►
From YouTube: Meshery Development Meeting (Oct 28th, 2020)
Description
Welcome @HomeroMarin7! @realDhruvPatel walks through @smi_spec conformance with @openservicemesh
A
B
C
D
D
D
D
Yeah
devops
nice
all
right,
okay,
good!
I
was
just
trying
to
think
of
topics
to
talk
about
today
and
since
you're.
Here
we
might
talk
about
some
of
those,
so
okay,
yeah
it'll
yeah.
It
takes
a
little
while
to
cool
okay.
So
now
I'm
having
a
one-way
audio
conversation,
so
I
apologize
shredi.
I
thought
I
would
just
make
sure
that
you
were
able
to
accomplish
your
mission
of
setting
up
the
agendas.
The
agenda.
E
Well,
I
just
wanted
to
ask
this:
everything
is
okay,
because
you
want
to
present
it
in
like
three
hours
in
the
small
call
right.
So
I
ran
a
test
like
half
an
hour
before
to
check.
If
things
are
working
properly,
it's
it
is
supposedly
working
properly,
except
the
percentage
passed
is
yet
not
fixed,
which
I
think
I
will
get
back
to
abhishek
and
they
will
check
what
we
can
do
before
the
meeting
itself
and
yeah.
That's
all.
D
Nice
yeah
drew
thanks
for
thanks
for
pulling
this
up.
I
figured
that
we
should
yep.
We
should
put
drive
at
home
in
the
smi
call
later
today
that
hey
you
know,
here's
the
tool,
here's
all
the
orchestration
that
it
we've
got
to
use
a
different
term,
but
here's
all
of
the
here's.
What
the
conformance
test
tweet
looks
like
currently,
we've
only
have
we
only
have
so
many
tests
defined
assertions
defined
and
actually
drove
from
my
part,
I'm
slightly
confused
as
to
the
count
of
assertions.
D
E
D
There's
four
specifications,
each
of
which
have
a
number
excuse
me,
a
few
different
features
and
we've
written
tests
that
validate
whether
or
not
when
you
invoke
a
feature.
Basically,
when
you
interface
with
the
smi
api,
these
tests
will
validate
whether
or
not
one
any
one
of
those
seven
service
meshes
respond
in
kind
like
respond
like
they
should,
and
so.
Traffic
split
is
a
good
example
of
something
that
like
it's,
is
a
great
example
of
a
reason
why
we
would
use
measures
because
you've
got
to
spin
up
seven
different
types
of
service
meshes.
D
There
aren't
many.
You
know
any
other
tools
that
do
that.
You've
then
got
to
generate
a
bunch
of
load
consistently
and
then
measure
that
load
and
look
look,
how
the
traffic
is
being
split.
So
hence
mastery
is
a
perfect
tool
for
this
stuff.
You've
got
to
tran,
you
know,
collect
up
the
results,
transmit
them
back,
guarantee
their
provenance.
D
You
don't
as
an
individual
user,
you
don't
that's
not
the
use
case,
the
use
case
here
then.
Actually,
all
the
use
cases
are
written
down
and
the
unlike
normal.
The
use
cases
are
actually
written
down
here.
Seven
service
meshes
even
for
one,
even
for
a
service
mesh
team.
That's
not
exactly
the
use
case
either.
D
D
D
Anirud
the
reason
I
called
your
name
out
is
because
right
now
there's
a
bug
I
would
say
or
a
feature
that
we
should
just
remove,
which
is
in
the
smi
results
report
right
now
inside
of
measuring
there's
a
percentage
passed
column
which
is
intended
to
total
up
the
30
tests
that
are
there
right
now
and
give
an
overall
percentage
now.
I
assume
that
that
calculation-
or
I
think
the
assumption,
is
that
that
calculation
is
going
to
be
done
on
the
back
end,
but
dhruv
that
may
be
where
the
disconnect
is.
Is
that
maybe
there's?
D
E
D
Remind
me
of
the
some
of
the
tests
may
pas
like
of
the
12
they're
there.
Six
of
them
may
pass
the
seventh
might
fail,
not
all
of
them
are
executed
in
serial,
though
so
like
it's
possible,
for,
I
think
two
branches
of
the
twelve
tests,
two
of
the
total
12
tests.
There
might
be
two
parallel
sets
of
tests
going
on,
and
so
you
could
actually
technically
have
two
failures,
and
so
two
reasons
for
failure.
Does
that
sound.
F
Right,
we
don't
segregate
based
on
that
currently,
so
so
that
right
now
the
logic
behind
computing,
the
result
of
if
it
is
failing
or
passing
or
what
is
the
capability
and
what
is
the
percentage
passing
all
of
these
are
just
very
blunt.
It
is
not
properly
as
in
it's
just
it's
just
taking
average
of
all
the
assertions
that's
been
done,
so
it's
it's,
not
it's
not
accounting
for
the
set
of
sets
that
run
in
parallel
or
the
independent
ones.
It
does
not
care
about
that
right
now.
It's
very
blunt.
D
F
F
D
F
Ideally,
it
should
be
working
so
so,
basically,
there
were
two
discussions
that
went
down
parallely,
so
we
had
planned
on
putting
certain
features
and
then,
after
a
couple
of
comments
on
the
smi
talk,
we
tried
we
planned
on
removing
them.
So
some
of
these,
like
the
capability
we
planned
on
not
having
half
it's,
either
none
or
full
and
and
the
percentage
we
try
to
get
rid
of
like
we
have
we
plan
to
get
rid
of
the
percentage.
F
D
D
D
D
D
C
D
D
G
G
D
G
H
D
No,
I
just
I
I
just
I
know
that,
there's
a
number
of
people
on
this
call
who
are
who
take
pride
in
what
you're,
showing
because
they've
been
working
and
working
for
a
long
time
to
see
some
really
really
pretty
cool
things
so
I'll
just
this
is
good.
I'm
seeing
a
little
bit
of
this
for
the
first
time
as
well,
which
is
nice
so
just
context
for
everyone.
There's
a
couple
of
important
components
to
mesh
re
that
I've
been
missing
for
a
while.
D
Mesherie
is
focused
on
performance
management
and
trying
to
help
give
people
easy
to
use
tooling
to
stand
up
a
service
mesh
play
with
it.
Learn
it,
but
a
lot
of
that
functionality
is
just
for
teaching
and
learning
up
until
the
point
that
mesh
think
is
brought
in
then
the
man
life
cycle
management
becomes
a
real
thing
from
eshri
in
the
meantime,
is
the
performance
management
that
is
real
and
is
useful,
and
people
can
use
here
and
now.
D
C
D
D
So
I
can
no
longer
be
blamed
for
all
the
horrible
names,
but
anirud
beneath
and
a
collection
of
others
are
trying
to
bring
forth
really.
You
know
this
a
new
way
of
visualizing
things
inside
of
mystery.
You
could
think
of
this
particular
visualization
as
being
an
alternative
way
to
perform
the
same
tasks
that
you
can
today.
So
today
you
could
go
run
a
performance
test
by
filling
in
a
form.
D
Nikhil
has
been
working
on
bringing
forth
and
and
cush
have
been
working
on,
bringing
forth
a
performance
test
profile
so
that
you
don't
have
to
fill
in
the
same
form
every
time
you
want
to
run
the
same
test
and
then
beneath
and
dhruv
and
others
on
your
route
have
been
working
on
the
ability
to
come
in
here
and
visually
invoke
such
a
performance
test.
So
that's
what
vinith
is
showing
is
anirud
has
beneath
right.
Now,
if
you
click
on
one
of
these
nodes,
does
it
provide
you
the
context,
menu
to
invoke
a
performance
test.
G
G
G
G
D
D
Yeah
yeah
right,
there's
an
icon
on
your
desktop
there's
it's
over
in
the
applications
drawer.
You
can
do
it
from
the
terminal
anyway.
My
point
point
is:
is
like
yeah,
it's
an
open
consideration
that
we
should
refine
like.
Oh
for
certain,
like
hey.
What's
the
task
at
hand
for
the
user?
How
and
how
few
clicks?
Can
the
user
execute
that
task?
Okay,
it
might
mean
that
some
prefer
prefer
the
details
drawer.
D
Some
might
prefer
the
you
know
the
the
action
right
on
the
thing
and
it's
one
of
those
discussions
that
that
will
it'll
be
probably
much
better
had
as
we
get
toward
the
end
or
the
mvp.
If
you
will
for
this
because
then
you'll
be
able
to
refine
and
re
and
factor
like
what
all
the
choices
are,
can
you
exp?
Can
you
play
with
the
details,
pain
for
everyone
and
just
get
the
sense
of
what
the
details
pain
does
outside
eventually.
G
Yeah,
so
this
is
the
this
is
what
you
see.
First,
it's
an
audio
tab
and
actually
the
other
tabs
are
not
filled.
Yet
these
are
traffic
management
tab
and
this
is
security,
and
this
circuit
breaker
tab,
and
this
is
load
balancing
time.
These
aren't
exploded,
and
this
is
the
only
thing
we
are
right
now.
E
D
D
F
G
H
G
D
H
A
I
C
C
C
A
E
C
F
So
this
is
mainly
regarding
people
who
wanted
to
work
and
work
with
work
on
measuring
ctl
mesh
commands.
There
are
a
few
commands
that
can
be
done
as
of
now
at
least.
I
think
they
can
be
done.
Let
me
know
if
they
can
so
you
have
mesh
list
create
or
deploy
with
like
we
didn't
come
up
with
a
fixed
one.
So
if
aliasing
is
possible,
keep
both
then
it
they
will
naturally
remove
and
delete.
So
I
added
destroy.
It
sounds
better
sometimes,
and
we
have
measuring
workload
in
workload.
F
We
have
list
and
apply
so
these
commands
can
be
done.
The
apis
are
at
least
functional
for
now.
So
there
are
some
prerequisites.
First
thing
you
need
to
know
golan
because
obviously
you'll
have
to
look
into
the
server
for
apis,
at
least
for
now,
then
you
need
to
have
some
experience
with
cobra
and
flag
usual
flag
is
actually
united,
so
cobra
in
its
flag
from
the
lang
flag
library.
It
also
has
p-flat,
but
you
still
need
to
know
about
flex
and
we
are
using
cobra
in
mercury,
mercury
ttl.
F
So
a
bit
experience
on
how
cobra
works
is
good,
so
and
also
the
architectural
idea
of
measuring,
because
you're
working
on
the
cli-
and
these
are
a
few
properties
of
service
of
business,
so
yeah
service
mesh.
So
a
few
properties
service
mesh,
you
might
have
to
consider
namespace,
it's
basically
kubernetes
namespace,
the
name
of
service,
mesh,
port
and
host
of
service
mesh,
like
I
guess
there's
I
defined
it,
but
I
know
like,
should
we
go
with
this
for
adapter
type
type,.
F
Then
I'll
have
to
look
at
those
like.
I
didn't
find
the
time
for
it.
Okay,
so
I'll
look
into
the
smp
protos
I'll
update
this,
but
we
have
at
least
name
space
name
port
host
are
here.
Then
you
have
supported
operations
that
is
supported,
ops,
then
the
example
workloads
or
template
workloads
currently
like,
for
example,
you
have
book
info
and
I
guess
one
some
http
or
something
on
sql.
Then
you
have
example
configurations
these
are
the
template
configurations
and
the
ui,
so
my
ui
is
dumped
right
now.
F
F
Then,
once
you
are
once
you
know,
actually
what
you
need
to
implement
or
what,
like
you,
have
the
required
functionality
in
mind,
then
you
go
for
the
second
step.
That
is,
you
use
a
rest
line
like
postman
or
any
desk
line,
whichever
you
prefer
use
it
create
a
mock,
create
an
api
or
a
request
and
ping
the
message
mastery
server
and
see
if
it
reacts
to
the
certain
functionality
you
want
to
implement.
F
So
these
this
part
is
critical,
because
here
you'll
be
doing
all
the
logic
part
and
once
you're,
once
you
have
a
api
working
or
who
wants
basically,
once
you
once,
you
have
a
request
that
actually
interacts
with
measuring
server
and
gets
the
job
done.
You
simply
go
into
golan.
You
you
change
the
code
in
postman.
You
can
export
a
request
on
into
golang
and
but
the
first
thing
you'll
have
to
look
look
into
measuring
ctl
find
out
what
you're
getting
already.
F
Obviously
you
run
the
binary
after
make
and
test
it
out,
and
if
that
is
working,
fine,
you
simply
create
a
full
request,
mention
it
on
measuring
cpl
or
ping
any
like
anyone,
literally
lee
kush.
We
need
anyone,
so
this
is
actually
how
I
get
mesh.
The
commands
done
in
case
someone
hadn't
doubt
so
these
are
all
taken
from
the.
D
D
I
use
this
is
a
really
it's
really
timely
that
any
root
is
going
through
this,
because
this
may
very
well
be
an
ideal
spot
for
you
to
make
a
mark
there's
a
lot
of
functionality
and
measuring
that
is
yet
to
be
exposed
into
the
mesh
free.
Ctl
anirudh
has
championed
a
few
new
ones
as
a
matter
of
fact
anirud
I
don't
know,
did
you
speak
to
the
whole
new
command
experiment
and
and
how
that
works
like
what
the
intended.
F
F
So
if
the
command
like
it's
obviously
every
command
that
simply
goes
into
the
workflow
is
not
a
stable
command.
You
need.
We
need
certain
feedback
for
it
and
need
to
know
if
people
actually
use
it.
Obviously
people
will
maybe
so
those
go
under
experimental
sub
command.
Currently
I
guess
lee
can
you?
Can
you
show
the
experimental
subcommand?
I
guess
the
repo
is
merged.
If
not
you.
H
D
Well,
yeah,
so
so
a
couple
things
on
my
way
to
showing
that
maybe
I'll
I'll
say,
yeah
hey.
I
should
check
to
see
what
version
of
measure
ctl
I
have
and
by
the
way,
there's
been
some
enhancements
here
recently,
where,
when
you
go
to
check
measuring
ctl's
version,
it'll
check
the
client
and
finally,
the
server
as
well
it'll,
also
supposedly
check
for
the
latest
version
of
mesherie
out
on
github
and
anyway
in
my
environment.
D
D
D
Behind
which
the
the
up
and
coming
command
and
its
intended
structure
would
be
laid
out
so
the
first
one
that
is
the
first
experimental
command,
that's
being
worked
on
right
now
is
one
called
mesh,
so
there's
a
lot
that
mashery
does
around
service
mesh
lifecycle,
management
and
and
so
the
command.
The
thought
is
that
that
structure
would
be
messy,
ctl
mesh
and
then
things
that
you
want
to
do
to
a
service
mesh
and
so
of
the
initial
sub
command.
D
There
is
validate,
and
actually
it
very
much
it
is
exactly
the
thing
that
dhruv
was
showing
earlier
well,
the
validate
can
be
useful
for
a
few
different
things.
It
can
validate
configuration
best
practices
of
a
service
mesh
or
it
could
validate
your
the
smi
conformance
of
a
mesh
or
maybe
yeah
or
there's
a
few
other
use
cases
to
come
on.
Do
you
want
you
want
to
talk
further
about
the
structure
of
this
command
or
like.
F
Yeah,
so
the
whole
message
of
mesh
commands
is
basically
manipulating.
The
service
meshes
through
adapters
on
measuring,
so
we
already
ui
can
do
that.
We
don't
have
a
way
on
having
mystery
server
measuring
cpu.
Do
that
at
least
for
now,
but
the
rest
apis
are
pretty
simple,
so
you
can
do
a
lot
more
with
cli
as
compared
as
compared
with
ui,
because
you
can
simply
create
an
automation
script
that
does
everything
through
the
cli
and
achieving
that.
So
we
I
started
working
on
metroid
ctl
mesh.
F
F
So
that's
more
on
exactly
what
service
mesh
like
at
least
what
measuring
sql
mesh
is,
and
we
like
we're
working
on
those
like.
We
need
at
least
the
functionality
of
ui,
and
once
we're
done
with
that,
we
will
try
to
improve
mercury
ctl
further
than
the
ui
if
possible,
because
it's
like
you
can
do
a
lot
of
you
can
do
a
lot
through
automation,
which
I
guess
wouldn't
be,
that
much
possible
by
the
ui.
D
It's
a
good
way
to.
I
don't
think
that
that
any
rude
that
you'd
necessarily
had
a
chance
to
kind
of
explain
and
demonstrate
the
experiment,
the
purpose
of
the
experimental
command,
I
want
to
say
something
other
than
command,
but
that
that
name
space.
If
you
will
for
up
and
coming
new
commands
and
the
mesh
validate
being
the
first
one
at
bat
and
that
yeah.
D
Am
I
sharing
my
whole
screen
now,
hopefully,
okay,
that
to
any
root's
point
that
this
measure
ui
it's
a
it's
a
client
of
mesherie
server's
rest
api,
just
like
mastery
ctl
is
a
client
of
mesh
reserver's
rest
api,
they're,
both
clients
to
anirude's
point
it's.
If
you
wanted
to
automate
something
it's
a
little,
you
can.
You
can
write
your
own
scripts
or
your
own
applications
that
interface
with
mesherie's
rest
api,
and
it
can
be
a
client
just
like
these
two
examples
that
we
see.
D
D
If
we
go
and
change
their
structure,
I
can
have
an
issue,
and
so
that's
in
part
the
reason
that
anirud
was
bringing
forth.
This
experimental
area
is
to
signal
to
people
like
this
is
alpha
beta
area.
So
you
know
user
be
warned
in
terms
of
functionality,
but
also
in
terms
of
the
potential
syntax
shifting
over
time.
D
D
One
item
to
convey
that
you
might
convey
to
kush
is
that
utkarsh
had
recently
landed
some
new
functionality
and
it
shows
up
in
the
ui
here
and
it's
beginning
to
tell
people
what
version
of
measure
they're
running.
What
release
channel
they're
subscribed
to
and
kush
was
trying
to
make
sure
that
we're
not
repeating
ourselves
in
terms
of
there's
a
specific
library
or
specific
function
that
that
the
go
code
uses
to
look
up
the
latest
release
on
github.
D
Take
that
version
number.
You
know
the
the
release
tag
v
x.x.x
and
compare
that
tag
to
the
currently
running
server,
and
that
way
you
know,
it'll
tell
people
you're
up
to
date
or
you're,
not
now
that
refactoring,
so
that
that
code
was
last
week.
It
was
in
measuring
ctl
and
so
the
ctl,
the
the
command
line,
client
itself.
When
you,
when
you
run
mastery
ctl
version,
it
has
the
ability
to
just
look
up
its
own
version
and
look
up
talk
to
the
server
if
available.
D
D
E
F
I
mean
like
I
haven't,
bought
it
up
like
so.
The
thing
is
in
case
anyone's
working
use
if
anyone's
using
the
apis,
the
right
way
or
the
right
area
to
look
into
is
this
place,
so
you
can
see
the
routers
and
you
can
get
the
functions
that
perform
certain
actions
if
in
case,
you
want
to
understand
how
the
apis
work.
Currently
we
don't
have
an
api
documentation
and
that's
upcoming.