
see
if
there's
a
percentage
because
percentages-
so
you
can
see
like
for
these
right
now-
it's
50/50
because
there
isn't
any
routing
being
applied
and
also
things
like
response
time.
So,
if
you're
doing
something
like
deploying
multiple
versions-
and
you
have
a
new
version
rolling
out
the
canary
and
you're,
seeing
the
response
time
is
really
high.
You
can
use
this
to
kind
of
view
that
information
in
the
graph
there's
also
some
things
like
badges
that
we
have
applied.
G
So
if
you
have
a
current
product
page
here
as
a
virtual
services
applied
to
the
product
page
service,
you
can
list
in
there
and
there's
some
other
things
in
here,
such
as
circuit
breakers
being
applied.
Have
it
all,
underneath
your
under
display
option
give
a
missing
sidecar.
So
there's
you
know,
we
have
a
check
that
if
you're
running
without
a
sidecar,
you
can
know
about
that
within
the
graph,
and
it's
also
some
security
information.
So
with
the
security
information,
it
will
show
you
what
what
connections
are
being
done
over
usual
TLS
or
not.
G
Have
a
badge
that
will
show
that
it's
configured
to
have
one,
if
there's
actually
a
circuit
breaker
that
has
been
triggered,
the
line
will
be
either
orange
or
red.
Depending
on
how
many
failures
there
are,
and
then
you
can
go
into
the
response
code
and
under
here,
there'd
be
a
flag
with
a
hover
over.
That
would
tell
you
if
a
circuit
breaker
was
activated
and
what
percentage
of
the
requests
were,
that
error
code
was
triggered
by
a
circuit
breaker
same
thing,
for
something
like
fault,
injection
or
other
things
in
there.
So.
G
This
is
a
brief
overview
of
our
graph
part
of
things.
We
also
have
a
lot
of
different
other
page
that
we
can
go
into.
We
have
a
list
of
applications.
This
goes
through
and
basically
groups
things
together
by
the
app
labels.
If
you
see
on
the
page
here,
your
product
page,
which
has
a
b1
work
load
and
a
product
page
service
instances
under
the
they
both
have
the
app
label
applied
to
it.
G
It's
grouped
under
the
product
page
the
staff
label
matches,
so
we
can
go
through
and
look
at
any
application
you
have
such
as
the
detailed
page.
This
will
show
you
what
workloads
are
available
under
that
application.
What
services
are
available?
We
have
health
checks
here,
so
here
we're
checking
the
pod
status
and
what
the
pods
are
healthy.
We're
also
checking
things
such
as
the
network
traffic.
G
So
if
there's
errors
in
the
network
we'll
mark
something
as
being
unhealthy,
we
have
ethic
pages
which
will
give
you
a
overview
of
the
traffic
that's
flowing
through
this
particular
application.
We
also
have
things
like
the
inbound
and
outbound
metrics,
so
you
can
see
what's
happening
through
this.
You
can
also
expand
it
via
full
page.
If
you
have
Crisanta
configured
and
installed,
you
also
have
a
link
that
will
take
you
to
brief
fauna.
We
have
similar
things
for
all
the
for
something
like
a
workload
you
can
go
through.
G
Workloads
see
what's
happening
with
things
in
your
graph,
so
you
can
see
here
for
the
details.
V1
workload
we
have
you
know
what
type
it
is
is
deployment
the
athens
version,
labels
that
are
applied
the
health
of
this
workload.
We
also
can
see
what
pods
are
in
here
at
the
pods
healthy
or
not
other
information.
With
that,
we
also
see
what
services
this
work
was
being
accessed
through.
You
can
get
traffic
around
the
workload,
so
you
can
see.
What's
this
workload
is
talking
to
was
talking
to
this
workload?
G
If
you
want
to
see
logs,
we
have
the
logs
in
here
as
well.
So,
if
you're
noticing
something
is
acting
a
little
bit
weird
in
your
graph
and
you're
curious,
what's
happening
in
terms
of
the
logs
you're
saying
the
area
or
something
else
you
can
jump
in
here
and
see
where
the
errors
are
and
other
information
is
that
you
found
and
about
metrics.
We
also
have
for
workloads.
G
If
you're,
we
have
an
optional
way
of
adding
in
your
own
custom
dashboards
for
workloads,
and
I
think
we
have
this
under
the
galley
one
in
this
environment,
not
sure
if
it
is
or
not,
but
for
kelly,
yeah,
it's
not
in
there,
but
there's
an
option
that
you
can
use
that
will
allow
you
to
create
your
own
dashboards
within
qiao
a
so.
If
you
have
custom
metrics
that
you
want
to
show,
you
can
show
them
to
running
a
go
application.
You
can
have
your
go
application
metrics
being
shown
in
here.
G
If
you
have
custom
metrics
that
you're
having
here,
you
can
also
do
it
as
well.
We
have
a
CR,
a
custom
resource
that
needs
a
define
that
will
set
that
up.
That
can
be
quite
handy.
We've
use
that
quite
a
bit
with
key
ally
with
some
of
our
debugging
is
that
we've
made
it
easy
to
see
what
certain
functions
in
key
ally
are
taking
longer
than
others
to
warm.
So
we've
had
some
problems.
G
We
had
some
issues
with
certain
end
points,
taking
a
long
time
to
reply
back,
and
we
piece
that
kind
of
debug
and
look
into
those
issues.
There's
also
you
know
other
things
in
here,
but
I'm
sure
people
can
go
through
on
their
own
and
probably
look
at
these
a
little
bit
more
just
kind
of
looking
at
these
issues
on
their
owns
a
little
bit.
I
think
dry.
So
I'll
get
more
into
a
little
bit
of
a
demo
here.
So
in
this
instance,
when
I
get
rid
of
these
two
system,.
G
We
have
something
like,
let's
say,
ratings
where
we
have
z1
and
z2
and
let's
say
I
want
to
send
90%
of
my
traffic
to
v1
+
10
%
to
v2.
So
if
you
want
to
do
that,
first
of
all,
you
can
check
right
here
by
saying
the
percentages
and
seeing
what
it
is
right
now.
So
we
see
that
the
traffic's
about
50/50
between
the
two,
but
we
want
to
change
that.
So
we
can.
Click
on
ratings
then
go
into
the
rating
service,
and
from
here
we
have
some
actions
that
you
can
perform.
G
So
you
can
create
a
weighted
routing
route,
so
I
can
say:
I
want
v1
to
have
what
I
said,
90%
and
then
v2
to
have
10
and
I
can
create
this.
This
has
now
been
created.
You
can
now
see
that
I
have
my
virtual
service.
Here,
that's
been
created.
You
can
see
the
the
weights
that
we
have
with
this.
You
can
also
go
into
the
animal
here
if
you
want
to
edit
the
animal
directly.
A
G
D
G
G
A
G
Yeah,
well,
if
you
get
a
like
a
the
yes,
it's
pods
aren't
really
a
concept
in
what
we're
dealing
with
the
telemetry
and
I.
Don't
think
it's
even
a
concept
with
any
of
the
yamo
files,
it's
all
dealing
with
workloads
themselves
which
for
a
lot
of
simple
examples
here,
like
the
the
v1.
This
is
composed
of
items.
The
workload
here
is
just
one
pots.
If
you're
like
dealing
with
the
default
hello
world
example
a
lot
of
times,
it's
just
a
single
pod
in
the
workload,
but
there's
there's
nothing
to
stop.
D
G
You
can't
have
you
can't
have
a
deployment
that
would
have
different
v1
v2
v3.
It
would
all
have
to
be
the
same.
It's
not
possible
to
have
odds
under
deployment
that
have
different
labels.
Technically,
if
you
have
a
replica
set
based
workload,
then
you
could
have
pods
underneath
it
that
would
be
from
the
same
replica
set
but
would
have
different
version.
Labels
apply,
but
it's
kind
of
a
different
situation
that
it's
not
really
I,
think
it's
more
of
an
anti-pattern
be
doing
things
that
way
for
most
things,
we're
expecting
users
for
doing
deployments.
G
C
G
Then
you
can
do
that
as
well.
Using
this
other
configuration
wizard,
there's
also
options
that
we
have
in
here,
which
is
suspend
all
traffic.
So
if
you
decided
that
you
something's
wrong
and
to
kick
traffic
going
to
certain
workloads,
you
can
come
through
here
check
a
box
and
automatically
suspend
all
of
it,
so
traffic
stops
flowing
through
it.
G
You
know,
being
able
to
use
a
these
kind
of
bars
to
specify
how
much
traffic
goes
to
each
one.
I,
don't
need
to
worry
about
any
mo
files.
I
can
create
hit,
create
it's
automatically.
It
creates
the
virtual
services
and
destination
rules
for
me.
So
that's
what
we
found
to
be
useful,
but,
yes,
we're
expecting
to
add
more
in
there.
Even
with
the
things
of
the
virtual
service
is
we
have
a
kind
of
a
page
here.
G
You
can
see
it,
but
there's
also
a
yamo
file
that,
if
you
want
to
you,
can
go
through
and
at
the
ample
file.
You
can
also
I
think
delete
it
from
here.
If
you
want
to
get
rid
of
it,
I
don't
think
we
have
create
at
the
moment
just
because
we
found
that
creating
things
at
least
just
from
a
pure
kind
of
text.
G
Wizard
is
not
that
handy,
we'd,
rather
use
or
say
for
like
a
text
fields,
not
the
handy
values
wizard
for
people
to
create
things
and
then
even
which
is
I
think
we
shown
before,
but
we
can
update
things
like
the
the
amble
directly,
so
they
say
I
had
it
called
ratings
instead
of
ratings.
It
can
like
update
and
show
you
that
you
have
errors
specified
in
your
EML,
because
you
know
the
hosts
here
is
not
found
and
can
fix
it
by
ratings.
It's
a
and
it's
it's
green
again.
C
G
C
Be
a
possibility
now
I
was
wondering
if
you
could
just
use
key
Olly
to
do
a
complete,
end-to-end
SEO
demo
like
to
the
point
where,
if
you
take
the
prerequisites
that
you
have
a
cluster
which
is
do
installed
in
an
application,
that's
not
part
of
the
mesh.
Can
I
just
use
key
Olly
to
add
that
application
to
the
mesh
may
be
like
enabling
site
core
injection
for
our
namespace.
You
know
somehow
trigger
it
to
restart
pick
up
that
envoy
and
then
add
virtual
service
routing
and
then
applied
and
and
do
the
whole
thing
that.
G
Would
be
very
cool,
I
think,
but
it's
not
something
that
we've
thought
about
yet,
but
that
is
very
interesting,
especially
be
able
to
take
something
like
an
existing
application.
Just
like
clicking
on
it
and
saying
you
know,
did
it
yet
if
you
apply
this
inject
the
the
sidecar
or
whatever
to
this
and
continue,
that
would
be
very
interesting.
It
might
also
be
interesting.
We
could
I
think
also
create
the
mo
files
may
be
in
here
somewhere
or
allows
someone
to
upload
the
mo
file.
G
C
E
I
guess
this
was
a
kind
of
like
it's
do:
CTL
mesh
if
I
command
that's
been
discussed
in
this
workgroup
early
arm,
but
it's
more
on
the
UI
level.
What
can
we
do
to
make
a
user
life
easier
if
they
come
to
coyote?
They
are
looking
at
this
not
just
up
to
what's
going
on
within
their
mesh,
but
also
adding
services
to
their
mesh
right
from
the
UI.
G
G
C
G
There's
been
a
some
interest
in
that
on
our
mailing
list.
So
far
so
right
now
we
have
three
different
dedication
mechanisms
you
can
log
in
with
a
simple
username
and
password.
You
can
not
have
any
authentication
login
as
like
an
anonymous
user.
If
you
really
want
to
of,
though
that
means
anyone
who
accesses
key
ally
has
access
to
everything
in
there
and
then
we
also
have
a
integration
with
the
open
shift.
G
Go
off
that
allows
you
to
log
in
with
your
open
ships,
credentials
and
that's
also
tied
in
with
your
permissions
in
the
open
ship
cluster.
So
if
you're
running
openshift
and
use
that
option,
if
there
say
like
10,
different
namespaces
in
your
cluster,
but
you
only
have
access
to
one.
If
you
log
into
your
credentials,
key
Ali
will
only
show
you
that
one
name
space
so
we've
had
some
things
about.
G
You
know
potentially
using
something
like
generic
Roth,
where
anyone
can
configure
Roth
to
get
into
key
Olly,
where
I'm
not
sure
exactly
how
we
want
to
handle
that
at
this
point,
I
think
the
openshift
one
is
kind
special
case,
because
it's
tied
in
with
their
permissions
in
the
cluster.
If
you
wanted
to
see
a
lot
in
key
ally,
would
you
expect
that,
just
to
be
like
a
simple
pass
or
fail
filter
to
get
into
PLA
or
expect
more
things
to
be
in
there
from
that
or
kind
of
what
would
be
the
the.
C
C
G
G
E
G
This
presentation,
but
normally,
if
you
go
into
something
like
the
metrics
page,
oh
we
have
it
right
here,
I,
don't
this
is
going
to
work,
but
you
can
go
directly
to
the
dashboard
from
a
link
and
it
should
pop
up
and
show
you
your
your
dashboard
in
in
Cortana.
So
here
you
can
go
from.
You
know,
seeing
our
simple
graphs
that
we
have
display
to
going
to
the
Griffons
dashboard
and
seeing
more
information
in
here
I
view
them
I.
G
You
know
something
simple
to
look
at
and
get
things
done,
whereas
people
going
to
profound
I
might
want
more
configuration
options
so
from
our
key
ally
page,
you
can't
configure
what's
happening
with
the
dash
course.
We
want
to
add
in
something
else
or
modify
the
the
graph
anyway,
you
kind
of
can't
do
that
from
the
key
ally,
one
where
a
sink
refiner
you
would
be
able
to
so
I
think
it
might
be
useful
to
still
have
the
griffon
ax
dashboards
but
I
think
for
most
users.
G
It
probably
could
continue
to
use
key
ally
to
go
through
and
view
things
like.
There's
some
interesting
things
in
here.
We
can
do
things
like
see,
metrics
going
directly
from
different
versions
of
applications,
I'll,
just
not
a
good
one,
because
it's
coming
from
the
ingress,
but
look
at
something
like
ratings.
I
can
go
in
here
and
metrics
and
I
can
do
something
like
show
me
what's
talking
to
rating,
so
the
remote
app
is
reviews
and
I
can
show
like
what
version
is
coming
through.
So
I
can
see
the
metrics
just
for
the
1
and
v2.
G
So
that's
the
metrics.
For
me,
that's
for
the
three!
At
the
same
time,
you
can
also
do
things
like
if
it's
you
know,
have
multiple
versions
of
my
ratings
app
as
well.
I
can
see
that,
like
these
are
the
metrics
are
going
from
v1
of
ratings
to
be
true
of
services.
Same
thing,
this
is
the
one
of
ratings
to
be
3
of
reviews.
It's
kind
of
nice.
G
E
G
I
think,
like
the
idea
is
we
want
to
provide
enough
information,
get
someone
you
know
started
or
solve.
You
know
most
of
their
issues,
but
there's
going
to
be
use
cases
that
refine
is
going
to
be
better
for
fittings
for
workloads.
We
can
show
something
like
logs
in
there,
but
we
not
going
to
we're
just
getting
the
logs
from
the
tumor
net
east
side
of
things,
we're
not
getting
it
from
any
other.
You
know
log
aggregator,
but
we're
not
kind
of
like
replicate.
G
What
elasticsearch
is
doing,
for
instance
like
if
you
want
that
kind
of
functionality
should
go
to
other
service
or
time
to
provide
a
good
kind
of
default
view
of
things,
and
if
you
need
kind
of
this
more
power
user
things,
you
might
want
to
go
to
something
else,
kind
of
get
a
view
there.
But
for
you
know,
if
set
of
things,
we're
hoping
you
can
get
that
you
know
complete
without
any
problems
in
key
ally
itself.
G
Ok,
I
sent
one
more
slide
and
that's
it
I
know
it
was
brought
up
a
couple
years
ago
about
the
kelie
roadmap.
So
Cali
really
focuses
more
on
things
like
ethics
and
stories.
You
know
trying
to
make
things
easier
for
people
who
are
using
SEO
how
to
provide
some.
You
know
insight
to
what's
happening,
we're
hopefully
fairly
good
at
you
know,
getting
suggestions
from
the
community
into
key
ally
itself.
G
We
can
be
fairly
active
on
our
mailing
list
and
IRC
is
people
have
issues
to
go
there
and
ask
and
we
tend
to
go
through
and
get
those
into
calendars
pretty
quickly.
You
know,
as
part
of
this,
we
know
we
are
looking
for
feedback
and
working
more
of
the
east
of
community
to
get
or
features
and
functionalities
into
key
ally
and
to
see
where
we
can
go
with
that
in
terms
of
something
more
like
a
road
map,
we're
not
really
using
that
too
much
right.
G
Now
we
kind
of
look
at
things
more
in
three
weeks,
prints
which
people
should
be
able
to
see
from
our
JIRA
instance.
So,
at
the
start
of
the
three
weeks,
we
look
at
what
users
have
requested
or
what
issues
been
brought
up
prioritize
those
and
add
it
to
the
sprint
at
the
end
of
the
Sprint,
we
kind
of
redo
that
again
and
that's
how
we've
been
adding
in
future
functionality
Kelly
all
along
and
then
kind
of,
like
our
main
goals,
are
just
to
make
it
easy
to
manage
them.
G
Observe
was
half
things
in
a
service
mesh
and
to
also
you
know,
kind
of
expand
and
become
a
more
active
community
around
other
micro
service
observability
for
things.
So
that's
sort
of
our
roadmap
I,
don't
know
if
that's
helpful
to
users
or
if
it's
nope,
maybe
a
little
bit,
not
too
much
detail,
but
that's
mainly
how
we've
been
dealing
with
things
so
far.
Thank.
A
So
we
we're
running
a
little
low
on
time.
We've
been
thinking
about
talking
about
the
still
controls,
Michel
completions,
but
Dimmick.
Nothing,
quite
that
is
that
okay,
Eric
yeah,
that's
fine
I
mean
Jason,
wasn't
going
to
make
it
today
anyway,
so
moving
into
the
week
will
be
fine,
so
we
I
went
I
went
through
it
made
a
list
of
those
UX
items
that
have
issues
that
are
flagged
for
one
two
that
have
not
been
done,
don't
have.
Honors
I
was
hoping
to
see.
A
Configuring,
the
grifone
and
Prometheus
dashboards
to
respect
the
helm
settings
this
one
could
be
small,
maybe
it's
impossible
to
I'm,
not
sure
it
looks
pretty
easy,
so
maybe
somebody
should
check
it
out.
The
next
item
is
more
important.
I
think
it's,
this
sto
controllable,
install
item
happy
to
work
with
anybody
on
the
design
of
this
or
implementation,
is
using
this
to
your
control,
to
make
it
sort
of
easier
to
install
sto
in
the
way
that
we
think
maybe
link
or
D
or
other
service
meshes
are
take
less
work
to
install
some
of
the
profiles.
A
We
have
an
item
for
it.
Do
control
proxy
status.
I
think
we
have
an
idea
for
how
to
implement
it.
Proxy
status
has
not
worked
properly
on
the
multi
cluster
case,
and
this
one
is
for
proxy
status
and
proxy
config
to
have
them
deal
with
two
different
clusters.
At
a
time
one
cluster
running
pilot
and
one
cluster
running
kubernetes.
A
If
in
this
sort
of
split
horizon
yield,
we
have
a
documentation,
item,
I,
don't
know
if
we
have
Frank
here
to
document
at
a
level
of
the
attributes
that
are
being
produced,
and
then
we
have
an
old
item
that
I
believe
Sri
Ram
recently
tagged
for
one
to
to
improve
his
do:
control
validate
to
handle
kubernetes
items,
leap
kuat
is
making
it
not
complain
about
kubernetes
items
but
I
believe
sure.
Once
more,
he
requested
that
we
check
the
SEO
conventions.
A
A
E
Is
really
quickly
on
the
tubers
two
issues
it's
still
install
is
going
through
a
three
in
the
source
code.
We
just
moved
the
steel
modular,
install
our
farm
ecosystem
to
it
till
Ark
yesterday.
So
it's
going
to
be
rebased
on
the
new
thing,
so
it
needs
to
be
a
education
gone
through
the
new
thing
because
it
doesn't
make
sense
for
us
to
build
the
configuration
or
the
tooling,
based
on
the
older
version
of
over
install
four
one.
Two.
A
E
I
do
too
all
the
work
items
provided
for
one
to
do.
We
have
because
I
noticed
the
other
workgroup
has
been
doing.
Is
they
have
one
centralized
work
item
and
then
they
have
many
work
item
link
to
that
one
to
reflect
they'll
work
for
1.2.
Do
we
have
something
like
that?
Was
this
workgroup
to
kind
of
have
a
history?
A
A
good
point,
so
I
can
make
an
epic
for
all
the
work
that
has
been
tagged
as
one
two.
There
are
some
items
that
don't
have
a
version
or
have
nebulous
future,
and
we
should
the
steering
committee
and
this
workgroup
should
try
to
decide
what
features
we
want
to
get
in.
For
one
two
I
know
everyone
has
their
own
little
pet
projects,
but
I
want
to
make
sure
it
did
nothing
important
that
we
think
is
going
to
go
and
get
smashed
when
to
was
supposed
to
be
a
lot
about
code.