►
From YouTube: WebAssembly Filters Meeting (Nov 22nd, 2021)
Description
WebAssembly Filters Meeting - November 22nd, 2021
Join the community at https://layer5.io/community
Find Layer5 on:
GitHub: https://github.com/layer5io
Twitter: https://twitter.com/layer5
LinkedIn: https://www.linkedin.com/company/layer5
Docker Hub: https://hub.docker.com/u/layer5/
A
Good
morning,
good
afternoon,
good
evening,
wherever
you
joining
from,
if
you
are
new
to
the
meeting,
which
I
don't
think
anyone
is,
you
can
introduce
yourself,
but
if
there
is
no
one
new
drop,
your
name
in
the
attendees
list
in
the
meeting
minutes.
Let
me
okay,
the
instruction.
B
A
Sorry
so
we
have
both
abhishek
and
mario.
Here,
that's
the
next
topic,
so
it's
just
about
how
you
guys
tested
out
the
cyprus
based
testing
of
the
ui
or
have
you
I
mean.
How
are
you
still
in
exploration
stage
or
there
is
a
pr
or
something
like
that,
so
obviously,
on
mario
some
things.
C
Yeah,
sorry
so
yeah
on
pat
sorry,
what
why
I
was
yes,
cyprus,
yeah
so
on
cyprus
like
there
is
no
update
according
to
like
quotes
in
this
prs,
but
I
think
there
will
be
after
some
time
since
I
was
busy
on
something
else.
So
no
update
from
my
side
I
mean
mario,
could
have
so.
D
Hello,
yes,
this
is
mario,
so
yeah
I've
been
taking
a
look
at
the
code,
looks
like
we
could
leverage
one
of
the
usual
patterns
on
cyprus
looks
like
all
of
the
three
pages
use
the
same:
the
same
mechanism
to
work
on
the
upload,
so
I'll
just
work
on
an
approved
concept
and
if
it
works
I'll,
just
create
a
pr
based
on
that.
D
There's
this
plugin
cyprus
plugin
for
file
uploads,
so
I'll,
just
see
if
it
works,
which
it
should
it's
like,
really
something
many
people
rely
on,
so
I'll
just
try
to
use
it
and
I'll
also
be
our
ones.
I
have
a
working
solution
for
that.
E
Actually,
I
think
if
I
press
file
upload.
Oh
okay
got
you
got,
you
got
you
yeah,
okay,.
B
B
B
They
deploy
a
couple
of
different
types
of
service
meshes
like
actually,
those
are
great
functional
tests
for
well,
at
least
for
performance
benchmarking,
some
of
that
functionality.
They
do
approach
the
testing
from
measuring
ctl's
perspective.
From
that
client's
perspective
you
know,
and
so,
which
makes
it
perfect
sense
that
we
would
be
able
to
reuse
part
of
those
workflows,
but
not
all
of
them,
because
from
cypress's
perspective
will
hit
it
using
mesh
reui
as
the
client
so
yeah.
C
Can
I
show
the
pattern
configurator
right
now
there
were
quite
a
few
bugs,
but
I
think
I
should
show
once
because
maybe
I
I
I
won't
be
available
tomorrow
on
meeting.
That's.
Why.
B
Yeah
boy
we're
having
a
challenge
with
so
nation
said
that
he's
got
he's
kind
of
I'm
out
for
the
week
and
gaurav
he
was
busy.
He
was
online.
This
weekend
was
walking
through
some
things,
but
I
could.
B
B
A
So,
let's
move
on
to
the
next
item
and
then
we
can
also
talk
about
what's
unplugged
and
still
that's
a
new
thing
and
probably
is
of
use
to
us
also
so
yeah.
Let's,
let's
move
on
to
ashley's
item,
I
sure
actually
like
to
demo.
C
A
So
well,
while
ashley
is
researching
his
system,
let's
do
this.
The
next
item
is
actually
from
and
ashish.
Let's
talk
about
talk
a
bit
about
wasn't
plug-in.
Instio
is
a
new
thing
which
was
just
recently
added
to
istio,
and
it
is
so.
A
A
So
basically,
what
this
particular
a
wasn't
plugin
interface
is
allowing
allowing
to
do
is
extend
the
functionality
of
sd
proxy.
So
here
they
are
not
actually
mentioning
that.
Are
we
extending
onward
proxies
ability
or
a
generic
is
your
proxy,
and
that
is
probably
we
would
get
to
know
because
because,
because
what
is
along
is
sort
of
inhibiting
us
to
do
from
is
specifying.
Where
exactly
do
you
want
to
drop
the
filter?
So
you
can't
what
you
can
actually
specify
using.
A
This
particular
interface
is
where
in
the
generic
filter
chain-
and
when
I
say
generic,
I
basically
mean
that
assume
that
there's
a
filter
change
in
that
filter
chain.
There
is
a
huge
blob
like
there
is
a
huge
block.
You
don't
know
what
what
it
exactly
does,
but
you
know
that
there
is
an
authorization
filter
in
the
in
the
beginning
of
it
and
a
huge
block.
It
doesn't
matter
which
block
it
is
it's
onward,
filter,
string
and
then
the
last
stage
of
that
particular
block.
A
So
you
know
the
beginning,
you
know
the
end
and
that's
where
that's
basically,
what
you
can
actually
choose
here
with
that?
Basically,
that's
all
choice.
You
get
that
you
want
to
drop
the
plugin.
You
don't
get
to
specify
that.
Okay,
you
have
run
my
pcp
filter
now
go
ahead
and
drop
my
filter.
This
is
the
stage
I
want
to
capture.
Probably
you
want
to
capture
it.
After
http
processing
has
been
done
so
yeah.
You
won't
be
able
to
specify.
A
You
won't
be
able
to
specify
using
this
particular
api
that
we're
in
that,
but
where,
in
one
web
on
the
filter
screen,
you
would
want
to
drop
this
filter,
but
the
but
the
thing
this
particular
api
is
bringing
is
support
for
version
modules.
That's
something
available
in
online
proxy
as
well,
but
generic
oci
images
and
that's
that's
quite
basically,
that's
quite
important,
because
that
basically
means
that
you
can
drop
in
almost
anything
anything
that
is
oci
image
compliant
and
almost
everything
can
view
cmh
compliant.
A
If
you,
if
you
package
it
properly,
so
what
they
are
supporting
is
you
can
bring
in
your
oci
image.
You
can
bring
in
local
file
system
and
that's
actually
supported
by
onward
as
well
as
well
as
you
can
drop
in
http
as
well
as
https.
A
Based
was
the
modules
the
pitch
is
again
supported
by
onward,
but
it's
not
it's
not
easy
to
configure,
but
apparently
it's
easy
to
configure
using
this
their
new
particular
area,
again,
shadow
56
verification
and
all
that's
something
that
was
already
supported,
but
again
what
they
are
building
on
top
of
this
so
onwards
supports
was
extension,
wasn't
based
extension,
it
could
be
remote,
it
could
be
local
file
referencing,
but
what
it
does
not
support
is
basically
something
like
image
for
policy
or
which
is
which
is
mostly
like,
which
is
a
construct
that
is
usually
supported
by
an
orchestrator
which
onwards.
A
Not
so
one
web
doesn't
support.
Something
like
that
which,
because
is
a
control
plane,
is
a
service
mesh.
It
controls
your
cluster,
so
this
is
something
again
supported
by
istio.
How,
and
so
the
thing
is
that,
because,
because
this
is
a
this
is
a
nice
api,
this
is
being
supported
by
istio,
it's
built
into
istio,
so
this
is
of
extreme.
Basically,
this
could
be
extremely
useful
to
us
because,
right
now,
if
someone
has
seen
the
pattern,
let
me
try
to
open
it
up.
A
I
think
I
can
do
that
as
ashley
is
still
probably
spinning
up
his
machine.
So
let's
look
at
the
pattern
that
we
have
for
okay.
Do
we
have
that
pattern.
A
Yeah,
I
don't
think
that
is
here,
so
if
someone
has
a
link
to
that
particular
file,
I'll.
A
Yeah,
so
if
someone
has
seen
that
particular
pattern,
it's
quite
huge
where
we
are
trying
to
do
multiple
things
and
by
multiple
things.
It's
like
you
first
set
up
how
you
was
you
first
set
up
that
you
actually
want
to
drop
in
a
remote
filter.
You
would
not
have
to
that.
You
will
not
have
to
go
through
that
particular
process
if
you
want
to
refer
to
a
local
file
reference,
but
if
you
want
to
defer
a
remote
pattern,
in
that
case,
you
will
actually
have
to.
A
You
will
actually
have
to
go
through
that
particular
stage,
also
so
yeah.
So
the
actually
this
particular
pattern.
This
is
made
up
of
two
stages.
In
first
stage,
we
are
actually
doing
some
generic
stuff.
That
is,
we
are
setting
up.
We
are,
we
are
actually
setting
up
the
you.
You
can
say
you
are
said
we
are
setting
this
stage.
A
What
this
particular
stage
basically
does
is
that
we
are
specifying
that
our
remote
filter
is
going
to
come
in,
and
this
is
the
place
where
we
want
to
drop
it
and
that's
what
answered
before
and
those
things
are
actually
specifying
well.
The
second
part
is
actually
where
you
would
actually
reference
a
remote
file,
remote
voice
and
filter,
and
so
basically
this
is
a
lot
of
process
that
you
actually
go
through.
Just
in
order
to
set
up
was
a
module.
A
Now,
that's
all
good,
but
the
thing
is
that
if
you
would
actually
check
out
istio's
code
or
basically
underlying
implementation
of
how,
when
when
is
the
adapter,
will
take
this,
take
this
particular
thing
and
when
it
will
hand
over
the
responsibility
of
for
setting
this
up
to
istio,
if
you
would
go
and
look
at
the
code,
what
you
would
actually
realize
is
issue
doesn't
actually
do
much
it.
It
does
so
from
say
1.9.
It
actually
inspects.
A
Okay,
there
is
a
reference
to
http
based
module,
and
now
what
I'm
supposed
to
do
is
I'm
supposed
to
set
up
xdp,
so
that
onward
can
actually
onward
can
actually
pull
in
filter
put
in
a
remote
file.
So
that's
something
that
is
there
does
do
does
from
1.9,
but
other
than
that
it
doesn't
do
much
so
yeah.
It
doesn't
do
much.
What
it
does
is
it
will
take
the
value
it
will
just
drop
into
it.
A
It
will
just
pass
it
on
to
the
data
plane,
and
that
is
one
way
and
then
it's
on
voice
responsibility
on
where
it's
now
going
to
do
its
stuff,
and
it
can
feel
you
can
have
messed
up
configuration
you
can
have
yeah,
you
can
have
messed
up
configuration.
Multiple
things
can
actually
fail,
because
this
is
not
doing
its
validation.
It's
just
by
passing
it
on
to
no
it's
just
passing
it
on
to
onward,
and
d18
is
also
a
long
process,
but
this
wasn't
plug
in
the.
A
The
first
thing
is
that
it's
beneficial
to
us
it's
beneficial
to
to
measuries
implementation
of
filters
is
for
because
it
provides
a
stable
api,
and
second
thing
is
actually,
of
course,
it's
it's
supporting
a
lot
of
different
types,
including
horsey
image,
the
only
limitation
that
it
at
least
I've
been
like.
A
I
have
looked
into
it
that
how
it's
exactly
doing
its
thing,
but
right
now,
the
only
limitation
that
I
am
familiar
with
is
that
right
now
you
cannot
at
least
tell
that
we're
inside
the
onward
virtual
machine,
not
virtual
machine
in
the
onward
process.
You
want
to
drop
in
the
filter
other
than
that.
A
Two
weeks
ago
we
were
having
discussion
that
we
would
want
to
support
a
generic
configuration,
so
this
could
actually
sort
of
it
could
also
sort
of
sort
of
solve
that
particular
issue
partially
for
us,
because
now
they
have
a
generic
configuration
which
they
are
supporting
and
and
it's
not
at
least
as
far
as
I
have
seen,
it's
not
onward
specific
or
something
it's
just.
It's
it's
a
configuration.
A
It's
a
set
of
it's
it's
a
it's,
a
specification
sort
of
which
lets
you
configure
our
data
plane
and
that
data
plane
is
not
exactly
that
data
plane
is
basically
not
on
where
or
from
any
so
yeah.
Probably
we
can
also
if
we
extend
it
for
misery's
use
case.
A
What
we
can
do
is
you
can
we
can
actually
provide
a
similar
interface
for
linkedin,
probably
not
not
that
we
would
do
that,
because
linkedin
is
not
on
web-based
and
it
could
be
that
it
doesn't
allow
loading
wasn't
plug-in
or
something,
but
because
it's
not
wasn't
specific
it's
it
can
actually
load
in
ocr.
So
basically
the
thing
that
I
was
listening
was
that
this
could
be
a
our
generic
interface,
also
which
allows
configuring
any
data
plane.
So
yeah
seems
pretty
useful.
B
Yeah
in
part
yeah
in
part,
it
seems
pretty
reasonable,
based
on
the
assumption
that
you
know
like
before,
just
like
before,
before
1.12,
there
wasn't
really
much
added
there
or
istio,
wasn't
even
you
know,
was
even
verifying
validating
the
fact
that,
like
that
plug-in,
would
would
work
or
that
the
configuration
was
appropriate,
which
you
know,
which
is
beneficial
to
a
project
like
measuring,
assuming
that
they've
kind
of
it
looks
like
they've
continued
on
the
same
path
that
this
is
really
about.
B
Treating
envoy
generically,
providing
a
facility
to
retrieve
promote
binaries
and
define
both
the
configuration
of
that
remote
binary
of
the
filter
here
in
the
plug-in,
plug-in
config
and
then
just
where
it
hooks
into
the
chain.
Now
they've
gotten
opinionated
about
that
particular
chain
which,
which
is
great
for
istio,
may
be
a
good
thing
to
follow
to
just
help
users
to
help
provide
some
opinionated
defaults
for
users
in
terms
of
like
common
use
cases.
B
If
the
common
use
case
is
to
always
run
the
authentication
filter
first
and
then
run
whatever
you've
got
subsequent,
like
you
know
great,
but
but
and
then
providing
the
ability
to
go.
Rogue
is
what
they
should
have
called
this,
I
guess,
or
anyway
the
there's
one
bit
of
this-
that
I'm
curious
about
it's
not
entirely
obvious.
I
guess
it
yeah.
B
It
becomes
obvious
to
me
now
yeah,
so
so
the
use
of
oci
as
a
an
image
format,
wonderful
that
just
for
everyone,
who's
been
paying
attention
to
kind
of
the
challenge
that
we've
been
collectively.
B
This
api,
this
custom
resource,
like
it's
just
an
evolution
of
what
was
already
there
in
1
11
and
before
so
that's
putting
a
finer
point
on
the
type
of
construct
that
that
string
takes
and
not
I
don't.
B
I
don't
recall
if
currently,
if
that's
passed
as
it's,
not
it's
not
passed
as
a
secret,
but
yeah
they're,
basically
going
through
an
evolution
like
and
just
so
there's
a
couple
of
things
that
will
be
interesting
to
confirm
as
we
go
forward
like
a
is
there
any
validation
that
happens
prior
to
deployment
of
a
filter,
to
make
sure
that
things
aren't
going
to
fall
over.
B
Some
things
that
would
be
interesting
to
understand
about
the
behavior
of
the
system
or
is
the
level
of
dynamicism
that
is
occurring.
So
if
you
provide
how
many
times
can
the
plug-in
config
be
reloaded?
So
if
you,
if
you
load
the,
if
you
load
the
filter
once
and
then
you
go
and
want
to
do
it
again,
I
want
to
readjust
the
configuration
of
that
filter.
B
That's
very
much
so
part
and
parcel
to
the
use
cases
that
we've
been
describing
here
about
dynamically
adjusting
the
behavior
of
your
mesh
based
on
you
know,
real
time,
a
real-time
understanding
of
what's
going
on.
If
you
want
to
implement
class
of
service
and
have
that,
as
an
example
have
class
of
service
adjust
dynamically,
then
you're
going
to
want
to
update
your
plug-in
config,
ideally
without
loading
and
unloading,
the
plug-in
itself,
because
you're
in
that
particular
use
case,
it's
just
the
same
filter
that
you
want
to
use.
B
B
B
The
third
thing
to
understand
is
is
around:
is
it
still
around
dynamicism,
probably,
and
it's
around
the
pull
policy
it
really
has
to
do
with,
like
so.
B
Yeah,
like
there's-
I
I
don't
know
it's
not
clear
to
me
with
this
pull
policy
that
I
think
there's
an
underlying
assumption
here,
that
you
can
still
dynamically
load
and
unload
filters
on
the
fly,
hots
hot
swap
them,
while
your
workload
and
its
envoy
sidecar
are
still
deployed.
B
This
pool
policy,
though,
makes
me
it's
good.
It
makes
me
wonder
if
that
level
of
dynamicism
is
still
part
of
the
system
or
if
what
istio
is
really
expecting
is
just
that
you're
redeploying
your
you're
re-rolling.
B
B
Something
that
I
don't
know
is
evident
to
all
of
us
on.
The
call
is
how
the
mechanics
and
the
behaviors
around
potentially
more
than
one
filter
chain,
and
that
that's
an
ignorant
assumption
on
my
part
that
that
you
can
have
multiple
filter
chains.
The
phrasing
of
this.
The
way
that
this
is
described
makes
it
sound
like
that.
B
So
that
was
so
thanks
for
that
by
the
way
it
would
crash.
I
know
that
there
wasn't
a
lot
of
time
necessarily
to
kind
of
go
through
and
look
through
this,
but
of
the
description
that
you
gave.
That
was
really
helpful.
It's
very
encouraging
to
see
this.
B
This
evolution
happen
because
it
tightens
the
the
interface
to
your
point
about
reflecting
on
mescheri's
ultimate
desire
to
have
this
same
level
of
capability
for
other
envoy-based
service
meshes
that
may
or
may
not.
You
know
that
don't
necessarily
have
an
api
for
configuring
that
don't
officially
support
filters
like.
B
You
know
is
better
thought
through
than
what
we
have
thought
through
so
far
or
like
so
you
know
it's
a
great
place
so
envoy
itself,
oh
envoy
itself,
does
support.
Multiple
filter
chains
is
what
rudraksha
is
confirming
and
so
yeah
an
open
question
as
to
whether
or
not
istio
as
a
control
plane
allows
that
or
whether
or
not
you
know
how
that
would
be
specified,
because
you'd
have
because
conceptually
what
you
know
in
concept.
You'd
have
multiple
awesome,
plug-ins
crs
like
two
of
them,
potentially
that
both
have
the
same
selector.
B
B
The
thing
about
something
like
that
that
that
makes
me
think
maybe
they're
not
supported
or
not
well
supported.
Multiple
chains
is
because
there's
there's
a
lack
of
control,
at
least
expressed
here
in
terms
of
parallelism
versus
sequential
processing,
because
because
you
know
you
can
imagine,
you'd
have
a
chain
of
chains
if
you're
doing
sequential
and
that
doesn't
look
like
that's.
B
B
Anybody
else
have
comments
or
questions
about
that
like
that.
That's
really
intriguing,
that
that
level
of
those
level
of
capabilities
are
there.
If
you,
that
just
means
that
that
sophistication
only
adds
more
credence
to
the
notion
that
people
will
need
and
measuring
a
tool
like
meshri
to
be
able
to
understand
that
they've
got
plugins
and
where
those
plugins
are
that
those
plugins
are
in
a
sequence
and
a
chain
of
various
factors
that
affect
their
traffic
and
that
there
may
be
multiple
chains
and
then
multiple
sequences
within
those
chains.
B
A
So
it's
does
anyone
have
any
comment
or
any
questions.
C
Yeah
yeah.
So
after
that
I
didn't
actually
refactor
the
code
or
anything
because
I
didn't
know
whether
so
I'll
share.
C
C
Yep,
okay,
so
I
demoed
that
day
that
this
again,
this
column
I
haven't
removed,
but
this
column
is
just
so
that
I
don't
have
to
go
back
and
forth
between
the
json.
So
the
thing
is
that
I
have
added
two
extra
fields
into
the
json
into
the
json
output
that
we
get
when
we
hit
slash
api
patterns
and
slash
pattern
point
which
returns
the
entire
page.
C
So
for
each
pattern
we
can
have
these
two
fields
where
first
we
signify
that
either
we
can
support
that
particular
pattern
file
at
the
moment
or
not,
and
if
we
cannot
then
what's
the
error
and
currently,
as
of
now,
the
support
means
we.
We
are
going
to
check
for
three
stages:
basically
everything
above
the
above
the
basically
above
that
stage
where
we
actually
make
the
grpc
call.
But
we
don't.
C
Basically
we
don't
know
whether
the
particular
adapter
is
there
or
a
particular
adapter
can
handle
it
or
not,
but
everything
before
that,
basically
syntax
checking
and
all
that
all
can
be
handled
here
and
for
each
not
supported.
I
I'm
not
displaying
the
error
messages
here,
but
if
I
hit
slash.
C
I
think
it
this
is
the
correct
endpoint
yeah,
so
for
the
ones
we
can
support.
There
is
no
error
message,
but
the
ones
we
cannot
support.
There
is
some
kind
of
error
message
here
like
invalid
workload
of
type
hdmi.
Then
we
can.
Basically,
this
means
that
is
the
adapter
is
not
running
or
maybe
istio
or
any
such
component
has
not
registered
its
capability
to
handle
this
kind
of
this
kind
of
workload.
If
I'll
spin
up
s2
adapter,
like
I
did
that
day,
this
will
be.
C
This
will
be
supported
again,
so
we
can
have
some
kind
of
highlighting
to
highlight
the
ones
that
we
can
currently
support
and
gray
out
the
ones
that
we
cannot,
along
with
the
error
message
and
as
they
will
make
those
changes
from,
for
example,
as
they'll
correct
the
for
example.
C
Here,
if
I
remove
this
field
and
if
I
update
the
pattern,
this
does
not
comply
with
the
schema,
so
it
could
get
not
supported
now
and
along
with
that,
we'll
get
an
error
message
and
then
they
can
correct
it
and
basically
everything
that
that
will
that
that
will
happen
after
they
hit
that
deploy
button.
C
Everything
that
will
happen
after
they
hit
the
deploy
button
will
happen
before
will
happen
as
they
fetch
these
patterns
from
the
back
end,
and
they
can
see
all
of
these
things
in
the
ui
and
the
the
purpose
of
this
was
to
make
things
more
dynamic,
because
right
now
we
don't
have
that
notification
center.
That
can
make
things
a
lot
easier
to
know
what's
going
on
in
the
in
the
background,
so
because,
up
until
we
we
are
completed
with
that,
this
can
be
a
handy
tool
to
have.
C
So
that's
what
we
were
discussing
that
day
and
lee
suggested
that
which
was
quite
obvious,
but
that
we
can
have
these
basically,
the
unmarshalling
of
each
pattern
and
the
checking
of
each
pattern
can
be
offloaded
to
each
go
routine.
So
that's
what
I'm
doing
right
now.
So
it's
two
right.
Currently
it
takes
two
seconds
without
the
go
routines
it
takes
around
four
or
five
seconds,
so
yeah,
that's
something
and
I
have
to
refactor
the
code.
I
haven't
refactored
much
so
we
discussed
it
that
day.
C
Also,
if
someone
has
any
other
comments
now
or
something
I
haven't
made
the
pull
request
yet
because
I
don't
know
a
reason:
why
didn't
I
make
the
pull
request,
but
yeah.
C
Okay,
so
so
every
time
someone
reloads
the
website,
it
again
checks
the
current
status
and
against
the
server
in
the
left
right
whenever
you'll
hit
this
particular
endpoint
just
before
sending
the
output
to
you
just
before
this
add
pattern,
add
metadata
of
pattern.
Current
support,
I
mean
we
can
change
the
function
name,
but
basically
the
whole
response
is
a
byte.
C
It
takes
the
whole
response
as
a
byte
and
just
does
things
to
it,
and
those
things
are
adding
these
particular
fields
by
dry,
running
stuff
at
the
back
end,
using
the
database,
where
we
have
actually
registered
all
the
capabilities.
C
Okay,
so
it's
dynamic
like
okay,
okay,
so
it's
fine!
So
what
I
just
thought
is
like
if,
if
we
can
memoize
this
thing
because.
C
Thinking
about,
like
someone
may
not
have
east
adapter
running,
and
for
that
reason
it
should
have
not
supported
tag
right
there
so
yeah
that
that
functionality
will
not
be
achievable
if
we
are
going
to
memoize
and
only
change
the
status
or
when
the
pattern
is
edited
right.
So,
okay,.
A
So
I
think
so
I
just
you
have
added
this.
This
particular
attribute
only
in
api
slash
pattern
right
exactly
so
we
may
I'm
not
sure,
but
I
think
we
there
could
be
similar
need
of
in
the
slash
api,
slash
application,
endpoint
and
similar
need
in
slash
api,
slash,
filter
point,
although,
although
what
exactly
supported
filter
means
could
be
a
bit
different
than
what
supported
pattern
means
so
yeah,
but
I
think
that
we
may
have
a
similar.
B
Yeah
that
makes
perfect
sense
yeah.
I
think,
we've
a
long
kind
of
wondered
why
we're
implementing
three
the
same
thing
three
times
over
and
and
like
I
said.
I
know
that
sounds
funny
when
I
say
it,
but
but
like
it's
not
it's
also
not
obvious.
The
answer
isn't
necessarily
obvious,
based
on
some
of
the
different,
the
lack
of
our
foresight
to
what
some
of
those
differences
will
be,
and
so
but
yep.
So
this
this
same
capability
would
be
ideal
for
yeah.
B
B
What
do
you
say
cross
the
looks,
complex
cross-functional.
B
B
I'm
on
this,
since
we're
talking
about
this,
this
type
of
a
thing
with
respect
to
you
know,
dry
run
and
that
ability
do
we
have
course
we
don't
happen
to
have
a
another
portion
to
this.
Do
we,
in
terms
of
like
dry,
run
from
mesh
map.
A
So,
actually,
that's
something
I'm
spinning
up
badly
so
yeah,
so
we
I
suggest
in
like
just
to
elaborate
on
what
that
exactly
is
so
the
ability
to
dry
run
patterns
and
try
it
and
basically,
almost
everything
that
you
could
do
with
a
pattern
file
that
yaml
was
added
to
mishri
a
while
ago.
That
has
been
that
was
recently
that's
something
that
ashish
is
using
here.
That
was
also
something
that
that
was
something
also
something
that
is
hooked
to
mishma.
It
was
hooked
to
mishma
recently,
that's
something
I'd.
A
I
I'm
trying
to
demo
I'm
spinning
up
mishma,
so
yeah
it
is
integrated
into
meshmap.
Actually
so
now
that
gaurav
is
here.
G
I
was
not
yeah,
I
don't
have
much
of
a
bit,
but
I
was
not
able
to
work
proceed
with
the
github
import,
especially
I'm
clear,
with
the
approach
that
we
are
following
with
the
github
import,
but
in
the
implementation.
I
think
I'm
still
looping
myself.
E
C
C
So
so
there
are
a
few
changes
which
you
can
see
like.
Oh
I'm,
seeing
this
huge
pattern
so.
C
Probably
you
won't
be
able
to
see
that
okay,
that
is
like
very
huge
pattern.
Nevermind.
I
should
open
something
which
is
easy
to
see,
and
that
is
like
empty
ones,
very
good,
so
yeah.
So
getting
back
to
the
previous
thing.
There
is
quite
no
change
here,
but
I
have
recently
did
some
changes
which,
which
may
be
visible
right
here
so
yeah.
C
If
you
see
I
have
like
this
whole
window
is
fixed
because
whenever
you
are,
whenever
you
are
editing
this
pattern
file,
then
this
window
should
be
right
in
front
of
your
eyes,
because
that
is
kind
of
showing
you
how
the
how
your
changes
are
reflected
onto
the
yaml
files
like
the
actual
groundw
files-
and
this
is
like
a
dynamic
resizing
of
this
like
that-
is
like
a
good
experience
in
my
views,
point
of
view
so
yeah.
If
you're
seeing
it
like
the
way
it
should
be
so
yeah.
C
F
B
Yeah
yeah,
it
looks
like
one
of
the
things
that
you've
taken
into
account
here
is:
maybe
you've
incorporated
some
some
style
updates,
but
one
of
those
being
that
the
user
is
not
getting
any
number
of
levels
deep.
H
B
C
Like
that,
when
you
click
on
add
item,
then
it
will
definitely
go
so
I
don't
remember
where
it
was
happening.
There
was
some
something,
so
that's
definitely
not
having
here
once
you
click
on,
add
item
that
should
you
should
get
it
like.
B
What
what
what
are
you
showing
us
new
today,
abby
or
or
what
feedback
are
you
specifically
asking
people
for.
C
So
so
so
actually
this
dynamic
resizing
is
something
that
is
that
I
implemented
for
better
user
experience,
and
another
thing
is
like
there
is
some
breakage
in
this.
So
whenever
you
change
the
test
like
there
is
no
effect
since,
since
the
states
are
not
being
propagated
from
here
until
and
unless
a
new
state
is
being
rendered,
so
that's
the
bug
kind
of
that
we
are
getting
here
other
than
that
there
are
like
not
quite
big
changes,
just
minor
ones.
B
Who's
got
feedback
so
this,
so
this
is
in
measure
ui.
This
is
this
ultimately
becomes
the
really
powerful
any
service
mesh
configurator
right
now,
it's
under
an
area
called
patterns.
B
What
do
you
got?
What
do
you
all
think
like
there's
some
obvious
kind
of
aesthetics
to
update,
but
I
don't
know
that
that's.
C
A
C
C
So
there
yeah,
of
course,
some
most
of
them
are
there,
but
all
of
them
like
what
are
those
so
that
reached
to
hiding
this
thing
right.
So
that
is
one
one.
I
can
see
it
right.
You
know
so
it's
too
deep
and
then
okay
cool.
Let
me
give
you
some.
B
Feedback
separately,
so
this
is
this
is
good
and
it
sounds
like
nobody
else
has
this
on
feedback
necessarily,
but
there's
there's
a
number
of
specific
items
that
I'll
give
you
some
feedback
on,
and
so
it's
a
good
update.
Yeah.
I
see
that
the
colors
are
coming
through
in
the
drop
down
under
configure
by
the
way
just
to
clarify
or
confirm
the
level
the
level
the
number
of
levels.
B
You,
okay
and
then
part
of
what
you
were
saying
you
know,
there's
I'm
seeing
a
number
of
updates
since
last
time
that
you
had
demonstrated
this
that's
great.
But
one
of
those
is
the
notion
that
there's
a
some
some
dynamicism
included
in
the
vertical
height
of
the.
B
Okay,
good,
let's
see
if
we
can
that's
great
I'll,
be
thanks
for
that.
Let's
see
if
we
can
hit
a
couple
of
other
items
on
the
agenda,
so
yeah
so
definitely.
C
Please
write
it
down
on
the
slack.
I
will
be
doing
it.
Asynchronously,
like
today,
only
synchronously.
A
Yep
so.
B
So
what
is
what
do
you
like
doing
yeah
on
this
last
part
since
rudraksha
and
navendu
and
mario
are
here
with
three
of
them
and
piyush
isn't
here,
but
but
all
four
of
them
and
actually
ashish?
It's
like
they
all
have
they're
all
kind
of
iterating
around
something
something
related,
and
that
is
the
notion
that
we
have
github
actions
and
some
of
those
get
and
those
github
actions
need
to
be
used
and
can
be
used
for
testing.
We
also
have
existing
workflows
that
are
used
for
testing.
B
I've
noticed
mario's,
so
mario's
back,
but
so
what
workflows
do
we
have
today
and
how
can
we
be
using
those
as
part
of
mesherie's
release,
process
measures,
pull
request
process
so
today,
in
when
a
pr
is
submitted,
are
we
deploying
kubernetes
and
are
we
testing,
deploying
measuring
and
testing.
B
H
I
So
I
think
if
we
want
to
run
tests
in
the
pr,
then
using
the
current
actions
would
not
be
so
useful,
because
in
prs
we
would
want
to
check
out
the
pr
branch
itself
so
yeah.
I
don't
know
that
would
probably
need
like
just
some
script
in
the
workflow
itself,
rather
than
making
out
a
new
action
for
that.
But
if
we
also
have
other
needs,
then
we
can
probably
develop
an
action
for
testing
yeah.
C
The
test
I
was
writing
in
for
the
adapter.
I
haven't
reused
any
of
the
workflows
in
smp
action
because
because
I
don't
think
I
need
all
of
them,
so
I
can
go
over
the
logic
that
I
was
going
through.
So
I
I
started
up
a
mini
cube
cluster.
C
I
install
measuring
it
and
then
I
check
out
my
adapters
code
and
create
a
create
an
image
on
the
fly
and
then
inject
that
image
onto
the
deployment.yaml
and
then
basically
apply
those
manifests
using
cubecuttle
and
after
I
have
my
adapter
running,
I
use
mexican
pattern
apply
command
to
apply
a
pattern
which
basically
deploy
the
service
mesh.
So
all
of
this
it
does
not
require
any
of
the
core.
C
I
didn't
take
any
code
in
the
smp
action
and
I
don't
think
I
would
require
much
and
the
whole
workflow
is
not
too
big
either.
So
you
know
I
was
just
I
I
interjected
just
to
point
out
point
out
that
we
do
not.
I
don't
think
we
need
them
in
the
test.
C
That
would
that
we
would
write
for
for
the
adapters,
and
here
the
test
would
be
that,
after
deploying
that
particular
service
mesh,
the
way
I
was
thinking
to
do
it
currently
is
basically
to
basically
to
be
in
a
loop
and
to
constantly
check
whether
in
that
particular
name
space
where
I
deployed
the
service
service
mesh,
whether
each
pod
is
in
running
state
or
not,
and
if
it
it
doesn't
come
in
the
running
state.
B
Great,
so
for
the
three
or
four
of
you
do
you
want
to
clarify
the
goals
or
the
the
your
approaches,
so
you're
focused
on
test
on
adapter
testing
right,
so
you
need.
You
basically
need
the
same
thing
that
mesh
reserver
needs,
except
the
adapter
needs
to
be
dynamically
built,
whereas
in
a
mesher
server
it's
an
actual
server
that
needs
to
be
dynamically
built.
B
B
So
yeah
a
common
set
of
reusable
workflows,
one
that
either
builds
this
or
builds
that,
but
a
common
approach
and
workflow
that's
reused
and
that's
what
I
mean
by
the
actions
like
yeah.
Yes,
obviously
it
doesn't
make
sense
to
reuse
an
action.
If
it's
going
to
use
a
pre-built
binary
right,
we
can
do
that
for
on
a
nightly,
build
or
a
weekly
build
sometime
after
an
edge
deployment
has
and
when
there's
a
new
edge
release
available.
Sure
that
won't
work
for
mesri
ctl
as
a
component
in
terms
of
edge
release
testing.
H
C
The
script
is,
just
you
know,
messy
ctl
pattern
apply.
So
what
what
script
are
you
talking
about?.
H
Yeah
yeah
yeah,
like
so
one
issue
we
had
with
the
missouri
snp
that
scheduled
workflow
was,
we
were
using,
is
to
helm,
charts
and
deploying
it.
So,
instead
of
that,
we
can
just
use
measuring
ctl
and
the
pattern
that
deploys.
C
Yeah
yeah,
okay,
one
one
other
doubt
I
had
raksha's
on
the
call
if
he
knows
the
solution
for
it.
So
while
I
was
doing
it
in
order
to
communicate
with
the
with
the
mystery
server,
I
was
using
mini
cube
tunnel,
but
the
thing
with
the
mini
cube
tunnel
is
that
if
I
basically,
if
I
use
the
and
operator
and
to
you,
know
to
detach
it,
the
process
is
killed
in
the
the
process
is
killed
and
we
cannot.
Basically,
we
cannot
tunnel
anymore.
C
So
I
was
using
port
forwarding
instead,
so
I
get
the
ip
of
the
measuring
pod,
then
I
you
it's
kind
of
a
hacky
way,
but
I'm
using
the
I
take
out
the
ip
of
the
or
all
that
pod,
and
then
I
redirect
the
traffic
from
9081
localhost
to
inside
of
the
mesh
report
directly,
because
that
way
I
can
actually
detach
it
and
get
to
the
next
step
so
know
any
way
to
run
mini
cube
tunnel
in
a
detached
manner.
C
I
went
through
some
of
the
issues
that
were
already
opened
in
that
mini
cube
repo,
but
they
didn't
actually
have
any
solution
for
that.
One
of
the
solution
was
to
redirect
that
logs
to
slash,
but
that
also
kills
the
process.
So
do
you
know
any
other
way
or.
I
I'm
not
sure
I'm
using
slash
devs
last
night
in
both
of
the
action
and
it
seems
to
be
working.
Fine,
are
you
sure
it
kills?
The
development.
I
Okay,
so
might
need
to
look
yeah,
I'm
not
sure
can.
C
And
by
the
way,
while
we're
at
it,
when
we
are,
you
know
creating
actions
for
measuring
server.
What
things
do
we
need
to?
You
know
check
what
what
checks
do?
We
need
to
pass
to
make
sure
that
message?
Server
is
working
correctly
like
in
the
case
of
adapter
right
now,
I'm
considering
that,
basically,
a
service
mesh
should
be
deployed
successfully.
C
That's
kind
of
the
goal
that
that's
kind
of
the
task
it
has
to
fulfill
in
order
to
pass
what
are
the
tasks
we
have
in
message
service,
so
what
tasks
we
need
to
give
to
meshi
server
that
if
it
passes
those
checks,
then
we
can
be
sure
that
okay,
so
everything
is
working
correctly.
A
So
so
to
check
if
measuring
server
is
actually
working
correctly
like
and
the
most
basic
chip,
probably
we
can
do,
is
just
ping
it
so
just
like.
So
what
we
do
in
order
to
check
do
the
humanities
health
check
connectivity
check?
Is
we
just
check
for
its
version?
So
that's
one
thing,
but
that
isn't
actually
guarantees
that
mystery
is
a
mystery.
Server
is
working
and
it's
working
correctly
and
it's
going
to
work
correctly.
A
Probably
we
need
to
hit
our
route
for
which,
which
is
which
is
actually
which
relies
on
a
token.
That
is
basically
it
it.
The
action
needs
to
be
an
authorized
action
because
that's
is
probably
needed,
because
sometimes
sometimes
what
happens
is
that
we
have
bugs,
and
even
if
we
could
have
a
bugs
in
communication
with
provider
and
that
not
that
it
happens
frequently,
but
it
does
happen.
A
So
what
one
of
the
issue
that
that
I
recently
faced
was
that
if
you
are
not
setting
up
build,
attribute
build
environmental
variable,
then
the
request
to
the
remote
provider
is
is
being
so.
Basically,
there
is
no
version
getting
set.
The
version
that
is
getting
set
is
not
set,
so
promoter
gets
confused
and
doesn't
sense
the
right
capabilities
and
everything
gets
messed
up.
A
So
that's
something
so,
although
misery
server
other
server
is
working
perfectly
fine,
but
if
you
would
ask
it
to
do
to
perform
some
action,
it
probably
won't
be
able
to
so
so.
The
thing
is
that
the
thing
that
I
was
suggesting
was
probably
hitting
version
just
to
check.
If
machine
server
is
alive
and
it's
so
you
can
connect
to
it,
you
can
its
server
is
running
it.
The
process
is
not
dying
and
something
like
that,
but
to
verify.
A
If,
if
most
of
the
action
may
go
through
or
may
not
go
through,
you
would
probably
want
to
hit
a
route
which
is
behind
the
auth
hole.
That
is
something
like
what
could
be
the
route
which
is
which
is
either
important,
which.
B
I
think
I
think
yeah,
but
I
mean
if
measure
server
is
responding
with
its
own
version
number
system
version.
Then
I
mean
that
doesn't
flex
much
code,
but
that's
just
a
basic
ping
test
to
say
the
server
is
up
and
reachable
and
and
then
the
rest
of
the
testing
can
ensue
like
if
something
more
complex
is
chosen,
then
that
may
be
a
false
negative
in
the
regression
tests
in
the
tests
that
are
being
run.
D
Should
we
take
a
look
at
the
existing
workflow
yaml
file,
yeah
yeah.
B
B
When
you
put
up
your
pr
for
the
test
for
measuring
adapter,
I
think
that'll
be
there'll,
be
some
good,
reusable
components
in
there.
A
So
just
a
solution,
I
think
we
can
so
one
of
the
harmless
tests
could
be
just
hitting
system
version,
and
after
that,
probably
we
can
get
provider
capabilities.
The
the
reason.
A
The
reason
could
be
that
actually
just
to
just
to
ensure
that
not
that
so
one
of
the
thing
that
can
go
wrong
is
that
the
mystery
server
doesn't
communicate
to
the
provider,
but
that
that
won't
be
reflected
in
the
capabilities,
and
I
think
that's
that's
fine-
that
that
the
remote
writer
was
down.
But
if
the
capabilities
endpoint
itself
is
not
returning,
the
correct
data
and
by
correct
data
basically
mean
that
it's
referring.
A
It
is
different,
referring
to
version
like
not
set,
or
something
like
that,
then
basically,
that
particular
endpoint
sort
of
will
help
if,
if
mystery,
in
its
interaction
with
remote
rider
is
working
fine
and
if
there
is
no
issue
from
mystery's
side,
the
reason
actually
I'm
emphasizing
the
not
set
issue
is
because
I
recently
faced
that
one.
That
is,
if
you
are
not
specifying
the
build,
then
actually
actually
remote.
Foreign
communication
is
almost
collapsing
because
extension
are
not
coming
in.
A
Capabilities
are
not
loading
in
properly
just
because
version
was
not
set,
and
that
can
happen
quite
often
so
yeah.
That's
pretty
much
my
suggestion.
B
Just
that
yeah,
I
don't
know,
as
you
think,
through
this,
the
s
measuring
staging
is
there
as
another
remote
provider.
That
is
an
available
resource.
I
don't
know
how
that
factors
in.
A
So
I
think
we
are,
we
have
covered
all
of
the
items,
at
least
the
folks
that
were
on
the
call.
So
we
are
about
time,
let's
wrap
it
up,
it's
a
wrap.
H
Just
a
second
like
ashish,
when
you
like
work
on
your
the
workflow
that
you're
working,
could
you
ping
me
like?
Could
you
share
the
link
here.
C
C
C
I
think
it
should
rudrak
just
gave
me
the
link
where
we're
actually
but
he's
using
it
successfully.
So
if
he
can,
I
don't
know
why
am
I
having
those
issues
so
yeah,
most
probably
yes,.
H
Okay,
okay,
sure,
okay
yeah,
you
can
bring
me
yeah.
Thank
you.
A
Let's
conclude
the
meeting:
let's
thank
you
crash
yeah.
This
is
great.