►
From YouTube: Meshery Development Meeting (July 1st, 2020)
Description
Welcome Satya, Abishek, Keith!
A
A
A
A
Okay,
good
deal
well
we're
just
three
minutes
after
I
know.
There's
a
well,
I
know
there's
a
few
different
folks
who
are
going
to
join
today,
and
so
I'm
anticipating
a
couple
of
their
topics
if
you're
on
the
call
go
ahead
and
slap
in
your
name
in
the
attendee
list,
and
while
we
wait
for
a
couple
of
others
to
join
just
a
reminder
of
a
couple
things
one,
we
record
these
calls
and
post
them
on
youtube.
A
So
just
be
aware,
and
so
you
can
watch
them
at
your
for
your
viewing
pleasure
later.
Two
is
that
in
this
meeting
you
know
we
have
a
bunch
of
meetings
every
week.
This
one
is
intended
to
follow
the
format
a
little
bit
like
a
little
bit
of
a
daily
stand
up,
although
I'll
acknowledge
that
anytime
someone's
giving
an
update
that
it's
going
to
take
a
little
bit
longer
than
60
seconds.
You
know
than
two
minutes,
we'll
probably
end
up
talking
about
that.
A
A
A
Yeah
thanks
for
yeah,
so
thanks
for
being
here
today,
this
is
great
just
to
help
nikhil
and
abashak
and
naveen
and
and
the
others
that
are
on
the
call
get
familiar
with.
You
say
a
bit
about
your
current
focus
or
kind
of
what
you
what
you've
been
doing.
What
you're
excited
about
you
know
how
you
got
here.
B
A
A
A
B
A
Okay,
that's
that's
good
that,
because
that's
by
design,
that's
two
commands
and
boom
bam
you're
there
as
a
matter
of
fact
yeah
depending
upon
there's.
Actually
I
don't
know
if
you
believe
this
there's
actually
a
way
in
which
you
can
run
just
one
command
and
it'll
it'll,
bring
you
all
the
way
into
your
it'll,
install,
download
and
install
measuring,
run
it
and
bring
you
into
your
browser,
which
is
question
for
you,
though.
A
B
A
B
A
C
A
Okay,
so
so
satya,
you
kind
of
one
of
those
next
steps
to
take
with
measuring
is
probably
running
a
local
instance
of
kubernetes
and
it's
it's
pretty
easy
to
do
with
a
tool
called
mini
cube,
and
if
you
end
up
going
to
the
measuring
documentation
there
it
should
be.
You
should
find
that
it's
just
a
couple
of
steps.
It's
got
a
link
to
minicube
and
you
download
it
and
say
I
know
all
right
forget
what
you
even
say:
mini
cube
start
or
something
it's
fairly
straightforward.
A
A
A
A
Well,
or
I
guess
the
the
statement
is,
my
measuring
instance
is
not
automatically
connecting
to
the
available
kubernetes
cluster
that
my
user
has
access
to,
and
but
it
should
be,
it
should
be
auto
connecting
what's
your
experience
here.
Do
you
have
to
go
initialize
your
connection
to
kubernetes
manually.
B
C
A
And
it's
it's
disappointing,
it's
it's!
It's
not
the
best
experience,
because
users
generally
don't
like.
If
I,
if
I
disconnect
my
kubernetes
instance
here
and
when
they
load
up
mesherie
it'll,
look
like
this
and
the
it's
they're
gonna
they're
gonna
come
in
they're,
not
gonna,
be
guided
to
do
anything.
It's
just
gonna
say
not
connected
to
kubernetes,
they're
gonna,
say
okay!
Well,
what
can
I
do?
Click
on
things
they're
not
going
to
work
because
the
adapters
aren't
going
to
get
connected
because
we've
got
some
other
poor
logic
in
mesheri.
A
That
really
expects
to
have
a
kubernetes
connection
before
it's
using
the
adapters,
which
isn't
the
best
experience
I
mean
so
there's
a
dependency
here
and
if
we're
not
auto
discovering
the
user's
cube,
config
the
and
we
don't
even
have
a
button
to
configure
kubernetes
they're
going
to
fumble
there's,
no
doubt
the
clicking
is
going
to
happen
between
here
with
a
nasty
sign
or
which,
by
the
way,
that's
the
the
term.
For
this,
it's
not
it's,
not
a
toaster
notification.
A
A
E
A
Yeah
so
nitish,
I
think
you
you're,
if
you're
on
now,
yep
you're
tracking
well
that
docker
or
a
meshery
when
run
as
a
docker
compose
application
or
when
run
as
a
container.
You
know
docker
container
on
your
local
system,
it
it
does
mount
the
user's
home
folder.
I
don't
think
it
mounts
the
whole
folder.
It
mounts
specifically
a
couple
of
of
paths,
but
one
of
those
being
the
cube,
the
user's
local
cube
folder
such
that
then
it
can
read.
A
It's
a
read-only
amount
to
read
in
that
user's
local
cube,
config
meschery,
then
parses,
the
cube,
config
and
just
assume
makes
the
assumption
that
the
current
context
is
the
desired
cluster
that
the
user
wants
meshri
to
interact
with,
and
that's
you
know,
90
of
the
time.
That's
what
they
want
to
do,
the
other
10
of
the
time
they
can
come
back.
You
know
they
can
come
in
and
load
up
their
cube,
config
and
choose
a
different
context.
Name
from
here.
D
A
Nice
yeah,
please,
that
would
be,
that
would
be
lovely.
It's
been
an
outstanding
issue
for
way
too
long.
It's,
I
might
have
to
go
back
in
time
and
tell
you
guys
how
many
users
have
had
this
bad
experience,
or
rather
it's
not
a
bad
experience.
It's
it's
upsetting,
because
we've
had
such
a
smooth
and
aw
like
this
is
what
what
should
happen.
They
should
land
in
here
and
they've
got
colorized
buttons.
A
They've
got
things
connected,
I'm
sure
some
people
are
kind
of
wondering
like
I'm,
I'm
sure
we
haven't
done
a
user
study
yet,
but
I
suspect
well.
First
of
all,
this
is
not
the
dashboard
that
we
want
to
show
people.
This
is
this
is
not
a
dashboard
that
tells
you
about
your
service,
mesh
environment
and
the
services
that
are
running
and
their
health
and
their
workloads
and
how
your
meshes
are
configured
et
cetera.
It's
not.
This
is
the
best
that
we
could
do
given
what
we
had
had
at
the
time.
A
What
this
is
this
is
your
your
measuring
health
panel
more
or
less
your
measuring
configuration
panel
this.
This
is
really
much
more
of
the
the
the
settings.
What
you'll
find
inside
of
settings,
but
a
bit
of
measuring
health?
If
you
will
so
it's
not
entirely
inappropriate
that
it
that
these
things
are
listed
on
a
dashboard,
it's
just
that
you
would
rather
have
each
of
these
four
areas
be
a
small
panel
in
a
greater
dashboard,
or
you
know
so,
or
something
else,
but
anyway.
A
This
is
all
that
we
have
right
now,
and
so
what
I
was
saying
is,
I
suspect,
that
oftentimes
users
will
come
in
and
they'll
say
wow,
okay,
so
yeah
I
wanted
to.
I
didn't
know
that
you
know
so
it
already
deployed
a
console.
Okay,
great
I'm
gonna
go
yeah.
I
want
to
go
interact
with
console
and
they
click
to
do
that
and
they
get,
and
they
don't
understand
that
it's
it's
no!
No!
A
A
Think
about
it,
if
you
would,
we've
got
some
some
folks
in
the
community
who
are
like
nikhil
and
a
bunch
of
others
that
who
are
capable
in
terms
of
improving
this
user
experience
so
nikhil.
This
might
actually
be
something
for
us
to
to
talk
about,
maybe
yeah,
okay,
good,
all
right!
Thank
you
for
this
abhishek.
It's
it's
been
really
frustrating
for
me
for
a
while.
So
let's
go,
let's
go
back
to
the
meeting
minutes
get
back
to
satya
thanks.
A
F
Okay,
I
guess
the
the
first.
The
first
thought
is
we're
starting
with
seo
and
that's
that's
fine.
We
have
go
based
apis
and
client
sets
for
the
rest
of
the
service
meshes.
I
know,
linker
d
runs
golang
based
control
plane,
so
it
it
should
have
something
ideally.
But
what
about
what
is
it?
Kong's
yeah
get
what
it
is
or.
A
F
The
the
language
is
important
just
to
just
to
import
the
type
otherwise
we'll
have
to
we'll
have
to
look
at
their
crds
and
build
our
own
data
structures,
and
that
means
maintenance
will
be
a
pain.
I
mean
it's
not
too
bad,
but
we'll
always
have
to
keep
up
with
the
versions.
So
if
we
had
an
official
one,
that
would
be
a
lot
easier.
I
don't
know
I
mean
it's
worth
looking
at
what
solo
has,
because
I
think
they
they
built
their
own
like
before.
F
A
To
answer
your
okay,
I
guess
there's
a
couple
of
things
here,
one
to
answer
your
question
in
curiously:
it
should
be
the
case
that
our
landscape
would
help
us
answer
this
so
and
you're
right,
like
linkard's
control
plane,
is
in
go
now
the
we
just
talk
about
the
primary
language
on
each
of
these,
which
doesn't
explicitly
say
what
the
goal
I
don't
know,
I
don't
know
of
a
control
plane
that
is
not
written
in
go.
I
do
know,
I
think
ambassador,
so
ambassador
doesn't
count
as
a
service
mesh,
but.
F
F
So
so
something
like
that
will
again
be
a
a
main
sustenance
issue
or
a
maintenance
issue.
If,
let's
say
linker,
d2
doesn't
have
the
official
client,
the
cl,
the
client
go
supported
data
structures
or
whatever
you
want
to
call
it
like
the
like.
The
data
structures
that
implement
the
object,
interface,
runtime,
object,
interface,.
A
A
A
F
Yeah,
so
so
we
are
watching
with
this
year.
We
are
watching
the
virtual
services
and
the
the
other
custom
resources
right,
and
the
watchers
like
adi
mentioned
was
that
there's
two
two
types
of
clients:
you
have
a
you,
have
a
virgin
client
and
you
have.
You
basically
have
an
untyped
client,
so
there's
a
type
version
client,
which
is
something
like
how
you
have
it
the
deployments
and
in
istio's
case
they
have
their
their
apis
that
they
have
defined.
F
But
if,
if,
if
we're
we're
going
that
down
that
road
for
stone,
which
makes
sense-
but
I'm
just
thinking
about
other
service
meshes,
how
much
work
is,
I
don't
know
if
there
is
an
alternative
to
this.
We
might
just
have
to
bite
the
bullet
and
do
that,
but
I'm
just
it.
We
should
identify
all
the
meshes
and
whether
they
have
the
type
clients
or
not.
F
F
But
how
would
how
would
that
work
for
say,
octane
or
nsm,
or
whoever
else
is
doing
the
same
same
traffic
management
features
or
whatever
that
we're
trying
to
abstract
right?
I
don't
know
how
we
watch
these
resources
without
defining
data
structures
for
them,
if
they
they
must
be
defining
them,
but
if
they
took
sds
out,
that's
gonna
be
super
ugly
and
hard
to
interface
with
if
they,
if
they
didn't,
if
did
it,
they
did
it
cleanly
like
sds,
wasn't
necessarily
for
kubernetes,
but
some
of
these
were
kubernetes.
A
A
Yeah,
because,
just
because
we
go
interface
with
helm,
doesn't
mean
that
we're
gonna
support
a
brownfield
deployment
or
intelligently
discover
the
objects
in
the
environment
in
in
that
regard.
So
a
couple
of
comments
there,
like
in
that
regard,
all
right.
So
just
as
you'd
commented
in
the
past
that,
like
hey,
I'm
getting
economies
of
scale
by
doing
envoy
specific
things
with
envoy's,
pervasive
use
in
the
data
plane
that,
like
you'd
mentioned
octorine
as
an
example
and
citrix
says
another
that
are
these.
Are
these?
Are
istio
service
meshes
or
they're?
Based
on
on
istio?
A
Oh,
I
see
which
helps
or
there's
economies
of
scale
there.
Okay
istio
is,
I
think,
as
some
of
us
know,
is
shifting
away
from
supporting
helm
and
just
yeah
the
the
issue
with
them.
Just
support
like
pushing
and
supporting
istio
ctl,
as
the
you
know,
as
the
main
way
to
deploy
it
is.
I
don't
believe
that
the
istio
go.
Client
is
very
robust.
F
I
mean
they're,
not
they're,
not
pushing
justice
to
your
cto
right
is
the
operator.
Is
there
the
primary
one?
So
you
can
create
the
custom
resource
either
through
stu
ctl
or
you
can
do
it
yourself,
so
it
could
be
programmatic
to
or
manually
created,
because
it's
it's,
the
spec
is
laid
out
and
they
have
exposed
that
spec
in
their
operator
repository.
So
I
mean
we
don't
care
about
that.
They'll
always
have
the
virtual
service
objects
or
api
data
structures.
They'll
always
have
the
destination
rule
blah
blah
blah
that
they
support
so
so
yeah.
F
F
So
we
always
need
that
api
data
structure,
which
has
your
metadata
whatever
the
kind
and
all
that
information
and
it
implements
the
the
object
interface
because
in
in
an
in
an
untyped
plan
you
get
the
you
just
get
the
object.
Interface
and
then
you,
you
basically
cast
that
to
the
right
implementation
of
that
interface.
F
Yeah
anyways
this,
isn't
this
isn't
a
deal
breaker
I
mean
it
was.
I
just
wanted
to
bring
it
up
as
a
sustenance
thing,
but
but
we
can,
we
can
always
start
with
sto
and
either
either
hope
for
the
rest
of
the
guys
to
create
apis,
because
they
would
want
someone
to
interface
with
their
meshes.
It's
not
just
the
end
of
the
road.
With
we
created
a
mesh
and
there's
nothing
else.
You
can
do
about
it
or
build
on
top
of
it.
F
F
A
C
C
That
also
I'm
not
able
to
understand,
because
recently
he
asked
me
about
that
mismatch
part.
So
I
I
recommended
you
to
use
rbm.
If
that
solves
the
issue,
then
I
will
have
to
check
if
you
can
document
that
markings
preferred
using
rpm.
Otherwise,
we
need
to
check
if
actually,
we
are
making
some
issues
regarding
the
jam
file
on
the
dependencies
one
because
it
has
been
increased
by
quarter
in
use.
A
It
sounds
so
nikhil.
Is
it
right?
It
sounds
like
we
have
two
questions
to
to
address
one.
Should
we
update
the
documentation
with
our
contributor
documentation
with
a
small
troubleshooting
area,
and
to
should
we
should
we
or
should
we
not
be
checking
in
the
dot
lock
file
and
would
and
would
that
even
help?
I
guess.
C
G
Even
that
didn't
work,
I
had
removed
ruby
like
the
existing
ruby
from
my
laptop
and
I
installed
ibm
and
installed
the
data
screwb
version
using
rbm,
but
it
still
gave
me
that
issue
when
it
came
to
the
nokojiri
nokokiri
gem
that
didn't
allow
me
to
run.
You
know
to
be
able
to
run
the
layer
5
website
or
even
the
mystery
upset.
G
G
All
the
others,
okay
I'll
then
try
to
uninstall
rvm
again
because,
like
I
just
said
that
you
know,
I
had
uninstalled
all
versions
of
ruby
or
even
any
dependency
files
that
were
there.
I
installed
ruby
again,
I
installed
ibm,
but
even
then
it
was
given
again
just
try
to
set
the
default
version
after.
C
A
And
then
is,
as
is
obligatory
as
satya
just
suffered
this
earlier.
Anyone
who
joins
any
of
the
the
calls
for
the
first
time
gets
to
go
through
a
bit
of
a
bit
of
a
quick
introduction
and,
and
so
and
so
keith
welcome
to
the
metroid
development
call
pl.
You
know,
please
say
a
few
words
just
to
kind
of
introduce
to
get
to
know
the
crew
a
bit.
H
A
All
right,
I'm
withholding
all
sarcasm,
very,
very
nice!
Thank
you
for
joining
keith
very
nice
to
keith
has
actually
it's
great
that
he's.
I
suspect
you
might
have
some
insights
as
to
some
of
the
troubles
that
we're
kind
of
covering
here.
These
are
not
foreign
concepts
for
you
at
all,
and
so
we're
just
yeah.
We're
part
part
of
our
challenge
today.
A
You
know
it's
third
party
to
the
actual
service
mesh,
that's
doing
its
functionality,
and
so
operators
are,
you
know,
likely
to
walk
over
to
the
mesh
itself.
You
know
update
its
config
and
then
meschery.
You
know
as
an
out-of-band
piece
of
management.
Software
wouldn't
necessarily
be
aware
of
those
changes,
so
the
mechanics
by
which
yeah
mastery
you
know
is
continually
staying
in
sync
is
so
you've
only
you've
only
solved
that
not
cracked
that
nut,
I'm
not
sure
how
many
times.
A
H
Things
from
infrastructure,
automation
and
lifecycle
map
and
to
some
capabilities
related
to
service,
catalog,
aggregation
from
entities
like
amazon
or
vmware,
hybrid
clouds,
and
then
today
I
run
services
that
do
application
and
infrastructure
and
container
host
monitoring,
and
so
I
think,
you're
right,
subscribing
to
the
frequency
and
understanding
integration
points
between
the
different
systems
that
are
changing.
And
how
do
you
build
a
real-time
service
model
versus
keeping
that
information
for
querying
later?
That's
not
real
time
specific
is
something
I
had
to
deal
with
pretty
frequently.
E
B
So
yeah
last
week
the
blog
is
up,
so
I
guess
it
was
pending
in
the
last
development
meeting.
I
guess
it
wasn't
notes
like
a
blog
was
upcoming,
so
the
blog
is
there,
the
pr
I'll,
add
the
link
to
the
pr
and
the
meeting
minutes.
So
please
review
that,
and,
along
with
that,
like
I
have
started
working
on
the
next
thing,
that
is
graphic
split
testing
about
traffic
split.
A
Okay,
a
little
bit
of
context
for
others,
but
also
a
question
naveen
said
so
for
others,
the
part
of
the
conformance
tooling,
that
we're
building
into
measuring
for
service
mesh
interface
means
that
meshi
will
be
well,
hopefully
the
canonical
smi.
You
know
implementation,
validator
or
the
conformance
tool
for
that
project.
There
are
four
specs
inside
of
smi.
B
No,
I
was
saying
like
there
were
three
services
that
we
would
give
it
out
to
so.
First,
we
will
just
send
all
that
effect
like
we.
We
we
should
have
ideally
like
at
least
at
least
five
tests.
I
guess
like
first
one
should
be
asserting
when
the
traffic
should
go
to
entirely
to
service
a
service
b
service
c
and
when
it
should
be
distributed
evenly
and
like
there
are
various
aspects
of
the
spec
that
needs
to
be
tested.
A
B
Yeah,
so
traffic
access
is
like
complete,
but
we
we
can
add
tests
later
on,
like
maybe
in
a
week
or
two
when
all
three
are
like
traffic
metrics,
graphic,
excess
and
left
split
are
complete,
so
traffic
specs
is
basically
just
asserting
traffic
specs,
which
is
by
the
way
the
fourth
like
a
fourth
smi,
the
the
fourth
part
of
the
fourth
pillar
of
smi.
So
it
is
just
checking
whether
a
specific
crd
is
defined
by
service
mesh
or
not
so
there
we
don't
have
any
problem
like
it
is
complete.
B
Yeah,
I
guess,
by
this
week
we
would
like
you
would
be
able
to
do
graphic
metrics
and
by
the
next
week
I
guess,
traffic
matrix
sorry
traffic
split
by
this
week.
We
would
be
able
to
do
graphic
split
by
the
next
week.
We
would
be
able
to
do
traffic
metrics
and
by
the
next
week
I
guess
we
would
have
covered
all
the
test
cases
of
like
all.
We
would
have
to
like
configure
and
write
configuration
for
each
test
case
and
assert
the
logic
there.
A
A
Okay,
fair
enough,
but
since
we're
on
the
topic
of
delivery,
you
can
go
ahead
and
sign
in
you
know
in
blood.
If
you
like
right
right
here
for
for
when
you
will
have
have
this
completed,
I
think
everyone
heard
him
say
three
weeks
we
are
recording
the
calls,
so
just
initials
will
do
that.
Will
work
right
there,
okay,
good
good!
It's
good!
It's
good
progress!
So
far,
we're
going
to
make
quite
a
bit
of
noise
about
this.
I've
been
talking
to
folks
already,
so
all
right,
very
good.
A
B
C
I
guess
so
that
won't
fit
instead
having
the
hash
is
working
fine,
I
guess
the
issue
that
you
were
talking
about
in
the
description
is
with
the
performance
tab
on
submitting
the
preferences.
The
page
gets
reloaded,
reload
and
navigates
to
the
environment,
tab
on
the
settings
page,
I
guess
that's
the
issue.
You
were
mentioning,
so
I
came
up
with
that
and
I
saw
the
code
and
I
got
the
area
where
that
thing
is
happening.
C
Okay,
as
for
I
think,
on
saving
means
on
submitting
the
preferences,
there
is
an
on
change
function
which
will
automatically
save
the
props.
I
mean
it's
called
the
props
and
say
update
the
value
I
have.
I
haven't
checked
that
part
if
the
prompts
get
updated
and
the
value
is
reflected
or
not.
I
will
check
that
after
changing
it.
C
Although
we
are
using
react
routing,
but
basically,
if
we
are
creating
a
new
page
means
like
we
are
navigating
to
a
savings
which
it
has
a
dedicated
page
under
the
pages
section
in
the
uf.
So
for
few
of
the
pages
like
management
and
the
adapters
settings
performance,
they
have
dedicated
pages
created.
But
since
the
tab
section
is
a
means
same
page
thing
same
page
too,
so
it
gets
reloaded
when
we
say
switch
to
a
different
app,
the
components
the
different
component
gets
called
and
the
pc
gets
re-rendered
automatically
on
switching
to
a
different
time.
C
A
Okay,
okay,
so
of
the
two
challenges
that
we're
facing
one
is
the
one
is
navigating
the
user
to
not
only
a
specific
tab
within
a
component,
but
that
where
we've
got
like
sort
of
two
levels
of
of
a
menu
system-
and
so
it's
navigating
to
them
to
that
here,
let
me
let
me
I
didn't
realize
that
wasn't
sharing
the
screen.
Let
me
share
so
there's
so
there's
a
navigation
issue,
a
navigation
question,
but
then
there's
also
the
ability
to
refresh
individual
components
without
refreshing,
the
entire
page
right.
Those
are
the
two
challenges
that
we
have.