►
From YouTube: SONiC DASH Workgroup Community Meeting June 21 2023
Description
Kristina to download all slides from SmartNIC Summit & store - done
Kristina presented view of DASH project Issue board - done
Kristina to coordinate gaps in ReadMe.md; Andy provided content re: gap analysis (below):
Link to Andy's article describing gaps in BMv2 and P4-DPDK targets that they are missing in order to be a good DASH P4 reference model: p4-guide/dash at master · jafingerhut/p4-guide · GitHub
https://github.com/sonic-net/DASH/pull/392
Kristina created a new playlist in the DASH YouTube Channel for conference content, with 2 recordings added - done
Headers and Libraries question - Vijay -this relates to Issue388 - SAI from reshma-intel move to Master (which is being worked on at Intel. An OCP SAI fix is needed, then they can move the pointer, which is needed for DASH (for VXLan). Reshma will prioritize, hopefully 1 or 2 weeks out.
SAI Meta in bmv2 discussion
ACL groups in Metering policy
A
To
the
dash
Community
call
73rd
meeting
here,
so
I
wanted
to
share
with
everyone
a
couple
of
pictures.
If
you
have
a
second,
we
went
to
the
smartnext
summit
last
week
and
so
I
don't
know.
If
you
can
see
my
screen,
I
wanted
to
show
a
couple
of
pictures
of
a
couple
of
us.
There,
Me
reshma
hanif
Andy
Michael
zigmund,
so
this
is
kind
of
what
it
looked
like
there.
While
we
were
delivering
the
sessions
and
listening
to
everyone
else
and
watching
the
Keynotes.
A
A
So,
anyway,
a
couple
of
thoughts:
there
were
a
few
sessions
recorded
Gerald
had
one
where
he
had
given
a
presentation
at
a
cloud
launch,
and
he
we
did.
The
recessions
weren't
recorded
there
at
the
summit,
but
there
were
a
couple
of
different
ones
that
were,
and
so
I
was
thinking
inside
of
the
Dash
YouTube
channel
I
am
probably
going
to
create
a
another
playlist,
so
you
guys
know
we
have
the
different
playlists
I
have
to
navigate.
A
We
have
high
availability,
we
have
behavioral
model
and
then
the
regular
Wednesday,
Community
call
and
I
think
I'll
make
one
just
for
conferences
and
when
we
have
recordings
at
conferences
where
our
topic
is
discussed,
maybe
I'll
pop
them
in
there,
so
that
people
can
see
those.
So
just
an
FYI.
If
that's
okay
with
everybody
that.
C
A
Yeah
yeah
yeah,
so
these
are
the
playlists.
We
have
Community
call
High
availability
and
behavioral
model,
and
so
I
could
just
add
one
more
just
for
conferences.
I
can
take
an
action
item
to
do
that
in
the
next
day
here
so
and
upload
a
couple
I'll
put
those
in
the
meeting
notes
that
go
out.
I,
don't
know
if
we
have
anyone
new
on
the
call,
but
you've
all
seen
the
meeting
notes
that
go
out
yeah.
A
A
C
A
A
A
E
Header
files
in
the
ocp
repo-
and
there
are
sign
header
files
in
the
dash
repo
which
are
Auto
generated
from
the
BMV
too
and
I
think
the
recommendation
is
to
build
the
vendor
lipsi
with
the
ocp
sign
header
files.
But
what
we
are
observing
is
when
we
try
to
run
the
side
Thrift
test
in
the
dash
repo
with
the
lips
eye
with
our
lip
side,
which
is
linked
with
ocp
site.
There
is
a
mismatch
in
The
Meta,
because
I
think
the
ocp
side
lags
behind
the
dash
repo
PSI
version.
E
I.
Think
basic
question
is
what
is
the
what's
the
best
practice
for
linking
which
size
should
we
linked
to
when
we
build
our
vendor
lip
side
and
also
I
think
this
is
going
to
be
a
a
an
ongoing
problem
right
before
cpsi
and
the
dash
I
are
not
in
sync.
So,
should
we
be
more
proactive
about
pushing
the
ocp
side
headers
and
keep
keeping
it
in
sync
with
the
dash
version.
E
G
B
B
Yeah
we
already
had
made
some
changes
a
long
time
ago
to
get
the
vxlan
fix
in
the
P4
Lang
and
that's
added
into
PSI
in
between
all
our
PTF
changes
into
ocpsi
and
our
final
vxlan
fix.
There
was
another
fix
that
went
in
which
actually
broke
all
of
the
dash
PTF.
So
we
need
to
fix
that
and
it's
a
very
large
change.
It
seems
so
trying
to
work
with
the
prince.
Is
it
okay?
If
I
include
Richard
into
this
discussion,
maybe
I'll
call.
B
I
H
I
B
C
B
C
F
So
if
the
whole
Sonic
Image
is
built
from
the
ocp
master
or
samosity
release,
vendor
is
supposed
to
align
to
that
even
today.
Without
that
we
have
the
same
kind
of
the
similar
thing.
There
is
a
master
sign.
Then
there
is
a
side
release
and
Sonic
is
sitting
on
the
side.
Release
no
one's
asking
vendor
to
to
put
a
library
for
a
or
the
latest
master
I
see.
I
C
C
B
B
A
A
A
C
A
E
E
So
right
now
yeah
we
had
to
like
recompile
our
website
with
the
dash
repo
version
of
the
PSI
and
then
get
the
then
run
it
for
the
side
Thrift
test.
But
again
we
have
when
we
are
trying
to
run
with
the
full
Sonic
there,
because
everybody,
all
the
other
components,
are
using
ocp
side
like
cyrid
is
we
have
to
end
up
using
the
ocp
PSI
version,
so
we
are
having
to
like
maintain
two
versions.
H
E
Yeah
I
think
today
the
Sonic
build
image
is
pointing
to
the
tot
or
CP
site.
So
I
guess
it
just
pulls
whatever
is
in
the
master
or
cpci
version
yeah.
As
long
as
the
dash
headers
are
coming
to
the
ocp
side,
repo
and
the
dash
version
and
the
ocp
PSI
versions
are
in
sync,
except
for,
like
probably
local
changes,
which
somebody
might
be
doing,
then
I
think
we
should
be
good.
H
E
Okay,
yeah
I
think
just
a
follow-up
question.
Is
we
noticed
that
the
sign
meta?
So
in
the
in
the
side
report
there
is
the
same
meta
which
is
kind
of
a
generated
file?
Basically
the
resign
metadata
and
PSY
metadata
utils
undersci
meta
directory,
so
these
are
being
included
currently
in
the
VMV
to
libsi,
and
is
that
something
that
is
supposed
to
be
included
in
the
vendor?
Libsi
because
it
doesn't
seem
like
only
when
we
run
the
lipsi
with
the
scytheris
server.
It
requires
the
same
meta.
E
C
C
C
F
E
Yeah
I
think
so,
probably
Chris.
If
you
try
to
today,
BMV
to
libsai
is
getting
compiled
in
with
the
sign
meta
from
the
last
repo
and
Sky
Thrift
server
is
Happy
because
that
is
already
included.
But
if
you
try
to
run
the
same
side,
Thrift
server
with
a
vendor
lip
side,
then
it
complains
that
the
meta
symbol
cannot
cannot
be
found,
but
yeah.
C
Yeah
I
I,
don't
know
the
right
answer
to
that,
because
the
scientist
server
build
process.
One
thing
someone
else
you
know
came
up
with
all
of
that
and
just
kind
of
canned
out
of
the
side,
repo
and
as
far
as
the
bmv2
lib
PSI
house,
I
meta
got
into
that
I.
Don't
I,
don't
really
have
a
good
answer
for
that.
It's
been
so
long
since
I
looked
at
it
and
I
actually
built
upon
Marian's
original
work,
just
some
refactoring
in
the
build
process.
C
A
C
Issue
yeah
I
mean
I,
remember
having
to
create
the
build
step,
makes
make
SCI
meta
to
get
it
to
bind.
You
know
for
everything
the
link,
but
I
was
sort
of
dabbling
at
that
stage,
with
Marion's
original
work,
just
sort
of
let's
say:
refactoring
it
a
little
bit
and
trying
to
get
the
sci
Thrift
server
compiled
for
the
first
time
and
whatever
solution
that
came
up
with
finally
worked
or
maybe
there's
a
another
step
that
can
improve
things,
in
other
words,
this
stage,
where
there's
actual
vendor
lib
size
for
dash.
A
A
H
All
right
so
I
pulled
a
div
from
from
the
meter
PR
to
basically
do
the
integration.
You
know
the
PSI
and
integration
with
our
data,
pane
and
I.
Just
had
a
couple
of
questions
on
the
some
scale
on
the
mirror.
Policy
tables
is
also
the
mirror
policy.
Is
it
like
an
ACR
group
when
you
apply
it
to
an
enr,
you
basically
are
not
allowed
to
change
it.
It's
Frozen
that
I
guess.
Sorry,
that's
a
question.
E
E
H
It
is
documented
or
Echo
groups,
but
it's
not
for
these.
But
to
me
it's
similar,
especially
there's
the
churn
also
on
the
policy
table
because
of
prefix
lengths
right.
If
you
put
one
at
a
higher
priority
with
a
low,
very
small
prefix
Nance,
you
basically
hide
it
kind
of
hides,
potentially
some
subsequence
rules
at
low
priority.
So
there's
some
churn
there
as
well
right.
H
I
H
Yeah
and
those
are
very
as
opposed
fine
details,
I
suppose
like,
for
instance,
if
you
have
the
media
class
on
the
outbound
route
and
the
mapping
well,
there's
no
way
to
disable
those,
because
the
middle
class
is
just
a
16-bit
value.
So
how
does
that
work
with
does
if,
if
you
encounter
the
rod
or
an
eni
well,
then
he
and
I
better
have
that
bucket
ready?
You
know
that
sort
of
things.
E
H
So
forget
this
I
didn't
really
ask
the
question.
The
question
is:
is
this
the
rat
and
the
mapping
have
a
mirror,
and
eventually
you
also
have
the
eni
router
policy
media
policy
result.
Potentially
that's
just
optional,
but
if
you
don't
have
that
one
anyway
and
if
you
don't
have
the
mapping
overriding
the
rods
middle
class,
the
middle
class
on
the
rod,
you
get
it
no
matter.
H
H
E
H
Class
is
just
a
16-bit
value,
which
is
unfortunate
because
there's
no
way
to
know
I
don't
have
a
meter.
So
then,
because
you
have
the
meter,
you
want
to
make
sure
the
data
plane
is
ready
to
ink
to
find
a
bucket,
or
you
know
that
sort
of
things
right.
There's
no
great
answer
for
that,
but
you
know
we
need
to
know.
E
Yeah
I
think
there
is
a
logic
in
bmv2
where
we
picked
the
meter
class,
where
we
do
the
zero
check,
assuming
it
is
that
zero
value
is
an
invalid
value
and
everything
other
than
non-zero
values
are
valid
values.
So,
in
this
particular
scenario
yeah,
you
would
be
setting
the
meter
class
on
the
routing
outbound
routing
to
zero.
But
you
are
right:
we
probably
have
to
explicitly
document
this
somewhere
right.
That's
what
you're
saying
okay.
A
H
E
H
E
E
H
Right,
so
the
general
question
is
as
far
as
counters
now
is
the
metering
that
you're
defining
the
pr?
That's,
that's
it
that's
what
we
have
and
there's
no
like,
if
you
the
logic
with
the
eni
raw
table
and
the
the
writing
middle
class
and
the
the
mapping
middle
class.
Those
are
the
are
those
the
only
ones
defining
meters
like
counters.
Basically
or
are
we
also
supposed
to
increment
pair
Rod
counters
pair
mapping
counters
pair
whatever.
E
I
G
A
Of
these
two
places
it's
either
in
it's
either
in
this
one
or
this
one,
I'm
gonna
pick
this
one
and
see
what
happens
the
counters
I,
don't
think
it's
in
this
one.
It
must
be
in
the
other
one
yeah.
Let
me
go
back
one,
it
might
be
in
the
packet
transforms
document
at
counters
and
it's
not
awesome,
but
this
is
where
we
started
I'll
put
in
the
chat
window.
H
A
H
H
A
A
Way,
no,
no
problem,
so
we
have
about
15
minutes
left,
so
I
showed
you
guys.
The
project
issue
board
and
I
was
thinking
of
just
one
more
thing.
Just
so
we
all
stay
on
the
same
page.
I
was
thinking
that
possibly
here
I
think
it's
in
the
pipeline
area.
I
was
thinking
of
possibly
creating
a
a
readme
file
in
each
BMV
too,
and
maybe
create
a
folder
for
dpdk
and
outlining
the
gaps
between
the
two
different
data
sets,
and
just
so
we
all
know
when
we're
on
the
same
page.
A
C
B
A
A
J
A
I
H
E
A
So
another
big
day.
That's
all
I
had
for
this
week
and
I.
Think
Vijay,
I
and
Prince
I
also
created
issue,
387
I'm,
not
sure
if
you
guys
saw
it
in
that
project
about
addering,
adding
scale
into
the
documentation.
I
tagged
everybody
here
and
I.
I
know
that
we
have
this
one
for
reshma
routes
and
scales
is
for
me.
A
A
C
I
just
pasted
a
link
into
the
chat,
the
people
who
are
starting
to
integrate.
You
know
vendor
lib
site.
You
might
want
to
take
a
look
at
it.
It's
a
way
of
using
the
dash
project
as
a
sub
module
in
your
own
projects
in
your
company
as
a
way
of
sort
of
holding
in
all
the
test
cases
Etc
and
maybe
modifying
the
the
make
files
to
suit
your
own
workflow.
But
it
might
be
interesting
to
some
people,
so
anyone
who's
trying
to
use
like
the
dash
generated
files
Etc.