►
Description
[SIG-Network] Ingress NGINX Bi-Weekly Meeting for 20230316
A
Hello
good
evening,
good
afternoon,
everyone,
my
name,
is
James
strong.
This
is
the
Sig
networking
Ingress
in
genetic
subproject
meeting
for
March
16th,
which
means
it
is
a
cncf
meeting
and
it
is,
it
is
to
the
cncf
code
of
conduct.
If
you
have
any
issues,
please
report
those
to
myself
or
to
Ricardo
or
to
the
Sig
networking
chair
So
today
we're
going
to
recap
what
we've
done.
A
Let
me
go
ahead
and
recap:
the
the
issues
I
think
the
big
thing
we're
going
to
review
today
is
Sean's
going
to
do
a
demo
of
the
open
Telemetry
and
we're
going
to
go
through
our
normal
Cadence
with
the
meeting.
This
is
only
several
of
us.
I,
don't
see
any
new
members,
so
we'll
go
ahead.
A
I
can
already
say
before
we
go.
Look
at
the
action
items.
I
had
a
work
meeting
a
work
Summit
and
I've
only
been
focusing
on
trying
to
get
the
new
images
updated
with
the
cves.
So
none
of
the
action
items
that
are
on
the
last
meeting
I
know
I
haven't
intended
to
and
Ricardo
is
not
here.
So
the
only
one
we're
going
to
probably
do
is
the
the
demo,
so
we
need
to
again.
A
This
is
why
I
put
it
on
the
list
is
to
hold
us
a
little
bit
more
accountable,
and
it's
doing
exactly
that.
So
hopefully,
our
next
meeting
I've
had
at
least
check
some
of
these
old
action
items
off
with
that
it
is
yeah.
It's
316.,
okay,
I
thought
I
was
looking
at
the
wrong
one.
Sorry
folks,
so
DeSean
I,
think
we'll.
A
C
B
To
enable
one
we
have
to
do
two
things:
one
is
to
have
this
open,
Telemetry
enabled
in
the
values
Yama
and
the
other
thing
is.
We
have
to
add
some
options
to
the
config
map,
controller
config
map
and
the
most
important
ones.
Are
this
one?
This
collector
host
and
port
and
for
demo
I
will
use
open,
Telemetry
collector,
but
anything
that
accepts
grp,
otlp
grpc
should
work
now,
I've
tried
it
with
Jagger
and
graph
and
grafana
okay
and
some.
C
B
C
B
Have
a
demo
app
that
is
behind
Ingress
and
I,
just
make
a
curl
call
to
that.
It
does
nothing.
It
just
prints
returns
some
hello,
hello,
nginx,
that's
it
and
then
we
can
see
the
traces
in
younger.
Let's
just
look
for
new
ones,.
B
So
this
one
comes
from
the
module,
open,
Telemetry
module
and
it
has
these
attributes,
tags
and
so
on-
and
these
are
this-
the
other
ones
are
from
my
demo
app.
So.
C
B
B
A
With
the
modules
that
we
have
built
into
the
nginx,
if
I
needed
to
do
a
migration,
do
we
have
a
migration
path
upgraded
or
do
we
have
a
migration
path
for
folks
to
use
this
new
Telemetry
module.
B
So
this
is,
it
had
to
be
so
the
old
user
they
had
to
use
open
tracing.
Okay,
then
we
are
different
from
open
tracing
so
yeah.
They
have
to
do
some
something
for
the
migration,
which
means
they
have
to
change
this
config
map,
plus
the
values
gamma.
A
So
we
probably
because
I
know,
one
of
the
asks
has
been
to
remove
the
Jaeger
code
that
we
build
into
our
nginx
modules,
all
right,
the
nginx
base
image.
So
that
would
be
a
breaking
change
right.
So
we'd
probably
want
to
bump
things
to
like
a
2-0
if
we're
removing
those
modules
and
forcing
folks
to
use
the
open,
Telemetry
module.
B
Okay
yeah,
but
maybe
we
can
reuse
their
options
in
the
code,
but
I
don't
know
if
that's
the
right
way
to
do
it.
Yeah.
C
A
A
So
that's
I
mean
that's
something
we
can
think
about.
So
we
can
get
this.
We
can
get
the
documentation,
PR
accepted,
I!
Think
we've
already
got
all
the
module
is
already
there
right:
yeah,
yeah,
okay,.
A
A
A
This
out
there,
and
then
we
can
start
having
a
discussion
about
asking
people
to
move
towards
using
this
module.
B
A
And
then
once
we
actually
do
remove
it,
that's
a
breaking
change,
we'll
bump
the
chart.
Unfortunately,
there's
no
way
for
us
to
know
who
is
using
any
of
this
functionality
we'll
have
to
put
a
notice
out
somewhere
on
the
thoughts
in
the
timeline,
but
okay
yeah.
No,
it.
A
B
B
A
No
I
was
if
we
want
to
get
this
in
the
165
or
though
I'm
thinking,
we're
gonna
do
one
seven,
because
we're
bumping,
golang
and
we're
updating,
Alpine
and
I
know,
there's
probably
some
other
features
in
there
and
I.
Don't
want
folks
automatically
rolling
us
out
so
I
think
I'm
gonna
go
with
170
and
if
we
add
the
mod
this
feature,
I
think
that's
that
also
helps
justify
the
170.
A
C
A
C
A
A
C
C
If
you,
if
you
bomb
open
SSL
as
well,
then
it's
like
double
make
sense.
A
Yeah
there
is
an
open
new,
open,
SSL
release
patch
release,
three
one:
it's
not
trickled
down
into
Alpine
Maine.
Yet
so
there's
that
one
as
well,
okay,
awesome,
cool
cool
is
Sean
anything
else
you
wanted
to
show
us.
No.
That
was
you.
B
C
A
B
B
The
pr
is
for
controller
the,
but
the
example
is
in
documentation,
but
it's
not.
The
code
will
not
be
in
the
main
repo.
It's
just
I
would
say
it's
I've
done
it
the
same
way
as
as
it's
done
for
open
tracing,
so
open
tracing
documentation
says
that
you,
you
could
go
to
some
example
that
is
hosted
by
someone
else.
A
A
B
A
A
A
B
I
shall
I
stopped
my
sharing
yeah
yeah
you're,
fine
I
just
stopped
it.
A
I
think
what
24
long
Let's
do,
let's
review
the
the
issue
triage
and
just
go
through
some
of
the
newer
issues,
and
then
we
can
discuss
all
of
the
issues
with
the
CI
that
you
know
how
I
broke
everything.
A
Okay,
I
already
I
saw
you
commented
on
this.
One
I
also
commented
and
yeah.
B
A
A
I
like
jumped
on
there
all
right,
we
really
need
to
put
a
ask
out
there,
for
we
need
documentation
on
all
of
the
client
IP
preserve
IP
stuff.
We
get
that
question
all
the
time
and
I
think
we
just
need
a
dock
that
clarifies
it
for
folks,
because
there's
It's
a
combination
of
if
you're,
using
the
proxy
protocol
or,
if
you're
not
and
then
how
do
you
actually
send
those
back.
So
this
is
probably
another
one
of
those
issues.
C
So
at
least
this
one
right,
they're,
basically
coming
with
a
very
odd
question
as
in
they
are
not
compliant
with
the
kep
and
they
just
try
to
apply
the
non-cuberative
infrastructure
idea.
So
this
guy's
got
a
h,
a
proxy
in
front
right
and
so
why,
let's
say
proxy
can
for
packets.
He
used
to
want
to
start
headers
and
everything
and
he's
not
configuring.
It's
a
proxy
for
proxy
protocol
and
stuff
like
that
and.
B
C
So
so
yeah
we
do
need
to
have
documentation,
but
on
the
other
hand,
how
Ingress
works
or
how
load
balance
service
type
works
and
all
that
is
completely
out
there.
So
this
is
how
you
have
to
deal
with
it.
Then
we
have
to
deal
with
documentation
on
many
many
things
so
yeah.
We
should
make
an
effort.
C
B
A
Updating
yeah
so.
A
C
So
if
we,
if
we
want
to
be
like
diplomatic
or
you
want
to
be
not
if
you
want
to
be
precise
on
support
that
we
are
providing
I,
think
we
should
open,
let
others
answer
or
whatever
and
after
we
put
links
in
there
for
documentation
or
we
should
kind
of
start
following
two
policies
or
two
things
that
are
already
well
defined
properly.
A
Them
I,
don't
know,
yeah
I,
don't
know
about
automatically
closing
tickets
if
they
don't
provide.
The
information.
Just
ask
I,
know
you're
really
good
at
that,
and
it
is
annoying
but
which
we
shouldn't
just
close
the
tickets.
If
they
don't
have
that
information,
so
just
ask
for
it.
I
am
inclined
to
close
this
one
with
saying
that
this
is
how
Ingress
works,
and
if
you
don't
have
proxy,
you
can
figure
to
do
that.
Then
configure
it
to
pass
the
information
and
it
only
works
with
cluster
local.
A
A
B
A
C
C
A
A
A
A
Sorry
guys
you
watch
me
type
I,
just
don't
wanna
add
this
to
my
continue,
adding
it
to
my
back
load,
we'll
wait
to
see
if
he
responds.
A
A
C
A
No
he's
just
asking
if
we
can
add
this.
This
is
one
of
those
where
I
think
the
pr
is
easy
enough
to
where
someone
could
actually
could
do
this,
but
yeah,
it's
probably
not
there
in
the
Helm
charge,
ammo.
B
C
A
A
They
need
to
add
the
value.
This
is
a
pretty
straightforward,
PR
that
I
think
anyone
could
I'll
add
the
Help
Wanted
I
can
accept
the
triage,
but
I'm
not
I'm,
going
to
put
it
on
the
backlog,
but
I
also
think
that
Cindy
wants
it
enough.
He
can.
He
can
also
put
in
the
pr.
A
A
A
Statement,
I'd
gladly
accept
this
PR.
That's
rude
right.
A
A
Hey
look
at
that.
I
was
right
on
all
of
the
priorities.
Sorry,
you
said
we
have
got
a
couple.
You
want
to
look
at.
C
Whoa,
basically,
what's
happening
is
I'm
surprised.
What
he's
saying
is
that
we
can
Con
in
the
options
to
configure
the
amount
of
traffic
you
can
send
to
canopy.
There
is
a
there
are
two
parameters.
One
is
what
percentage
you
want
to
sign
and
there
is
another
field
called
total
weight
total.
So
he
says
if
he
chooses
the
minimum,
which
is
one
it
has
to
be
numerator,
it
doesn't
allow
fractional
values
there.
C
Then
that
is
too
much
so
he
needs
to
send
less
than
one
maybe
0.5,
and
all
that
and
the
field
doesn't
take
fractional
numbers.
That's
one
thing.
So
if
he
increases
the
weight,
total
parameter
to
1000
and
then
uses
in
type
5
or
1,
then
that
kind
of
meets
his
requirement
as
because
they
have
a
very,
very
big,
very
big
volume
of
traffic
coming,
so
they
can
achieve
Point
first
by
point,
five
percent.
A
Oh,
is
this
a
difference?
I
thought
these.
A
C
Was
to
allow
the
fractional
part
and
make
the
weight
total
value
more
than
100
and
all
that,
but
it
was
not
implemented.
There
was
intention
that
in
future
we
will
do
that
right.
Now,
it's
not
there
so
right
now,
whatever
you
do,
it
takes
the
way
total
as
100
and
the
minimum
you
can
set
is
one,
and
if
somebody
is,
one
percent
is
much
more
than
they
want
to
send
to
candidate.
They
have
no
choice
and
and
kind
of
explained
everything
and
put
a
list
view.
The
only
Pro
only
problem
is
I.
C
A
C
Right
so
there's
a
lot
of
mess
going
on
with
Canary
there's
one
more
issue
on
Kennedy
but
I'm,
just
not
certain.
If
his
I
mean
how
good
is
bras
or
a
good
leads
and
things
so
up
to
you,
but
this
is
important
because
you've
done
enough
work
and
he's
given
the
right
information.
So
we
know
exactly
what
the
problem
is.
A
A
C
You
can
ask,
but
he's
saying
something
very
important,
he's
saying
this:
doesn't
work
thousand
doesn't
work
at
all,
because
if
you
see
his
issue,
he
has
actually
given
data
about
the
test
and
the
number
of
requests.
So
it's
very
clear
that,
regardless
of
in
55
000,
it
will
always
be
five
percent
of
the
whole
number
percentage.
C
A
Yeah
no
I,
that's
fine
I
mean
the
test.
Cases
are
passing
I
just
want
to
make
sure
that
the
what
he's
changed
has
a
test
case
and
it
actually
solves
this
problem.
So
if
it
solves
this
problem
and
when
he
adds
the
test
case
or
if
that
one
does
because
I'm
I'm
like
you
long,
it's
I've
not
done
any
of
the
canary
deployments
or
looked
into
it.
But
if
he
adds
a
test
case
or
if
it
says
that
it
passes
his
test
case,
then
I
think
we
should
be
fine
to
go.
C
A
A
A
A
A
This
is
the
one
that
takes
too
long.
Sorry,
yeah.
C
A
A
A
A
A
A
All
right
we
got
10
minutes.
We
did
a
lot
more
triage
than
I'd
planned
to
so
sorry
about
that
long.
What
do
we
need
to
do
to
get
the
CI
stuff
fixed
and
get
that
released
out
and
release
out.
B
C
Somewhere
is
working
on
it,
so
there's
no.
This
is
not
a
CI
problem
at
all.
It's
basically
a
problem
with
bad
image,
so
we
need
to
pick
which
image
to
add
domain
now
and
which
to
exclude
so
HTTP
has
been
ruled
out
test
Runner
image
with
the
individual
PR
audience
domain.
The
new
nginx
base
image
today
morning
upload
so
that
new
engine
X
base
image
already
also
in
so
the
remaining
ones
are
open,
Telemetry,
CF,
SSL
and
those
ones
so
I'm,
working
with
the
somnia
to
check
them
one
by
one.
C
C
C
That's
completely
ruled
out.
We
never
ever
are
going
to
use
a
new
image
on
HTTP
bin.
The
other
problem
was
cfssl,
probably
in
the
error,
when
I'm
running
tests
locally
I'm
finding
errors.
The
first
error
that
the
failure,
the
test
that
you
fails
first,
is
the
TCP
Services
test
and
something
you
do
around
external
name
and
it
points
to
an
ocsp.
So
I'll
I
will
do
more
testing
to
know
if
the
CF
SSL
image
is
worth
upgrading
or.
A
C
And
then
the
open,
the
small,
benign
error
about
the
jinko
CLI
being
wrong
that
we
fixed
by
putting
in
a
dedicated
PR
only
to
update
the
test
Runner
image
so
now,
yeah,
amazing,
okay,
yeah.
So
three
of
those
seven
images
taken
care
of
the
rest,
one
by
one
formula,
is,
is
looking.
B
A
C
B
C
If
you
want
to
ignore
these
and
move
ahead
with
the
next
release
that
you're
thinking
the
nginx
base,
image
is
already
in
place.
The
test
winner
is
in
place.
So
if
you
make
a
new
release
now
the
controller
will
build
with
new
patched,
go
one:
two:
zero
and
317
to
Alpine.
Let.
A
A
To
get
released
to
community
or
Main,
then
I'll
release
the
new
image,
or
at
least
kick
start
the
process
for
the
build
and
everything
because
I
wanna
I
don't
want
to
have
to
rebuild,
go
through
the
whole
process
again,
if
31
is
going
to
get
released
soon.
So
we
can
wait
on
those
images
that
you
were
just
talking
about,
but
yeah
I
don't
want
to
build
a
new
controller
if
three
one
is
going
to
get
released
like
the
day
after
so.
B
A
We're
gonna
have
to
rebuild
nginx
again
to
get
the
three
one
and
then
build
the
controller.
So
I'll
just
wait
and
see
I'll
wait
on
three
one
and
then,
in
that
time
frame
you
guys
will
probably
get
all
the
other
images
fixed.
So
I'll,
let
you
and
what's
her
name
again,
I'm
sorry,
I
didn't
I
I,
see
it
but
I
don't
know
Samia,
so
yeah,
yeah,
okay,
I'll!
A
Let
you
to
work
on
getting
those
images
fixed
and
then
I'll
just
wait
on
three
one
and
then
we'll
have
everything
in
there,
and
that
also
gives
me
time
to
look
at
the
open,
Telemetry
and
get
it
actually
merged
in
so
we'll.
A
C
C
Yeah
specifically,
if
you
want
to
bump
the
minor
version
to
one
seven,
we
yeah
it's
a
really
really
prudent
thought
to
wait
until
open
s3.1.