►
From YouTube: Kuma Community Call - December 9, 2020
Description
Kuma hosts official monthly community calls where users and contributors can discuss about any topic and demonstrate use-cases. Interested? You can register for the next Community Call: https://bit.ly/3A46EdD
A
Welcome
to
the
the
monthly
making
meetings
of
the
community.
So
let's,
let's
quickly
start
this
here,
please
add
your
names
here
and
attend
the
list
so
that
we
can
for
track
record
of
what's
going
on.
I
just
quickly
wanted
to
go
to
the
events
and
already
spoke
about
this
so
many
times
but
okay.
Finally,
some
videos
are
available
for
those
that
are
interested
on
the
service
mesh
con.
A
I
have
linked
marco's
introduction,
opening
remarks
for
the
story
mesh
con
and
also
he
did
a
talk
about
command,
so
it's
multicultural,
multi
cloud
service
mesh
with
cncs
command
envoy.
These
are
the
links.
These
are
also
part
of
the
whole
the
whole
list
here.
So
if,
if
you're,
if
you
are
interested,
all
the
talks
are
actually
here,
so
if
you
want
to
go
and
check
things
about
web
assembly,
whatever
is
there
so
interesting,
interesting
also
from
the
main
cubecon
north
america?
I
have
selected,
I
have
put
here
links
to
two
videos.
A
So
if
someone-
and
I
definitely
I
was
on
one
of
these
talks-
so
you
know
building
your
envoy
control
plane-
that's
rather
interesting,
starting
from
some
examples
showing
more
or
less
what
are
the
goods
of
where
yeah
like
a
go
control
plane
like
kumar
what
it
could
be
like
in
its
core,
but
of
course
it's
much
more
complex
and
then
binding.
I
mean
talking
about
complex.
A
Of
course,
there
are
other
talks
that
might
be
interesting.
The
videos
are
out
there
but
yeah
and
also
while
looking
for
for
the
videos
at
this
cncf
youtube
channel.
I
also
found
out
that
the
envoy
com
videos
are
available.
So
here's
a
link
to
this,
so
there
were
some
some
interesting
talks.
There
definitely.
A
A
Okay,
talking
about
cons,
the
next
virtual
cubecon
eu,
which
will
be
held
online
in
the
beginning
of
may,
is
go
for
paper
and
this
sunday
so
so
eat
your
talks.
Okay,
I
mean
again.
If
someone
is
interested
to
reach
out
to
us,
we
can
figure
out
what
we
want
to
do
so
for
to
today's
agenda.
I
have
added
two
two
topics.
The
first
one
is
a
quick,
quick,
quick,
quick
chat
about
the
the
latest
reviews,
the
latest
releases.
A
So
last
time
we
talked
that
it
was
100
and
102
was
101
and
102
was
on
its
way.
So
we
did
that,
and
we
also
did
a
quick
patch
release
for
103
to
fix
some
some
small
timeout
issues
there
and
then
a
rather
bigger.
I
want
to
share
some
some
some
thoughts
around
upgrades
and
would
also
like
to
call
for
people
on
the
here
in
the
in
the
community
go
to
hear
their
their
thoughts
around
this.
A
This
is
something
that
is
kind
of
still
pending
for
us
to
to
fully
resolve
okay,
if,
if
anyone's
interested
in
adding
something
else,
please
just
add
it
to
the
you
know
to
the
agenda
here
or
we'll
have
time
to
actually
talk
about
it
in
the
end
so
releases.
As
I
said,
we,
our
current
list
is
103..
Just
where
are
the
release
notes,
changelog.
A
Yeah
so
again,
103
is
just
disabling,
a
timeout
that
was
apparently
crucial
for
some
some
of
our
users
we
default
to
disable
it.
Our
plan
is
to
actually
be
able
to
configure,
have
a
fully
dedicated
comma
resource
that
you
will
be
able
to
to
configure
all
the
timeouts
on
your
data
paths
so
that
all
the
the
different
aspects
of
a
particular
call
depending
on
is,
is
it
just
on
the
plain
tcp
level?
A
Is
it
a
http
request,
so
we
have
request
timeouts
response,
timeouts,
the
stream
timeout,
all
the
time
of
that
invoice,
exposing
we
will
be
will
be
able
to
configure
at
least
the
one
that
we
think
are
most
significant.
I
know
if
we
do
each
and
every
one
of
them,
but
yeah,
so
just
because,
of
course,
the
logical
question
is:
why
can't
we
do
this
with
the
provided
proxy
templates?
We
can.
A
Some
of
these
are
configurable,
but
there's
some
some
some,
I
don't
know
small
detail
and
some
corner
case
where,
actually
you
cannot
do
this,
so
we
wanted
to
have
a
more
holistic
approach
and
just
say:
okay,
timeouts.
This
is
the
resource.
This
is
what
you
can
configure.
You
go
and
configure
them
for
various
case
also,
you
can
apply
for
different
scenarios
and
etc,
etc,
of
course
much
easier
than
figuring
out
the
complex
proxy
template.
A
If
you
want
to
go
on
that
level
of
details,
102
is
here
I
I
don't
know
what
what
we
can
highlight
here.
These
are
mostly
at
least
according
to
to
what
I
remember
it
was
like
a
week
ago,
service
inside
10
points
we
added
some
some
some
more
endpoints.
I
know
that
there
was
something
about
fixing
things
with
helm
secrets
and,
of
course,
we
bumped
the
invoice
version,
which
was
current
at
that
time.
A
B
A
I
would
definitely
would
like
to
more,
like
kind
of
sparkle,
a
discussion
on
what
you
want
to
do,
where
we
are,
what
what
could
be
done
better,
I
mean
I'm
sure
that
a
lot
of
things
can
be
done
better,
so
I'll
just
share
whatever
I
have
in
my
mind
and
then
and
then
we'll
talk
about
it.
So
one
of
the
one
of
the
biggest
problems
is
helm
and
how
it
help.
How
I
mean
on
kubernetes
is
the
helm
deployment
parts
where
actually,
sir,
this
appears
to
be
a
problem
there.
A
So
what
we
noted
is
that
in
one
of
the
latest
version
I
think
102
we
have
added
a
new,
a
new
custom
resource
and
apparently
during
the
upgrade
from
1-0
to
102
or
whatever
upgrades
help
doesn't
want
to
to
do
any
any
upgrades
on
crds.
Although
it's
just
adding
a
new
one,
the
moment
that
you
do
helm
upgrade
it
just
it
just
just-
doesn't
touch
anything
anything
related
to
surges.
A
So
so
so
this
is
a
huge
problem
for
us,
because
this
means
that
now,
for
example,
that
when
we
want
to
add
this
new
resource
about
the
timeouts
which
we
spoke
like
five
minutes
ago
and
then
again,
we
have
to
provide
some
custom
upgrade
parts
where
you
download
the
crg
on
your
own
and
you
deploy
it,
and
then
you
upgrade
the
chart
and
blah
blah
things
like
that,
which
is
which
is
which
is
not
really
the
best.
We
were
thinking
about,
adding
some
internal
hooks,
pre-installation
hooks,
and
things
like
that.
A
B
A
A
I
don't
think
that
there's
one
one
size
like
one
answer,
fits
or
answer
so
here,
because
deploying
standalone
kuma
on
kubernetes
is
one
thing
and
then
having
a
multi-zone
hybrid
deployment
with
multiple
gates
and
universal
clusters
is
probably
going
to
be
another
another
challenge
so
for
force,
but
yeah.
If,
if,
if
we
take
standalone
as
an
example,
what
we
try
to
do
so
far,
at
least,
and
it's
kind
of
our
guiding
guiding
principle
when
we
do.
A
You
know
minor
releases
with
this,
this
shouldn't
be
broken
right,
I
mean
you
don't
need
to
you:
shouldn't
need
to
restart
ddps,
for
example,
so
you
just
upgrade
the
the
control
plane
and
then
the
data
plane
should
be
should
continue
working
because
we
try
to
keep
all
the
the
contracts
still
in
place.
We
don't
we
don't.
We
change
these
just
cross
minor
releases
on
the
on
the
major
releases.
We
are,
you
know,
gathering
the
breaking
changes,
and
these
usually
are
harder
to
do
so
effectively.
A
We
have
to
plan
some
downtime
or
restart
of
services,
or
things
like
that,
just
because
you
know
the
changes
on
the
on
the
conversation
that
the
contract
that
is
being
put
in
place
between
the
cp
and
dp
or
sometimes
you
know.
Lately
we
had
a
lot
of
these.
I
don't
think
if
we
have
something
something
big
in
mind
yet
to
to
be
done.
Like
the
cpdb
communication,
I
think
that
we
have
most
of
the
things
that
we
wanted
to
do
there
jacob.
A
Okay,
so
more
or
less
the
same
would
be
would
be
with
multi-zone
deployment.
So,
apart
from
adding
new
new
resources,
which
eventually
need
some,
some
new
way
to
synchronize,
which
could
be,
could
be
problematic.
If
you,
if
you,
for
example,
if
you
upgrade
your
remote
control
plane
and
then
it
tries
to
to
fetch
something
from
the
global
control
plane
that
the
global
control
plane
is
still
not
aware
of,
and
that
probably
could
could
be,
could
be
the
problem.
A
But
for
the
time
being,
I
think,
due
to
the
distributed
nature
of
the
of
of
this
solution,
the
most
problems,
if
there
are,
they
would
be
essentially
on
the
global
control
plane
side.
So
if
there
are
new
resources,
if
there
are
new
yeah,
essentially
resources
policies-
that's
that's!
That's
probably
going
to
be
the
the
major
change
that
can
eventually
raise
some
problems
on
the
dp
side,
as
we
said,
we
mostly
have
finished
whatever.
Is
there?
I
don't
at
least
on
in
this
minds
this
time.
A
This
point
in
time
we
don't
have
anything
into
visibility
that
we
would
require.
You
know
restarting
dps.
Of
course,
we
we
we
will,
we
we
do
upgrade
and
we
will
continue
upgrading
and
voice
version
to
the
to
the
latest
important
one.
So
there's
some
some
new
things
coming
like
I
don't
you
know,
I
don't
know
windows
support
or
you
know,
wasm
support
and
probably
no
no.
A
These
things
are
going
to
be
interesting
for
us
at
some
point,
so
envoy
upgrades
will
still
be
relevant
but
probably
not
necessary
from
what
I
can
tell
at
this
point
in
time
we
have
started
to
push
towards
moving
to
the
next
xds
api,
which
is
v3,
so
v2
is
becoming
obsolete
by
the
end
of
this
year.
So
in
a
couple
of
weeks
and
we
started
to
move
to
the
v3
api
and
as
far
as
the
internal
discussions
are
going
at
least
what
I
just
chatted
with
jacob.
A
A
A
Restarting
them,
then,
if
you
systemd,
probably
you
know,
you
will
start
you
can
restart
the
service
or,
if
you
use
some
other
deployment
methods
for
universe,
so
you
can
use
various
other
toolings
to
do
this,
but
I
think
that
at
least
this
point
of
the
lifetime
of
the
project,
I
said
that
kuma
dp
is
pretty
much
save
their
crosstop
grades
and
then
all
these
thoughts
have
some
value,
but
we
we
tend
to
do
this
very
occasionally.
Let's
say
that
the
upgrades
are
not
really
something
that
we
have
any
automation
or
ci.
A
Definitely
we
we
would
benefit
any
any
ideas
or
in
discussions
within
the
community,
of
how
how
we
can
make
this
better.
So
I
think
that
automation,
ci
kind
of
you
know
it
can
be
tight
in
a
way.
So
if
you
have
an
automated
weight
of
you
know,
I
deployed
the
odd
version
they
deployed
on
the
new
version.
I
go
through
the
crds
or
you
know,
new
resources
that
needs
to
be
instantiated
on
global
or
whatever.
Then
you
can
put
this
in
a
ci
and,
of
course,
the
other
way
around.
A
If
you
want
to
do
a
ci,
you
have
to
somehow
to
make
this,
but
for
the
time
being,
at
least
I
don't
have
any
any
any
specific,
specific
ideas,
except
maybe
just
starting
to
draft
some
some
some
some
testing
plans
that
can
be
become
some.
You
know
upgrade
scenarios
and
see
how
these
were
close
versions
and
things
like
that.
Okay,
so
with
this,
we
have
10
minutes
so
any
thoughts
around
upgrades.
D
I
have
a
question
to
ask
you
around
the
conflict.
Thinking
part,
as
you
talked
about
the
global
control
plane
and
the
remote
control
panes,
I
want
to
understand,
as
in
how
is
the
configuration
passed
on
from
the
global
control
plane
to
the
remote
control
plane
and-
and
previously
I
had
heard
about
this
data
plane
syncher
that
we
do
so.
I
want
to
also
understand
the
correlation
between
these
two
as
in,
why
are
both
of
both
of
them
important?
I.
D
I
know
that
there's
a
grpc
connection
from
the
global
control
plane
to
the
remote
ones.
If
that
is
then,
then
why
is
the?
Why
is
the
polling
done
on
the
gate
on
the
data
on
the
database?
There's
a
database
polling
that
you
do
right
for
each
each
data
plane.
So
just
want
to
understand
this.
This
entire
scenario.
A
B
D
B
D
D
A
A
A
I'm
not
I'm,
I'm
not
really
familiar.
Why
why
we
have
to
pause
for
each
and
every
data
plane?
Is
we
take
a
pillow?
Are
you
aware
of
about
this?
I
mean
how
would
we
implement,
because
I
was
assuming
that
the
pole
means
like
you,
go
and
dump
all
the
whole
data
plane
table
and
just
just
try
to
figure
out
if
there
are
any
any
new
new
ones
that
that
would
be
appalling?
For
me,.
B
C
C
If
you
have
two
data
planes,
you
do
a
poll
for
all
the
policies
once
because
this
is
shared
across
all
the
data
planes.
But
you
need
to
do
the
poll
for
one
data
plane
and
the
second
data
plane.
So
technically
it
is
growing
with
the
number
of
data
plans,
but
it's
not
that
much
of
a
growth
that
every
data
plane
cause
polling
of
all
the
resources.
D
D
We
saw
a
few
database
errors
as
in
it
did
not
sync
up
some
because
of
some
reason
from
the
carpool
db,
and
then
we
added
we
added
new
configuration.
We
added
new
data
plane
configuration
from
a
vm,
but
that
did
not
take
effect.
There
were
like
new
listeners
and
all
they
were
not
taking
effect
and,
and
it
never
worked,
I
mean
we
did
a
lot
of
testing.
We
tried
to
understand.
D
D
C
It's
kind
of
hard
to
discuss
about
this
issue
without
the
config
dump.
I
mean
the
next
time.
If
you
end
up
in
this
situation,
it
would
be
really
helpful
to
have
a
config
dump
to
see
what
is
going
on.
Just
one
idea
on
top
of
my
head
is
that
maybe
the
listener
was
there,
but
it
was
in
a
warming
state
for
some
reason.
D
That
also
we
saw
so
not
sure
I
mean
I
can.
I
can
see
if
I
have
the
config
them
and
I
can
post
it,
but
100
we
went
to
the
control
plane.
We
saw
that
the
control
pin
is
updated
with
the
latest
and
greatest
of
of
the
configuration
what
we
wanted
to
put,
but
that
configuration
was
not
pushed
to
on
y
and
and
there
were
kind
of
few
errors
with
respect
to
databases
that
the
connection
was
not
going
through.