►
From YouTube: Istio UX Working group meeting June 25
Description
Istio User Experience Working group meeting June 25, 2019
A
A
A
So
the
first
item
is
some
post
sto
control
commands
that
we
want
to
make
it
easier
for
people
to
get
sort
of
information
about
their
mesh.
So
let
me
just
give
an
overview
of
the
stuff
I
want
to
talk
about
today.
Either
proposal
for
this
be
a
puddle
inspect.
I
want
to
talk
about
briefly
about
the
status
field
for
events
and
Ciardi,
which
has
been
talked
about
for
1-3,
but
isn't
happening,
and
maybe
we
can
do
some
talk
about
by
the
control,
plane
and
issues
or
controlling
that.
A
So
a
lot
of
this
work
is
is
about
sort
of
helping
the
user.
Other
pieces
are
more
about
manipulating
the
control
plane
to
help
the
user,
maybe
they're
in
the
realm
of
the
operator,
I'm,
not
sure.
So
that's
what
I
have
for
an
agenda
today,
I'm
going
to
leap
right
into
this
first
proposal.
The
inspect
proposal.
A
A
This
is
going
to
be
sort
of
a
little
sort
of
an
ad-hoc
command
that
looks
up
all
of
these
to
your
rules
maybe
makes
a
snapshot
of
them
or
just
equivalent
of
it
yeah
to
see
things
that
are
affecting
the
services
to
talk
into
that
pod
or
set.
So
imagining
that
you
could,
of
course,
you
inspect
pod
our
service
or
the
commercial
service
and
sort
of
give
you
some
of
the
kinds
of
things
that
you
see
with
queue.
A
Control
describe
the
name
in
the
namespace,
maybe
the
ports,
but
it
would
also
then
give
you
a
list
of
the
things
to
sort
of
are
operating
on
that
object,
so
it
might
get
ports
from
the
Kumud
engine
service.
Give
you
warnings
if
kubernetes
service
doesn't
follow
is
to
your
rules.
It
might
tell
you
if
any
cluster
nation
rules
tied
and
if
what
subsets
destination
rule
defines
it
might
tell
you
what
the
TLS
settings
in
the
destination
rule
were.
A
With
virtual
services
talk
to
it,
it
would
break
down
the
virtual
service
more
telling
you
match,
rules
the
way,
and
you
read
a
regular
interval.
Maybe
it
would
ask
the
pot
itself
what
envoy
version
is
there?
You
give
a
summary
of
the
stuff
that
you
get
from
sto
control,
proxy
config
and
maybe
give
you
a
summary
of
what
you
see
missed
your
control
proxies
Pettus,
a
human-readable
summary
of
what
is
tio
thinks
your
pod
state
should
be
in
so.
C
This
is
great
for
debugging
and
getting
information
from
the
configuration
point
of
view.
The
one
thing
that
you
talked
about,
which
is
that
it
is
critical
to
understand,
is
that
you
are
trying
to
recreate
this
information
outside
of
the
pilot.
So
you
are
trying
to
look
at
the
sto
configuration
and
decide
which
of
these
configurations
affect
the
pod.
Is
there
a
way
to
not
do
it
aside
from
pilot,
because
as
soon
as
people
change
functionality
in
pilot,
what
will
happen
is
this
will
start
to
lag
so.
C
There
is
a
way
of
doing
this,
also
from
pilot
basically,
so
this
is
a
new
proposal.
Basically,
if
on
work
configuration
allows
a
slot
for
adding
arbitrary
annotations,
then
you
can
add
that
this
configuration
for
envoy
is
actually
coming
for
this
HT
or
high-level
config,
and
then
you
can
use
those
envoy
configuration
tags
to
generate
this.
So
basically,
the
idea
is
su,
sitio
should
always
inspect
on
web
configuration
and
on
word,
configuration
should
have
should
have
back
references
to
all
the
user
project
config
if
that
is
possible.
So.
A
C
A
That
would
be
completely
ideal.
So
if
that
was
the
case
in
a
great
amount
of
this
work,
when
the
pot
is,
the
target
would
be
parsing.
That
information
which
is
do
control
already
knows
how
to
get
from
envoy
and
just
parsing
the
JSON,
and
that
would
give
us
at
least
the
names
of
the
destination
rules
and
the
virtual
services,
and
then
whether
or
not
we
looked
into
them
would
be
that
maybe
the
first
version
wouldn't
look
in.
Maybe
the
future
version
could
look
in
see
for
that
click
rate.
A
A
A
A
So
the
the
next
thing
about
that
I
guess
you're,
just
a
right
one,
to
ask
so
I
propose
that
we
should
put
warnings
on
if
the.
If
the
kubernetes
items
are
up,
aren't
sort
of
correct
and
I've
seen
several
proposals
for
how
a
CEO
should
do
this
I've
heard
people
saying
that
is
to
itself
should
be
doing
this
or
we
should
be
using
code
from
Aspen
meshes
is
still
that.
A
So
currently
we
see
Oh
cuddle
has
a
validate,
but
it
only
validates
misty
ozone,
CR
DS.
It
doesn't
give
you
best
practices
if
you
pass
in
a
kubernetes
object,
such
as
a
service
to
tell
you
to
follow
the
name
convention
and,
furthermore,
I've
heard
we
want
to
get
rid
of
explicit
names.
There's
a
work
item
to
get
rid
of
explicit
names
on
services
as
well,
so
I
wouldn't
want
to
bake
in
a
whole
separate
checker
for
this
into
the
command
line.
If
that's
going
away
in
the
core
viseo.
C
And
it
feels
like
from
the
networking
group
meetings
it's
gonna
go
away,
like
you
said
additionally,
I.
Think
if
you
keep
the
scope
of
sto
CTL
inspect
narrow,
where
it
just
tells
you
the
configuration
which
is
currently
affecting
on
web
configuration
in
a
part.
I
think
you
don't
need
to
do
this.
I
yeah.
A
So
I
I
agree
that
one
possibility
is
to
make
this
as
narrow
as
possible,
just
to
tell
you
what
it
affects
it
and
then
for
things
like
this
or
things
like
this
you're
expected
to
go
deeper,
I'm,
not
sure
what
users
sort
of
want.
So,
if
I
make
a
initial
version
that
just
tells
you
what
affects
your
pod,
then
it'll
help
because
I
mean
a
lot
of
this
is
inspired
by
conversations.
A
I've
had
on
discussed
on
other
forums,
helping
people
debug
their
situation,
I'm
always
well,
let's
see
if
we
have
any
sto
rules,
but
this
you
know,
Cube
control,
get
vias,
comedy
are
and
all
kinds
of
things
so
having
a
single
stop
for
a
certain
pod
seemed
like
the
best
way
to
do
it.
So
our
people
on
this
call
favoring
a
narrow
approach
where
we
don't
show
a
lot
of
details,
but
instead
just
show
things
like
this
and
this
or
do
we
want
to
show
more
information?
A
B
C
B
So
the
hard
thing
is,
is
your
specific
configuration,
especially
you
are
ominous
as
a
sto,
CTL
command,
I,
think
the
so
the
name
and
the
namespace
should
be
given
to
user
just
to
help
them
unique,
identify
the
same
right
and
then
the
other
thing
ideally
will
be.
You
use
a
steel,
specific
I
think
that
will
be
really
cool
and
on
this
list
this
is
only
like
two
or
three.
That's
not
is
still
related
other
than
the
name
and
namespace.
A
Alright
well
I'm,
seeing
enough
support
for
this.
That
I'll
maybe
do
an
implementation,
or
we
can
work
on
one
together
if
I
knew
it
on
the
skull
sort
of
wants
to
get
some
early
feedback
on
what
it
would
look
like
to
have
such
a
command
I
think
a
command
like
this
is
going
to
help
people
who
are
debugging.
Our
current
is
geo
coded
debugging
tools
like
proxy
configure
to
low
level
and
sort
of
this
is
supposed
to
fit
in
the
middle
level
between
those
tools
and
something
that's.
B
Yeah
there's
also
a
nice
document,
I
think
I'll
say
you
might
own
that.
Why
I
wonder
how
these
two
can
be
used
conjunction
by
the
user,
because
I
almost
feel
like
when
I
inspect
a
path.
I
would
also
want
to
see
the
status
for
each
of
the
ischial
resources
and
that
I
know
which
one
of
them
have
being
applied.
B
D
E
I
think
that
there's
also
the
performance
aspect
of
it
right.
So
the
galley
is
a
Cuban,
a
piece
controller
and
it
will
need
to
be
acting
as
a
controller.
Many
things
for
Easter.
If
you
make
it
update
the
status
of
pod,
it
becomes
a
controller
for
pause
as
well,
and
that
made
them
not
to
be
like
a
costly
thing
to
do.
From
a
CPU
or
memory
usage
perspective,
I.
C
A
A
What
I,
in
and
sort
of
the
original
thought
that
people
had
was
that
we
could
use
this
for
several
things
for
ISTE
ozone
CRD,
so
one
thing
we
want
to
do
is
when
we
make
a
new
virtual
service,
we
want
to
know
we
did
sort
of
reach
the
pods.
So
that's
sort
of
a
status
of
have
has
this
rule
been
accepted
and
pushed
and
can
be
tested
that
would
make
our
own
tests,
for
example,
go
faster.
A
There
is
the
idea
of,
is
this
rule
globally
valid
so
emission
control
validated,
but
it
was
locally
valid,
but
it
didn't
validate,
for
example,
that
it
refers
to
a
subset
that
a
destination
rule
defines
so
with
the
new
code.
In
this
pull
request,
there's
like
a
snapshot
taken
and
galley
memory
of
all
the
pieces
and
they
validate
everything
and
they
put
it
all
together
and
that
status
could
easily
be
put
on
misty
ozone,
CR
DS.
It
could
also
be
put
on
other
CR,
dr
on
communities,
objects
itself
like
on
pods,
but
events
can
also
appear.
A
I,
don't
have
handy
a
screenshot
for
this,
but
it's
easy
to
create
events
that
appear
went
into
a
cube,
control
described
pod.
You
could
have
an
event
that
just
sort
of
says
you
know
the
this
pod
had
a
non-void
that
rejected
a
configuration
from
pilot.
That
would
be
super
huge
and
I
definitely
want
pilot
to
start
doing
that
and
I.
E
Think
they're
a
couple
of
professional
pitfalls.
There,
though
I
just
want
to
be
a
going
to
make
sure
that
we
are
kind
of
approaching.
This
is
a
problem
in
the
right
way
right.
So
the
first
thing
is:
there's
a
push
for
making
pilot
agnostic
of
kubernetes
right,
so
the
pilot
will
not
be
able
to
talk
to
the
status
of
the
pods
directly.
E
E
So
I
think
we
definitely
should
expose
this
information
and
I
believe
there
is
already
events
like
some
telemetry
exposed
and
we
can
also
do
community
events
which
I
think
is
cheap
enough
to
accomplish
I'm
wondering
if
we
should
also
target
status
as
well.
So
is
this
something
that
yeah
I'm
just
kind
of
thinking
out
loud
here
right,
so
I'm,
just
trying
to
figure
out
what
is
the
best
way
for
a
user
to
figure
this
one
out
like
I,
was
seeing
that
that
particular
nya
is
having
troubles
so.
F
A
The
meetup,
a
fellow
told
me
that
his
biggest
problem
running,
is
to
your
production
is
when
that
configuration
causes
pods
to
start
rejecting,
XTS
I
mean
it
makes
your
whole
mash
feel
like
it's
frozen,
because
nothing
you
do
has
any
effects,
because
some
other
things
you
did
earlier
is
making
things
go
away.
Whenever
there
was
a
reject.
If
we
sent
out
that's
that
status,
then
we
would
as
an
event
or
as
a
status,
we
would
sort
of
see.
You
know
it.
This
event
started
happening
at
this
time
we
related
to.
When
we
applied
some
rule.
A
B
C
D
D
I
mean
we
should
have
dashboards
and
a
like
a
trial
user
that
it
doesn't
use
dashboards
shouldn't
be
worrying
about
this
right,
like
it's,
not
a
common
case
where
you
get
a
XDS,
reject,
that's
more
like
someone,
that's
like
a
power
user
who
is
certainly
going
to
have
dashboards
and
alerting,
and
things
like
that.
So.
A
The
gentleman
I
met
at
the
meetup
told
me.
It
often
takes
them
a
day
to
figure
out
a
problem
that
like
this,
because
it
is
deep
and
we
are
reducing
the
number
of
excess
rejects.
We
have.
My
hope
would
be
that
if
we
put
the
code
in
it's
never
executed
because
pilot
always
fixes
up
things
before
it
passes
the
monta
envoy.
A
D
But
what
I?
What
I'm
thinking
is
that
so
pilots
goal
is
to
not
have
config
rejected,
and
so
in
doing
that.
Sometimes
that
means
that
internally
pilot
is
rejecting
can
fix
like
we
would
have
a
gateway
config.
We
would
drop
it
and
say
this
config
is
going
to
generate
invalid
config
and
now
that
gateways
not
tied
to
a
pod,
so
we
shouldn't
attach
it
to
the
pod
status.
It
should
be
on
the
Gateway
status
thing
directly.
A
E
That's
going
to
be
redirected
by
an
animal
right,
so
that's
more
of
a
easier
control
plane
in
distress
problem
rather
than
user
configuration
being
wrong
problem
and
I
think
an
approach
for
that
category
of
problems
in
general
might
be
more
interesting.
Where
you
know
we
I
don't
set
up
a
some
sort
of
standard,
alerting
out-of-the-box
for
a
you
know,
conference
control
being
in
distress
which
might
be
much
more
interesting,
I
would.
D
Agree,
I
would
be
worried
that
if
so
say
they
get
a
rejection,
there
should
be
an
alarm
that
goes
off
and
actively
tells
them
like.
There
was
an
X
just
reject
they
shouldn't
have
to
go
and
keeps
detail
describe
the
pod
because
they
may
not
even
know
their
traffic
may
be
functionally
fine
and
it
may
be
not
getting
updates
and
they
start
seeing
more
and
more
bizarre
errors
and
then,
finally,
you
know
five
hours
later
they
go
and
describe
the
pot
and
see
that
it
was
rejected.
D
D
B
So
along
will
be
useful
for
operating
environment,
but
on.
If
you
look
at
the
developer,
experience
I
search,
try
to
add
their
services
to
the
match
pop.
Now
the
new
set
of
an
alarm
and
I
made
you
one
services
time
so
I
would
be
equal
to
check
out
what
is
going
wrong
with
everything
like
services,
so.
C
Soul
in
in
that
case,
you
want
to
know
what
this
tio
configuration
is
affecting
my
sellers
right,
so
I
I
understand
so
I
don't
feel
like
that
support
status,
give
tools,
listen
Eid
has
proposed
where
they
can
use
the
tools
to
understand
why
their
pod
is
working
or
not,
and
for
the
failure
cases
where,
for
some
reason,
the
control
plane
rejects
a
config
or
the
control
plate
cannot
send
down
a
conflict
envoy
I
think
those
are
different
things.
Those.
B
B
I
could
bet
I
personally
hate
that
sometimes
sometimes
the
solutions
you
have
to
restart
the
pilot
that
you
have
to
go
through
proxy
config,
proxy
debug
or
the
the
two
commands
it's
actually
really
hard
to
use.
Sometimes
you
have
to
go
through
admin,
API,
so
I
think
we
want
to
hide
user
form
that
complexity,
so
they
can
just
using
maybe
one
command
to
be
able
to
tell
them
what
could
go
wrong,
whether
they
contact
a
skirmish
was
supported
during
a
supporting
contract
with
you
or
whether
this
is
something
that's
wrong
with
their
services
themselves.
B
A
A
Status
and
events
related
to
this
virtual
service
so
trying
to
parse
what
was
in
the
virtual
service
and
understand
it.
Looking
at
the
total
boss,
I
could
just
ask
kubernetes
give
me
all
events
related
to
this
virtual
circuit
status
field
and
I'm
going
to
display
that
so
this
my
cube,
control
describe
it
gives
the
status
field
of
kubernetes
objects
is
Tioga
to
inspect,
would
give
the
status
field
of
objects
that
influence
and
it
commands
object
and.
C
C
A
A
A
E
Would
it
make
sense
to
look
at
pilots
UN
strictly
so
you
know,
pilot
exposures,
like
tonometry,
is
stuff
right,
so
Easter
cuddle
can
actually
scrape
the
telemetry
and
then
create
you
pods,
so
I
think
so.
Okay
I
think
we
need
to
separate
the
scenarios
right.
So
one
scenario
is
the
case
where
there
is
in
production,
great
deployment
with
its
own
operator,
and
then
it's
on
reporting
and
alerting
right.
So
that
scenario
I
think,
is
already
covered.
From
the
perspective
of
my
rejecting
configuration.
E
I
think
the
scenario
brought
up
is
about
this
case,
where
I
do
iterative
development
using
some
configurations
locally
and,
like
you
know,
trying
to
debug
things
and
understand
things
and
I
somehow
run
into
this
case,
where
the
in
what
the
control
plane
is
in
distress
right.
So
one
of
the
things
that
I
can
do
is
I
can
use
eto
cuddle
to
go
and
query
like
the
control,
plane,
components
you
know
pilot
or
galley
and
see
if
they
are
in
distress
by
looking
at
their
telemetry.
C
E
C
Think
that
is
pretty
good
because
and
then,
if
you're
still
a
little
inspect
can
actually
give
you
both
the
outputs.
If
you
want
it
can
give
you
the
output
of
what
configuration
is
currently
affecting
your
board
and
if
the
control
plane
is
distressed,
it
can
just
show
you
the
output
of
the
event
stream
from
pilot.
A
So
I
think
we
have
some
good
thoughts
on
on
this.
A
lot
of
this
is
not
stuff
that
we
need
to
implement
I.
Just
sometimes
I
want
to
make
the
other
people
on
this
group
aware
of
what
I'm
seeing
on
the
other
working
groups.
One
item
that
I
did
want
to
discuss.
It's
just
a
pure
issue
is
Cube
control.
Explain
there
is
apparently
a
lot
of
stuff
in
kubernetes
1/4.
That's
going
to
allow
cube
control,
explain
to
work
better
for
CRTs.
A
If
anyone
wants
to
take
over
this
item,
it
would
be
great
I
tried
and
failed
on
earlier
versions
of
kubernetes.
To
get
this
end,
it
would
be
great
it
to
automatically
transform
our
protobufs
into
open
API
specifications
and
put
them
into
our
CRTs.
I
was
able
to
do
that
I
was
it
put
the
schema
into
the
CR
D,
so
you
did
cube,
which
will
explain.
A
It
would
show
you
the
schema
and
product
could
or
kubernetes
api
could
be
tilled
to
validate
the
series
according
to
that
schema,
but
I
couldn't
find
any
way
to
get
the
text
to
be
shown
by
kubernetes
I,
don't
know
if
that's
my
skills
of
kubernetes
or,
if
that's
a
limitation
with
1:14.
If
anybody
wants
to
take
over
this
item,
though
I
would
be
greatly
appreciated.
A
E
This
something
that's
related
to
this
in
config
work,
group,
Jason
banging
on
our
site,
actually
started
doing
some
tooling.
Work
correctly
can
generate
open,
API
schema
from
our
proto's,
and
this
is
useful
for
adding
validation
at
the
API
server
level
and
doing
some
limited
version
transformations
through
API
server
itself.
So
our
goal
is
to
actually
get
these
open,
API
schemas
into
cid
definitions
at
some
point.
E
A
Put
them
in
and
I
couldn't
and
it
it
the
schema.
Was
there
but
I
either
I
did
it
incorrectly
or
it
wasn't
supported,
I
couldn't
I
couldn't
get
the
text
in
I
really
really
feel
users
want
to
build
into
cube
control
explained
on
his
TOC
or
DS.
So
if
the
config
workgroup
could
take
over
this
issue,
in
fact,
let
me
tag
it
right
now
as
a
possible
config
item.
Oh
it!
No,
it
is
so.
If
anyone
wants
to
think
about
this,
what
I
think
it
should
be
a
our
priority
for
helping
people
understand
the
system.
E
Okay,
yeah
so
I
I'm.
What
exactly
should
cuddle
explained
us
with
the
schema
but
I
suspect
I
think
this
command
can
help,
and
then
this,
the
open
API
schema
that
the
API
server
supports
is
actually
a
subset
of
the
full
definition,
rightful
spike
and
I.
Think
up
until
recently,
they
weren't
even
supporting
dollar
Rev,
and
there
are
certain
limitations
like
that
that
API
service
team,
toodaloo
support.
E
A
A
So
I
started
working
on
some
of
the
control
plane
settings
that
users
commonly
need
to
adjust
I
had
a
bug
and
I
needed
to
take
a
look
at
the
or
I
wanted
to
take
a
look
at
the
access
logs
that
envoy
can
sort
of
make
it
I
really
had
difficulty.
So
the
first
thing
I
went
to
our
documentation
for
access
logs
or
for
troubleshooting.
If
you
look
at
them,
they
sort
of
they
sort
of
say
the
logging
is
on
by
default
and
that
it's
easy
to
set
logging
up
and
that
you
should.
A
You
should
control
the
foot.
You
should
set
access
log
format,
but
the
sort
of
link
to
format
it
was
very
confusing.
I
have
been
seriously
thinking
that
we
need
some
kind
of
commands,
either
in
the
operator
or
SEO
cuttle.
For
these
operations,
like
turning
on
MPLS,
seeing
if
MPLS
is
enabled
turning
on
envoy
access
logging
adjusting
the
access
logging
thing,
we
hide
a
JSON
within
a
config
map.
Is
we
had
Yamma
within
the
config,
which
is
in
JSON
confusing
difficult
for
users
to
play
with
this
stuff?
What
do
we
think
I.
D
A
It
is
there
so
that
you
can
do
it
through
hell.
Maybe
I
just
have
the
wrong
idea.
Whenever
I
install
this
to
you
with
hell
my
dimension
myself
as
sort
of
reinstalling
sto
I,
never
know
if
it's
going
to
restart
my
pods
for
the
control
plane
or,
if
I'm
running
a
custom
daily
build.
It's
going
to
put
me
back
to
some
other
build
that
I
may
be
install
with
the
template.
I
found
it
confusing
to
think
that
I
have
to
reinstall
on
way
to
turn
logging
on
and
off
yep.
We
want
to
do
it
so.
D
A
Right,
so
it
obviously
if
they
wanted
on
forever,
they
installed
it
on.
However,
first
I
user
might
be
sharing
their
control
plan
with
a
bunch
of
other
things.
Maybe
I
just
want
to
turn
it
on
for
the
stuff
in
my
namespace
that
my
right
tribe
is
maintaining
and
I
don't
have
access
to
the
control
plane
or
maybe
I
just
want
to
turn
it
on,
because
I
have
a
bug
and
then
I
want
to
turn
it
off.
Okay,.
D
A
C
Add
this
feels
like
a
knob
in
the
sidecar
resource
and
I'm,
saying
that
with
a
bit
of
skepticism
but
I,
think
that
has
a
proposal
out
there
for
configuring
or
fine-tuning
envoy
tracing
the
other
sidecar
resource,
if
the,
if,
if
that
is
available
for
tracing
I,
think
it
should
be
fair
game
to
say
for
a
specific
board:
I
want
to
turn
on
access
logging
and
then
disable
it
later.
I
think
I,
don't
know.
Who
else
from
networking
meeting
is
here?
I
think
John
can
answer
if
that
seems
reasonable.
That
makes.
A
F
A
No
samples
of
them
in
our
samples
to
hear
maybe
there's
none,
I
I
started
making
a
Hjelm
template,
so
I
could
make
a
sidecar
resource
because
they
were
too
hard
to
write
by
hand.
Would
it
be
nice
to
have
built
into
is
do
Codel
some
way
to
make
a
sidecar
or
some
way
to
manipulate
fields
in
the
sidecar,
like
this
field
were
proposing,
adding.
B
In
fact,
I
was
having
a
conversation
with
Louise
it's
a
few
days
ago,
to
not
specifically
a
psychologist,
but
specifically
on
some
of
the
English
configurations,
and
we
use
it
to
use
these
co-ceo
to
simplify
it.
So
I
think
it's
along
the
same
line,
but
not
so
much
I'm
speaking
whether
cyclist
is
stable
enough
for
us
to
build
into
it
still
CTL,
but
I.
Think
it's
good
to
look
at.
It
still
see
TR
how
it
can
help
user
to
simplify
some
of
the
configurations
so
yellow,
instead
of
using
yellow
just
using
it
still
CTL
command.
B
A
A
A
A
C
On
and
off
and
then
giving
service
owner
the
right
control
to
control
what
they
want
at
the
pod
level.
So,
however,
is
to
use
philosophy.
Is
that
at
the
pod
levels
you
can
figure
this
to
manipulate
things
for
the
Envoy
sidecar,
that's
what
it
should
live.
Currently,
it
looks
like
sidecar
is
that
resource
and
I
agree
it's
very
hard
to
use
it.
A
C
A
D
Think
I
think
one
thing
that
we
may
want
to
look
into
is
I.
Think
if
you're
not
familiar
linker
D
has
this
linker
DTaP
command,
which
I've
never
used
it
by
I.
Believe
it
just
like
it's
like
the
live
stream
of
requests
going
into
a
pod.
So
you
just
kind
of
like
tap
into
the
log
stream
and
envoy
recently
added
a
similar
tap
mechanism
that
we
don't
configure
it
on
East
EO.
So
maybe
that's
something
we
could
use
I'm,
not
sure
I
haven't
looked
too
much
into
it,
but
some
things
you
consider
I
think.
B
So
one
thing
about
how
now
we
were
in
the
meeting
earlier
today,
there
was
a
lot
of
concern
if
we
tie
too
close
to
home,
which
means
we
try
to
close
to
kubernetes.
At
the
end
of
the
day,
a
lot
of
people
are
looking
at
it
still
bianca
Bonetti.
So
we
need
to
be
careful
with
commands
that
try
to
close
to
hell.
A
See
that's
a
good
point
Lynne,
but
my
solution
was
no
better
because
I
just
want
to
use
the
kubernetes
api
into
a
patch
on
resources.
What
I've
never
used
sto
outside
of
kubernetes?
How
do
we
do
things
like
nipple,
8,
CR
DS,
without
cube
cuddle
patch
I.
A
So
I
know
a
lot
of
people
here
like
pipeline
approaches
and
they
expect
things
to
be
done
by
pipelines
and
stuff
I'm.
A
big
fan
of
like
one-liner,
crazy
admin,
monkey
type
stuff
working
so
I
have
lots
of
little
snippets
like
I
inject
and
remove
sidecar
annotations
with
cuddle
patch,
and
this
sort
of
thing
I'm
always
trying
to
see
how
I
can
quickly
get
a
lot
of
functions,
happening
and
I.
Just
don't
know
how
that
works.
If
I'm
not
using
kubernetes.
A
So
I'm
I'm
doing
cube
cuddle
patch
on
you,
know
the
ISTE
marketing
apps
and
on
the
sto
mesh
config
and
on
pod
annotations
and
all
kinds
of
things
that
make
certain
people
flutter
people
tell
me
no,
no
a
tree
install
with
with
helm
and
I'm
like
no
look
I'm
just
going
to
patch
this
object,
so
one-liner
takes
10
seconds.
So
we
want
to
I
guess
support
both
use
cases
can.
B
Yeah
yeah
I
just
did
I
just
didn't
want
to
tie
to
helm,
especially
we
went
to
helm
with
installation
and
now
we're
moving
kind
of
away
from
hell,
and
we
start
to
hear
more
and
more
people
that
is
still
just
for
queue,
so
we're
going
to
probably
spend
more
efforts
to
make
it
work.
Young
Cuban,
Eddie,
oh.
A
That's
what
I
was
thinking
too
so
I'm
going
to
have
a
meeting
on
Thursday
about
the
operator
and
the
new
Installer
and
how
they
work
together
and
how
we
can
get
the
functions
of
the
new
into
Sto,
cuddle
and
other
command
lines.
How
how
would
someone
with
console
be
installing
is
tio
with
two
control
planes
or
something
like
that.
F
F
A
B
B
A
C
B
D
C
B
B
B
D
C
F
A
A
Hopefully,
at
the
next
meeting,
Romana
will
talk
more
about
specific
items
related
to
the
sort
of
installation
and
of
mesh
and
upgrading
of
the
mesh,
so
we
saw
we've
been
frustrated
by
it
and
we
have
very
few
work
items
for
how
to
improve
the
user.
Experience
of
installing
and
upgrading
in
helmet
is
sort
of
great,
but
we
have
these
great
new
installers
and
it
there's
not
a
lot
of
discussion
about
how
a
user
does
that
interaction.
A
We've
seen
the
architecture,
you
know
how
it
makes
to
control
planes
and
does
it
canary
from
the
old
version
to
the
new,
which
looks
and
feels
super
cool.
But
now
what
happens
if
halfway
through
that
fails
and
that
sort
of
thing
so
I've
been
trying
to
understand
that
so
hopefully
that
will
give
us
some
things
to
talk
about
in
two
weeks.
Do
we
have
any
other
sort
of
agenda
items
that
people
on
this
call
want
to
present
in
two
weeks.