►
From YouTube: Meshery Development Meeting (Oct. 7th, 2020)
Description
All about Meshery Release Channels.
A
C
A
A
Okay,
so
first
topic
in
the
agenda,
we
have
to
prefer
speed
ups
this
month
because
it
is
hack
to
profess
for
october.
For
anybody
who
doesn't
know
how
to
prefest
is
a
month-long
celebration
of
open
source
where
you
basically
make
four
pairs
and
when
this
cool
t-shirt
or
have
a
tree
planted
in
your
name
and
get
to
learn
a
lot,
hopefully
so
for
hacker
fest
tutorial
meetups,
we
are
doing
tutorials.
A
A
D
E
A
D
C
There's
a
crash
and
burn
with
each
each
of
the
names
I
think
clement.
I
I
tossed
that
in
there
a
little
bit
late
for
him.
I
think
it's
probably
I'm
assuming
it's
early,
that's
why
he's
not
on
he
clement
did
put
forth
a
design
recently
on
letting
people
switch
between
measure,
release
channels,
the
stable
release
channel
and
the
edge
release
channel
and
how
that
would
look
like
in
measuring
cpl,
like
letting
people
switch
between
the
channels
using
the
command
line.
C
C
C
So
if
you
can
see
my
terminal,
you
can
see,
I
was
doing
some
work
yesterday
on
nginx's
service
mesh
go
that's
so
by
the
way,
sneak
peek
and
kind
of
a
good
topic
for
the
community
call
on
friday.
We'll
talk
about
nginx's
service
mesh
a
bit
all
right,
if
you're
a
messer
user,
then-
and
if
you
go
on
to
the
system
that
you've
run
meshri
on,
you
will
find
that
in
your
users,
folder
there's
a
hidden
folder
in
your
home,
folder
there's
a
hidden
folder
for
measuring.
C
So
if
you,
by
the
way,
if
you
check
out
your
home
folder
and
you
look
at
all
the
hidden
folders
in
there,
you'll
likely
you'll
likely
find
a
lot
of
hidden
folders.
The
hidden
folders
are
denoted
denoted
with
a
period
in
front
of
them
dot
and
mesh
re,
like
many
other
applications,
are
kind
of
following
the
convention
of
stuffing
a
little
bit
of
application
configuration
into
that
hidden
folder.
C
And
so,
if
we
go
into
the
the
mesherie
folder,
we
take
a
look
around
we'll
find
a
couple
of
things.
One
will
find
a
local
database.
That's
used
as
a
cache,
we'll
talk
about.
That's
not
the
we'll
talk
about
that
some
other
time,
but
the
other
thing
that
you'll
find
is
you'll
you'll,
most
notably
find
mesh
yaml,
or
at
least
you
should
take.
If
you
take
a
gander
at
this,
well
you'll
find
that
it's
it's
a
super
thinly
veiled,
docker
compose
file.
C
It
is
a
docker
compose
file,
that's
what
measuring.yaml
is
at
least
for
now,
and
so,
when
you
look
through
the
file
and
get
familiar
you'll,
hopefully
readily
understand
part
of
what
probably
aniruddh
and
and
the
rest
of
us
might
talk
about
with
respect
to
mesri's
architecture
and
how
mesherie
has
a
server
that
has
a
ui
as
a
command
line.
Client,
it
has
adapters.
C
You
know
eight
different
adapters
today,
tomorrow
or
in
the
future
there
will
be
a
meshery
kubernetes
operator.
It's
called
measuring
operator,
there's
a
repository
out
there
that
you
can
go
check
out
if
you
like.
There
are
some
of
you
that
are
not
on
the
call
right
now
that
have
been
working
on
mesh
sync,
which
is
a
custom
kubernetes
controller.
C
It's
all
it's
all
out
there
and
it's
available
to
read,
but
but
we're
working
on
the
docs
to
improve
them
for
this
part
for
measuring
for
any
piece
of
software
when
people
want
to
download
it
and
use
it.
There's
always
some
friction
between
trying
to
make
sure
that
they
they're
running,
something
that's
stable
and
doesn't
have
you
know,
is,
is
of
high
quality,
but
also
trying
to
get
them.
C
The
latest
features
you
know
sooner
than
later,
and
so
for
meshary
the
path
that
it's
following
is
in
establishing
a
release
channel
by
the
way,
the
this
it's
not
like.
What
I'm
saying
here
is
undocumented.
There's
long
design,
specs
called
the
measuring,
build
and
release
strategy
that
outline
what
these
two
channels
are
and
how
they're
used
and
the
mechanics
by
which
they've
been
created,
I'm
going
to
show
part
of
their
mechanics
right
now.
C
There's
also,
I
think,
a
docs
page
on
on
this
right.
Now,
though,
these
two
release
channels
every
time
that
someone
merges
a
pull
request
into
the
meshery
master
branch,
a
new
edge
release
is
performed,
meaning
the
artifacts
of
the
temporary
build.
That's
done
are
those
artifacts
that
build
measury,
server,
messaging
adapters
and
things
mastery
ctl.
Those
are
persisted
to
docker
hub
anything,
that's
created
in
a
container,
it's
persistent
to
docker
hub
for
this
project,
and
if
it's
it
each
and
every
time
that
a
pr
is
merged,
a
new
edge
release
is
formed.
C
There's
a
public-facing
repository.
It's
it's
under
this
org
it's
under
this
container
repository
name
and
each
of
the
images
that
get
pushed
up
if
it's
a
stable
release
and
a
stable
release
means
that
it's
one
of
those.
You
know
what
let
me
share
my
whole
screen
because
we're
talking
about
a
lot
of
things.
C
C
C
Docker
hub,
you
can
find
the
same
repository
that
I
was
just
talking
about
quickly.
You
can
find
it
by
kind
of
clicking
on
that
docker
pulls
so
here's
the
public
facing
repo.
If
you
go,
look
at
the
tags
for
this
image,
you'll
find
that
there
are
some
that
have
us
that
say:
stable,
there's,
one,
the
only
one,
and
only
ever
one
that
says
stable
latest.
C
There
are
others
that
say:
edge
and
they'll
have
a
commit
hash
or
part
of
a
commit
hash.
Others
will
have
like
stable
in
a
commit
hash
or
stable
and
the
version
number,
and
so
it's
these
and
here's
the
an
earlier
version.
It's
these
numbers
and
these
two
release
channels
a
stable
release,
channel
and
an
edge
that
control
how
control,
how
quickly
new
features
get
into
the
hands
of
users,
every
time
that
a
user
downloads
and
runs
meshri
by
default.
C
C
It
actually
says
well:
you're
subscribe,
you're
subscribed
to
the
the
stable
channel
and,
as
such
you
will
you
will
deploy
when
you
deploy
messaging,
you
will
deploy
the
image
the
container
image,
that's
associated
with
stable
latest
with
that
tag,
and
the
notion
that
you're
subscribed
is
facility,
okay
and
is
facilitated
by
another
container.
That's
running
in
your
environment
that
that
other
container
is
described.
C
C
Should
I
be
paying
attention
to
which
one
should
I
go
out
and
check
for
a
later
version
of,
and
when
I
find
a
later
version,
it
will
automatically
download
that
later
container
image
and
reload
reload
that
image,
so
so
people
when
we
make
a
stable
release,
those
that
are
running
meshry
under
stable
under
this
default
configuration
they'll,
be
auto,
updated,
so
merging
a
pr
can
be.
You
know
we
want
to
have.
C
Yeah
this
is
great
and
actually
that's
a
great
question.
Poor
and
the
answer
is.
The
answer
is
that,
yes,
this
is
something
that
users
can
control,
but
which
is
a
good
answer.
The
where
we
would
like,
for
that
answer
to
be
better
is
to
make
it
convenient
for
users
to
control
by
way
of
measuring
ctl
to
to
let
them
do
something
like
measuring
ctl
system.
Config
update
frequency
like
like
th
this
flag
right
here,
is
something
that
I'm
making
up
like
like
we
don't.
C
A
A
You
know,
docker
hub
that
if
a
new
image
is
created-
and
it
continuously
does
that,
so
I
I
personally
think
that
it's
not
that
of
a
fish
efficient,
but
I
was
also
not
able
to
find
it
is
anybody
using
watchtower
tower
have
have
done
it
before
that?
Every
time
when
there
is
a
new
image
created,
pulls
the
image
or
creates
a
container.
Only
then
like
yeah,
is
there
a
way
to
control
watchtower
that
ways.
C
B
C
C
I
want
to
pin
to
stable
v04.11,
and
that
would
be
a
way
and
then,
at
that
point
like,
if
I
pin
to
a
version
well
then
like
hey,
I
can
just
disable
watchtower.
I
can
just
either
remove
the
labels
or
remove
watchtower
or
both,
and
so
when
I
go
to
start
measuring
I'll
just
say:
hey,
don't
here's
the
config
that
I
want
you
to
use,
but
first
of
all
I
could
just
go
change
meschery.yaml
itself,
the
default
config.
C
C
C
One
of
the
things
that
we've
been
able
to
provide
so
far
is
a
flag
on
most
of
the
system
commands
because
the
system
commands
that
the
the
set
of
commands
under
system
that's
really
about
controlling,
measuring
not
about
controlling
service
meshes
and
other
things.
That
measure
does.
If
we
look
at
the
system
command,
it's
it's
this.
It's
like
start
messaging,
stop
measuring,
tell
me
the
status
semester.
You
get
the
message
logs,
but
one
of
those
commands
is
to
reset
reset
measures
configuration
reset
that
file
that
I
just
screwed
up.
C
C
What
should
happen?
Reset
message?
Configuration
file
to
default
settings
is
that-
and
maybe
this
is
a
bug
for
this
flag-
is
that
messy
ctl
will
reach
out
to
github
grab
the
publicly
available
docker
compose
the
publicly
available
measuring.yaml,
the
publicly
available
docker
compose,
that
is
spelled
correctly,
pull
it
down
overwrite
the
local
file
and
then
go
from
there.
C
C
C
Anyway,
a
little
bit
of
a
you
know,
there's
more
questions
as
you
think
on
this
problem.
Space
there's
like
well,
there's
there's
like
a
lot
of
stuff
that
comes
up
so
one
of
the
things
that
clement
who
lives
in
the
pacific
coast,
and
so
it's
like
7
30
a.m
for
him.
So
so
he's
given
a
pass
for
not
not
coming.
Is
it's
like
okay?
C
Well,
you
know,
I'm
not
the
only
one
that
doesn't
know
how
to
spell-
or
at
least
I
hope
not
because
that's
embarrassing,
but
and
so
it
wouldn't
be
nice.
If
you
guys
just
gave
me
a
command
to
switch
back
and
forth
between
edge
and
stable
and
that's
what
he's
looking
at
is,
I
think
he's
got
a
small.
I
haven't
seen
it
just
yet.
We've
spoken
on
it.
C
You'd
identify
one
of
two
things:
either
that
you
know
channel
stay
stable
or
edge
and
mesh3
ctl
will
manip.
You
just
do
a
simple
yaml
manipulation.
It
goes
in
and
does
a
find,
replace
on
stable
versus
edge
and
that's
helpful.
That's
a
little
more
convenient
eventually
and
by
the
way,
if
you're
running
measuring
edge,
which
maybe
let
me
try
to
do
right
now,
you'll
see
in
the
ui
we're
beginning
to
expose
the
ability
to
switch
between
release
channels
in
the
ui
or
we're
getting
closer.
C
C
C
It'll
then
pull
up
probably
the
login
screen.
Hopefully,
we've
got
the
latest
or
I
shouldn't
say:
hopefully,
we've
got
the
latest
like
this
should
be
edge
and
what
should
yep?
This
is
the
first
time
I'm
seeing
it
is
this
there's
a
new
and
which
means
this
is
probably
the
first
time
anyone
of
you
are
seeing
it,
but
there's
a
dysfunctional
set
of
controls
to
switch
back
and
forth.
So
not
only
would
it
be
convenient
to
be
able
to,
you
know,
switch
back
and
forth
between
channels
on
cli,
but
also
in
the
ui.
C
If
I'm
actively
running
the
thing
that
I'm
switching
the
channels
of
and
how
do
I
reboot
myself
and
and
what,
if
I'm
right
now,
I'm
running
mesh
read
locally,
so
I
can
manipulate
my
own
yaml
config
and
my
local
file
system.
If
I'm
running
measuring
kubernetes
and
how
does
this
work,
it
works
a
little
bit
differently
and
so
a
lot
of
interesting
surmountable
challenges
to
to
go.
Take.
C
C
The
the
pause
and
intentional
the
silence
is
intentional
in
case
there's,
questions
and
the
poor.
This
is
actually
a
very
timely
and
kind
of
a
great
I'm
looking
forward
to
depor's
tutorial
on
on
thursday
there's
a
lot
to
talk
about
within
mesri
ctl,
and
hopefully
some
of
the
things
that
you
know
will
be
built
out
over
time.
C
C
And
then-
and
then
so
so
here
are
you
running
and
so
for
my
right
yep,
it's
a
good
good
question
around
like:
where
do
you
can
you
run
meshery
and
and
and
all
that
so
so
in
my
particular
system,
I'm
running
I'm
sitting
in
front
of
an
imac,
so
I'm
running
this
on
docker
desktop
is
what
I'm
using
to
run
mescheri
locally.
For
me,
there
are
there's
been
quite
a
bit
of
effort
put
into
making
mesherie
compatible
with
it.
C
And
that
is
you're
you're,
either
looking
to
run
mescheri
as
a
on
a
system,
that's
running
docker,
so
it's
kind
of
like
you're
either
you
either
have
a
docker
environment
and
want
to
deploy
mesherie
as
a
docker
compose
app
or
you
have
some
form
of
a
kubernetes
environment,
and
you
want
to
run
meshri
as
a
kubernetes
app
in
both
cases.
It's
the
exact
same
meshery
bits.
C
The
reason
that
I
say
kind
of
don't
matter
is
because
well,
if
like
a
convenient
way
like
there's
a
universal
way
of
installing
meshri,
and
that
is
to
at
least
from
a
docker
from
a
docker
perspective
that
you
would
just
you
can
run
a
bash
script
and
napoor
is
familiar
with
this,
because
she's
had
a
helping
hand
in
creating
this
and
or
if
you
use
brew
on
mac
or
brew
on
linux,
you
can
use.
You
know
brew
to
install
it.
C
C
What
was
that?
Oh,
but
mesherie
itself,
it'll
run
even
if
kubernetes
is
absent,
and
so
actually
on
my
machine
that
I
was
showing
you
guys.
I'm
running
docker
desktop
on
os
x
and
I
don't
have
kubernetes
enabled
here
so
I'm
actually
just
running
docker,
that's
it
and
and
yet
I
was
able
to
bring
up
and
log
into
mesheri
and
as
a
matter
of
fact,
it's
more
than
I
don't
know
where
more
than
just
logging
in
to
mesheri.
C
If
I
wanted
to
do
a
performance
test
against
an
endpoint,
some
any
other
endpoint
doesn't
have
to
be
running
on
a
mesh
it'll.
Do
that
and
it'll
do
its
statistical
analysis.
It'll
save
those
results.
You
can
recall
your
results.
You
know
kind
of
all
that
stuff.
There's
about
half
of
messages
functionality
you
can
do
without
kubernetes
and
the
other
half
you
need
it.
C
E
E
System,
so
the
conformance
testing
is
currently
under
above,
as
in
it's
not
being
dissolved,
yet
basically
it
it's
kind.
It's
kind
of
crashing
the
adapters
right
now
and
the
buggers
are
all
block
is
already
been
raised
and
I've
assigned
it
to
someone
and
it's
been
under
development.
That's
what
I
would
say.
E
B
E
E
E
B
E
B
E
E
E
E
For
this
view,
oh
yeah,
we
got
a
success
message
for
sd
as
well,
which
would
have
been
installed
in
sd
system
and
yeah.
We
have
those
control
playing
up
as
app
as
well
so
yeah,
so
these
are
pretty
much
the
changes
that
went
through
with
the
linkedin
issue,
adapter
and
yeah.
Apart
from
that,
as
I
said,
the
smi
tool
is
facing
some
bugs
right
now
and
yeah
should
be
fixed
by
next
time
that
be
made.
E
E
C
C
C
E
Yeah
abhishek
did
you
also
notice
the
osm
adapter
like
whenever
you
try
to
invoke
this
smi
test?
It
breaks
down
yeah
yeah,
so
that's
that's
what
the
bug
is
already.
It
didn't
exist
before,
like
I
used
to
run
a
smi
test
before,
but
that's
yeah,
it's
a
pretty
recent
one.
It
came
in
because
either
the
release
is
not
done
been
done
properly,
as
in
I'm
going
to
check
this
cd
if
the
release
has
already
actually
published
an
image
that
needs
to
be
checked
that
that's
mostly
should
be
the
issue.
E
C
E
C
D
A
All
right
thanks,
abhishek
nopper,
pointing
at
you,
you
want
to
talk
about
the
issue
that
you
put
up:
yeah
yeah,
so
I'll
first
demonstrate.
What's
the
issue
is
and
somehow
maybe
because
how
okay,
if
you're,
sharing
the
screen?
Okay.
So
this
is
the
issue
that
every
time
I
run
mastery
ctl
version
after
starting
like
if
measure
ctl
server
like
system
start
is
not
done.
It
will
show
up
in
error
that
connection
refused
and
it
will
pop
up.
This
error
error
occurred
invalid
character.
A
B
A
B
A
Okay,
cool,
so
what
actually
happens
is
I'll
share.
The
screen
when,
like
mystery
system
is
not
yet
started.
It
will
show
up
this
error
that
connection
diffuse,
which
is
fine
by
me,
but
when
I
start
the
server-
and
I
see
here
so
I
I
get
this
error.
I
know
the
reason
if
I
open
this
in
browser
as
well.
B
B
B
A
A
E
B
E
C
C
That's
the
opening,
a
tag
that
we're
seeing
back
where
it's
trying
to
unmarshal
the
json
and
it's
not
json,
it's
html,
so
what's
happening
is
she's
trying
to
do
measuring
ctl
version
without
without
mescheri's
server
actually
running
just
yet,
or
maybe
that's
the
wrong
way
of
saying.
Is
it
like
measuring
server
that
slash
yeah
yeah?
So
that's
what
I
just
said:
that's
probably
the
wrong
way
of
saying
it.
The
api
server,
slash
version
needs
to
be
exposed.
Unauthenticated
prop
is
one
way
of
coming
around
this
issue.
C
B
C
Oh
yeah,
it
yes
and
right.
Yeah
I'd
like
to
talk
about
that
for
like
yes
and
no
it's
like
well,
if
you're
a
hacker
you
go,
you
walk
into
the
building
wireless
network
or
something
I
don't
know
just
and
you've
got
you've
got.
You
know.
You've
got
a
your
wireless
access
point
has
snmp
running:
it's
got
two
different
community
strings.
One
of
them
is
protected
by
snmpv3.
C
That
makes
you
off
to
really
get
at
much
of
the
metrics
for
the
interfaces
that
are
on
the
wap,
but
the
other
one
does
expose
just
the
system
version
of
the
wap.
That's
all
that
you
can
get
well.
It
turns
out
that
it's
running
such
and
such
version
such
and
such
version
you
know-
has
such
and
such
exploit
in
it.
So
it's
like
I
don't
know.
I
mean
we're
not
in
that
situation.
This
isn't
we're.
C
So
the
user
in
that
yeah,
maybe
the
so
yeah
I
mean
in
that
case,
like
the
user,
would
have
off
to
the
access
point
already,
but
you
wouldn't
necessarily
want
an
unprivileged
user,
a
guest
user
to
get
access
to
your
version
potentially
but
anyway,
maybe
bad
example,
or
that's
a
sort
of
errant
conversation
on
the
example.
The
point
being
is
like
it's
a
good
question
as
to
whether
or
not
we
want
to
you
know,
make
make
authorized
users
have
or
authenticated
users.
I
guess
have
that
access
anyway
newport.
A
Yeah
yeah
yeah,
like
I
get
that
what
you
were
talking
that
if
I
have
not
selected
a
provider,
but
the
point
is
that
I
have
initially
selected
the
provider
I
have
logged
in.
Even
if
there
is
authentication
issue
I
am
logged
in
now.
I,
if
I
share
my
screen,
I
am
logged
in
still
the
page
is
showing
404.
So
that's
that's
something
which
I
am
really
concerned
about.
If
you
see
I'm
logged
in
here
from
my
account
and
the
api
is
still
oh,
it's
showing
four
zero.
Four
page
note
fine.
C
A
A
Even
like,
if
I
think,
if
it's
a
some
like,
if
something
is
wrong
with
the
end
point,
I
tried
running
it
locally
on
my
machine.
If
I
do
make
run
local
and
then
check
it,
it
shows
me
the
exact,
like
version
and
get
sha
so
it's
working,
then
it
doesn't
show
me
four
zero,
four
page
not
found
when
I
try
running
it
locally,
so
I
also
had
to
talk
with
kush.
He
told
me
that
it
could
be
an
issue
of
build.
A
E
E
E
E
C
E
C
C
Yep,
oh
yeah,
to
make
sure
we
qualify
what
the
issue
is
yep
as
part
of
that
it
would
be
good
to
it'd
be
good
to
circle
back
that
next
week
to
follow
up
on
like
hey.
When
is
it
that
tokens
are
required,
tokens
universally
required
for
all
endpoints
of
measuring
server,
some
of
that
stuff,
and
I
think
that
that's
universally
true
all
endpoints
would
mess
with
server,
require
authentication
or
authentication.
C
E
A
E
Okay,
so
the
thing
is
with
the
architecture,
so
the
main
focus
is,
is
measuring
more
focused
on
adapters,
or
is
it
more
focused
on
adapter
having
their
own
identity?
So
is
it
adapter
or
is
it
a
service
mesh
like
the
current
architecture,
focuses
measuring
on
having
adapters
and
then
adapters
have
their
own
instance
of
service
matches
right.
A
A
C
The
yeah
the
way
to
re-characterize
and
thank
you
sridi
the
way
to
re-characterize
that
on
a
route
is
what
you
said
about
mystery
server
is
accurate
on
it
measuring
adapters
that
their
purpose
is
to,
by
and
large,
hold
service,
mesh,
specific
knowledge
or
service
mesh.
Anything
that's
specific
to
a
given
mesh
that
that
logic
would
be
in
each
individual,
specific
adapter,
and
so
they
they
hold
knowledge
of
like
how
to
provision
a
specific
mesh
or
how
to
configure
it
or
how
to
deploy
an
app
onto
that
mesh.
C
E
So
the
thing
I
was
thinking
it's
the
main
thing
I
think,
like
current
measuring
ui,
you
see
all
those
adapters.
Why
not
remove
them
completely,
so
don't
over
overwhelm
people
with
so
many
service
meshes
just
give
them
a
plus
icon
with.
So,
if
you
click
on
press,
you
get
to
choose
a
name
for
a
service
mesh.
You
get
to
choose
the
type
for
the
service
mesh
and
like
if
the
adapter
is
running
fine.
E
If
it's
not
running
give
the
user
an
invocation
that
this
adapters,
like
you,
don't
have
this
plug-in
running
or
this
adapter
running
so
either
use
it
or
like
either
boot
it
up
or
use
the
current
adapter
that
we
that
you
have
running
so
you
know
yeah
giving
an
identity
to
service
mesh.
So
the
thing
is,
if
like
currently,
the
issue
I'm
facing
with
measuring
cli
commands
is
what
should
I
focus
on
like?
E
Should
I
focus
on
having
adapters
do
this
stuff,
because
the
app
because
the
actual
area
of
operation
is
the
service
match
and
not
the
adapter
adapter
holds
the
config
whole
c
raw
configuration
and
the
like
how
to's
so,
basically
the
how
to
guide
right?
But
the
main
thing
that
has
to
happen
is
on
the
service
mesh.
C
But
yeah
nope
the
message:
cli
master,
ctl,
yep
it'll.
It
should
only
ever
speak
to
mastery
server
message.
Server
will
then
decide
that,
hey
that
your
that
that
this
client
is
looking
to
perform
is
is
hitting
this
rest
end
point
or
is
looking
to
perform.
This
invoke
this
action
that
action
might
be
taken
care
of
by
measuring
server
or
message
server
might
push
that
action
off
to
one
of
its
plug-ins
one
of
its
adapters
for
execution.
C
E
Like
the
entire
issue
here
is
like
I'm
measuring
server,
so
I
get
a
request.
I
get
a
request
to
you
know
like
let's
say,
install
osm
right.
So
if
I,
if
I
have
installed
like
I
get
an
instruction
like
create
a
new
service
mesh
for
osm
I'll,
go
I'll,
go
ping.
The
osm
adapter
I'll
say
that
we
need
a
service
message.
We
need
an
instance
for
osm.
That
adapter
would
create,
would
create
an
instance
on
on
my
cluster,
and
so
I
have
service
message.
E
The
thing
is
next:
if
I
want
to
configure
the
the
service
mesh
a,
I
would
have
to
go
to
the
same
procedure.
So
what?
If
I
have
two
instances
so
like
like?
I
have
service
mesh
a
I
have
service
match
b,
both
running
on
the
same
cluster,
and
if
I
say,
if
I
say
to
the
config
adapter
like
work
configured
adapter
that
configures
osm
to
route
50.
C
Yeah
you're
right,
you
have
to
include
a
flag,
and
this
is
why,
like
you
and
I
haven't,
sat
down
to
look
at
the
design
of
the
mesh
commands
and
it
doesn't
have
to
be
me
per
se.
But
this
is
why,
in
the
past
you're
like
hey,
I'm
ready
to
review,
I'm
like
I
haven't
had
the
time
having
to
that
yep.
It
has
to
you're
right.
It
has
to
have
a
flag
for
it,
or
rather
you
want
to
design
it.
C
So
it
has
as
few
flags
as
possible
and
there's
some
cheats
to
potentially
get
there,
but
in
concept
the
command
needs
to
have
context
for
exactly
what
you
said
it.
Well,
it's
probably
context:
it's
not
just
it's
more
complex
than
what
you
said,
which
is
unfortunate,
but
it's
it's
what
adapter
it's?
What
mesh?
C
Maybe
it's
it's,
what
name
space
sort
of
these
these
details
within
there?
Okay,
that's
what
I'm
targeting,
but
even
before
that,
it's
what
kubernetes
cluster!
Oh
before
that
it's!
What
measuring
server
instance,
because
guess
what
you
get
a
copy
of
my
ctl,
you
might
have
multiple
messages
out
there.
Oh
crap!
So
yeah,
it's
a
good!
It's
a
surmountable
set
of
problems.
C
It
is
something
that
you
want
to
give
a
holistic
focus
on
as
you
as
the
command
structure
is
defined,
because
once
you
deliver
that
structure
to
someone
somewhere,
they
start
writing
bash
scripts
around
it.
They
start
teaching
themselves
and
familiarizing
themselves
with
a
certain
set
of
commands,
and
then
it's
not
that
you
can't
ever
change
it.
It's
just
that
we
actually
did
it
once
very
early
on
in
measuring
and
folks
came
back
and
said
man.
I
missed
that
other
command.
It's
like
a
so
do
I,
but
this
is
the
right
structure,
long
term,
so.
E
I
mean
the
idea
here
is
to
have
a
name
name
flag
or
an
id
flag,
or
something
like
a
unique
identity
for
each
service,
much
instance.
So
if
I
say
the
adapter,
we
service,
like
let's
say
osm
adapter-
that
I
need
one
instance
of
name
so
sma
and
the
other
instance
of
smb.
It
should
create
two
instances
and
not
give
me
an
error
on
the
second
one
saying
that
an
osm
service
mesh
is
already
done.
C
Yep,
that's
a
good
way
when
I
was
alluding
to
sort
of
cheating
the
system
before
that's
a
good
way
of
kind
of
cheating.
The
system
that,
if,
if
there's
a
map
or
if
there's
a
a
list,
a
key,
a
set
of
keys
that
says:
hey,
look
the
instead
of
the
user
specifying
a
flag
for
what
measures
server.
Instance,
what
kubernetes
cluster?
What
adapter?
What
service
mesh?
E
That's
that's
the
entire
issue
here.
So
when
I
was
thinking
about
what
flags
I
should
have
for
each
command,
the
issue
is:
do
I
need
to
pass
on
an
id?
Do
I
need
to
pass
in
the
name
and
the
other
thing
is:
does
the
adapter
actually
or
does
the
measuring
server
api?
Do
they
actually
support
these
parameters
that
are
passing
because
I
would
not
be
able
to
test
until
and
unless
the
adapter
does,
the
specific
function
through
the
ui.
C
C
Releasing
some
early,
some
new
capabilities
in
mastery
ctl
we
can
and
those
who
probably
persist
that,
even
if
we
got
to
a
point
where
we're
able
to
cheat,
you
know
quote
unquote
and
just
let
the
user
provide
like
one
flag.
We
would
still
want
to
support
the
user
specifying
any
of
those
other
flags.
So
structurally,
it's
still.
E
E
So
that's
the
name
so
for
that
we
currently
don't
have
an
option
from
the
adapters
and-
and
the
thing
is
like
my
whole
concept
here
was-
is
measuring
focusing
on
adapters
or
is
it
focusing
on
service
mesh?
Because
if
we
are
focusing
on
service
mesh,
then
we
should
have
a
separate
id
based
system
where
we
would
actually,
we
will
classify
the
service
meshes,
but
the
each
service
mesh
has
its
own
identity
and.
C
Yeah,
that's
where
mesh
sync
comes
in
yeah
you're
right,
like
hey,
we
want
to
name
them
and
track
them
and
manage
their
life
cycle.
Mesh
sync
will
help
with
that
today.
It
is
possible
to
differentiate
between
two
different
service
mesh
instances
of
osm,
for
example
in
the
same
cluster,
because
osm
and
no
service
mesh
supports
being
deployed
to
the
same
name
space.
So
inherently,
if
you're,
identifying
a
namespace
you're
identifying
a
different
service.
Mesh
instance
you're
right
about
the
dashboard,
and
that's
why
I
had
some
the
the
dashboard
in
meshary
today.
C
All
of
this
stuff
is
a
point
in
time
thing
like
the
entire.
Almost
everything
is
it's
an
ugly
baby.
The
the
thing
is
is
like.
If
we
didn't
have
an
ugly
baby,
we
wouldn't
be
here
like
we
have
to
so
it's
an
iterative
thing
where
yeah
the
sys,
the
dashboard.
That's
there.
I
I
just
had
people
change
the
name
of
that
to
system
dashboard,
because,
ultimately
we
don't
want
to
show
people
any
of
that
info
unless
something's
wrong
with
measuring.
Then
they
go
look
at
it.
C
So
for
now,
that's
why
we're
showing
them
that
this
to
them,
when
mesh
sync
gets
going,
we'll
show
them
better
stuff.
Right
now,
with
the
mesh
sync,
the
poor
man's
version
of
mesh
sync
that
we
have,
which
there's
an
endpoint
for
information
about
a
bunch
of
information
about
istia,
anytime
istio,
is
deployed.
That
info.
Is
there
available
in
an
endpoint
that
different
uis
can
be
built
off
of?
We
should
be
trickling
some
of
that
stuff
up.
C
We
should
be
auto
connecting
to
prometheus
and
grafana,
because
we
already
know
that
they're
there
and
there
are
already
endpoints
for
doing
that
and
on
a
route.
Yet
please
like
go
create
some
new
like
propose
a
different
dashboard
like
different
widgets,
or
what
have
you
yeah?
That's
totally,
that's
so
much
needed,
and
irrespective
yeah.
F
C
It
it's
fine
like
the
naming
convention.
That's
not
even
I
mean
that
there's
categories
for
operations
already
and
yeah
there
should
be
conventions,
but
even
if
that's
not
gonna
solve
your
problem,
because
your
problem
is
you're
sitting
in
a
stateless
environment
with
mastery
ctl
and
you
don't
know
what
the
capabilities
are.
That's
where
grpc
or
something
like
that
would
come
in
so
okay,
thanks
treaty,
yeah,
thanks
honey,
rude
keep
on
go,
go
execute
on
those
that
dashboard
stuff
you're
talking
about,
and
the
measurements.