►
From YouTube: KubeEdge Community Meeting 20200624 (Europe Time)
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
A
C
A
A
A
Okay,
next,
the
metric
server
in
the
cloud,
any
updates.
A
I
think
we
have
the
red.
The
pull
request
is
ready
to
merge,
just
not
sure
if
it's
already
in
or.
A
A
Okay,
it's
still
it's
still
on
the
review.
Okay,
so
we
will
need
to.
A
Okay,
so
we
just
need
another
approval
about
the
pull
request
and
the
pure
request
is
still
open.
So
if
anyone
have
any
comments,
please
reply
on
the
pro
request
and
the
upgrade
upgrading
golan
to
114.
So
so,
actually,
the
one
the
pull
request
upstream
is
merged.
I
think
we
can
start
the
or
we
can
start
the.
We
can
start
this
one
in
cubic.
A
Okay:
okay,
the
another
thing
is
that
okay,
let's
bring
it
back,
okay,
maybe
we
also
need
to
take
a
look
about
the
kinetics
dependency.
Shall
we
wait
and
upgrade
to
119
directly
this
time
or
not.
D
A
Okay,
so
the
gateway
gateway
support.
It's
still
ending
and
I
think
we
can
add
another
item
that
we
may
verify
and
support
ingress
integration
on
the
edge
nodes.
A
Okay,
the
upgrade
180,
I
think
we
discussed
maybe
so.
The
only
thing
is
that
we
need
to
think
about.
Should
we
wait
and
upgrade
to
119
directly,
but
it's
okay.
If
either
either
is
okay,
just
we
need
to
with
some
feedback
from
the
community
and
the
agenda
certificate
rotation.
Any
updates.
E
Yeah
we
have
updates.
We
have
submitted
the
pull
request
for
review.
A
A
A
One
incubate
cubase
and
me
and
installer
get
data
in
hd
core
db
for
troubleshooting.
Any
updates.
A
All
right
crd
improvement,
so
so
this
one
we
had
some
discussion
and
feedback
in
the
device
iotc
meeting.
I
think.
A
Already
refreshed
the
pull
request
so
just
need
another
run
and
review.
A
Okay,
so
so
the
edge
core
as
the
system
process.
I
think
this
one
pr
has
some
conflicts.
A
A
A
F
A
A
Okay
least
watch
from
major
notes.
A
I
think
this
one
is
still
on
the
discussion,
so
maybe
we
we
need
to
need
another
way
to
think
about
the
the
design.
A
Okay,
the
connected
connection
manager
inside
cloudhub;
okay,
this
one
is
already
in
and
we
are
turning
cherry
pick
to
back
to
release
one
two:
three:
okay,
the
pull
request
is
under
reveal
for
the
gce
yeah,
okay,.
E
A
A
E
B
B
A
Let
me
change
there
for
men,
so
we
can
easy
to
check
the
status.
A
A
A
Oh
don,
I
mean
the
pull
request
accept.
The
edition
should
be
considered
done
if
you
just
submit.
B
A
So
this
or
under
review,
so
this
all
on
the
review.
Okay,.
A
So
what
about
the
no
issues.
A
G
B
G
B
Level,
the
first
it's
depends
on
token
network
interface,
because
this
needed
to
need
a
network
interface
to
bend
to
service
and
dns
server.
We
can
create
a
fake
virtual
network
interface
to
face
to
fix
this.
The
second
is
now
hms
only
supports
ip
ipv4,
dns
resolve
and
and
the
data.
This
is
the
right.
B
Let's
use
the
open
source
library,
we
can
switch
to
open
open
source
library
to
support
all
dns
reviews
resolver,
like
coding,
as
did.
B
B
Yeah,
okay,
also,
there
are
there's
two
two
two
aims
for
each
mesh.
One
is
dns,
one
is
approx
for
the
part
part.
Now
we
only
support.
A
Okay,
so
for
the
proposal,
it's
very
helpful
to
highlight
the
I
mean,
clarify
the
background
and
the
goals
yes,
and
also
for
the
content.
I
think
you
need
to
also
list
the
advantages
and
the
disadvantages.
Why,
for
example,
replacing
the
dns
library
it's?
It
should
be
very
helpful
for
the
others
to
understand
the
the
pros
and
cons
and
also,
I
think,
maybe
from
the
code
level
you
need
to
clarify.
A
If
there's
any
external
behavior
would
be
change.
What
about
the
background
backward
and
forward
compatibility
consideration,
and
if
is,
is
there
any
potential.
G
A
Yeah
and
also
you
may
also
add
the
section
about
the
development
plan
and
the
criteria
of
the
feature
maturity.
I
mean,
for
example,
alpha
the
criteria
to
beta
and
criteria
criteria
to
stable.
A
All
right,
let's
go
back
so
for
the
crossover
net
communication.
I
think
we
still
need
need
more
breakout
discussion
about
this.
A
So
basically
the
idea
for
I
want
to
clarify
the
relationship
with
the
existing
cni
plugin
so
force
for
inside
one
subnet
communication.
A
So
we
can
just
develop
some
plugins
or
add-ons
to
to
make
it
able
to
work
with.
B
The
first
turnout,
so
something
something
I
can
support
the
subnet
called
sub
laser
communication.
B
Well,
it's
as
if
we
use
the
cli
to
deploy
and
as
a
cloud
a
cross
side
and
each
side,
then.
G
A
Turn
to
c9
implementation,
only
only
able
to
support
important
for
the
communication
inside
one
subnet.
C
Here,
the
closer
net
we
need
to
spot
the
across
the
public
network,
the
current
code
don't
start
the
across
the
public
network.
C
B
A
Okay,
so
this
storage
from
the
age.
F
F
F
F
F
F
F
F
F
A
Okay,
so
let
me
just
rephrase
in
english
so
basically
for
supporting
least
watch
from
educore
for
the
applications.
Currently,
there
are
two
two
approach
to
implement.
One
is
that
it's
also
referring
the
the
approach
in
open
yard.
Basically,
we
have
a
proxy
server
to
switch
if
it's
online.
A
If
it's
connected
directly
forward
the
the
request
to
the
to
the
cloud
and
then
we
need
to
duplicate
the
duplicate,
the
reply
of
the
the
received
data
in
hd
core,
as
well
as
sending
it
to
the
client
for
the
watch,
it's
basically
similar,
and
the
second
approach
is
that
we
use
reuse
the
cloud
and
edge
channel
so
for
the
any
request.
First
go
to
meter
manager,
and
we
can
you
reuse
the
query
to
fetch
data
from
cloud
and
in
this
format,
the
watch
my
little
bit
different.
A
We
still
need
to
some
change
in
the
cloud
to
to
keep
forwarding
all
the
watch
event
of
the
watch
updates
to
the
edge
okay.
So
that's
basically
a
very
high
level
idea
and
with,
I
think
both
are
fine.
We
just
need
to
add
more
details
and
at
least
a
list
of
the
advantage
and
the
disadvantage,
and
we
can.
A
Okay,
another
thing
is
that
discussed.
Is
that
whether
crd
is
required?
I
think
it.
I
think
it's
also
a
very
valuable
item
to
be
added
in
the
discussion
if
we
want
to
generally
generally
support
the
as
much
as
as
much
third-party
obligations
as
possible,
we
need
to
support
crd.
A
F
G
We
have
already
implemented
the
logs
and
the
metrics
supporter
for
for
lg
ports,
but
but
there
are
several
several
problems
we
need
to
discuss
and
currently
we
are
using
the
using
the
custom
host
name
mode
for
the
for
the
edge
nodes
and
then
the
coop
api
server
and
the
mesh
server
will
will
try
to
connect
to
the
nodes,
the
the
edge
node
name,
with
the
with
the
the
the
103501
250
port
to
get
bulbs
and
metrics.
G
Then
we
we
found
that
the
kubernetes
server
and
the
max
server
they
are
under
the
cool
system
namespace
they
they
cannot
access
to
the
accordions
tns
supported
resolving.
So
we
are
not
able
to
use
the
software
implemented
service
for
to
forward
for
to
forward
the
request
to
the
cloud
stream.
H
G
F
G
These
steps
are
all
manual
and
in
our
product
we
want
to
make
it
automatic
when
the
when
the
edge
nodes
are
registered
or
I'm
registered,
we
just
least
watch
them
and
simply
sync
the
entries
for
the
hosts
and
ip
tables
automatically
from
the
resource
and
the
nodes
resource.
G
A
G
So
we
just
this
is
our
proposal,
but
we
we
want
to
discuss
with
the
upstream
about
the
how
it
will
be
coexist
with
the
upstream.
A
Yeah
so
automating
the
country
for
configuration
when
installing
the
kubernetes
cluster
is
definitely
a
easy
way
so
yeah
we
we
also
so
actually
when
implementing
the.
When
initially
implementing
this
feature,
we
had
some
discussion
but
not
yet
become
a
very
easy
solution
about
that.
So,
just
like
john
j
said
for
hoster
name.
If
you
use
the
xip.io.
A
G
H
Sure
sure,
so,
with
the
xip.io
method,
we
needn't
to
modify
the
house
name,
the
host
fail
inside
the
kubia
server
or
magic
server,
pod.
A
Yeah,
so
if
you,
if
you
visit
so
xip
dot,
io
is
basically
so
it's
actually
changing
a
ip
address
to
a
domain
to
a
domain
name.
A
So
you
can,
you
can
add,
like
you,
you
can
put
the
the
private
ip
of
a
node
with
the
xib
dot
io
suffix
as
they
host
the
name
for
the
node
and
then
when,
when
api
server
trying
to
access
it
will
get
the
result
ip.
A
So
then,
then,
when
it's
trying
to
send
out
the
request,
the
dip
tables
still
would
get
the
actually.
We
still
need
to
use
ip
tables
to
hook
the
request.
H
Yeah,
so
if
xip
dot
io
is
working
well,
then
the
only
automation,
automation
work
we
we
can
do
is
to
set
up
a
demonstrate
to
modify
the
happy
tables
for
every
node
yeah
yeah
yeah.
That's
great.
A
Any
other
discussion
cube
h.
Many
in
manage
kubernetes
cluster
with
age,
core
or
remote
devices.
D
Yeah
I
placed
it
there
because
I
haven't
found
any
documentation
where
cube
edge
is
deployed
in
any
managed
when
it
discussed
like
eks
like
from
amazon,
azure
or
google,
and
I
just
wanted
to
discuss
if
there
is
any
successful
implementation,
because
what
we
have
seen
in
amazon
manage
kubernetes
cluster.
He
said
it's
not
allowed
to
register
automatically
an
edge
device
as
a
node
in
the
kubernetes
cluster,
as
it
seems
to
be
not
allowed
by
amazon.
A
Have
you
tried
that
or
not.
A
Okay,
maybe
maybe
so,
which
model
you
tried
you
you
tried
it
with
eks
right.
Yes,
okay,.
A
So
automatic
agent
registration
may
be
maybe
there's
some
issue.
We
need
to
double
check.
We,
we
did
only
tested
with
the
kind
of
manually
installed
kubernetes
before,
and
we
have
some
background
investigation
that,
because
the
all
the
managed
kubernetes
service
they
have
their
own
node
status
management.
I
mean
the
underlying
relevant
to
the
infrastructure
layer,
so
they
will,
for
example,
the
cloud
cloud
provider
they
would.
A
They
would
call
the
infrastructure
api
to
check
if
the
virtual
machine
exists
or
not
so
the
current
currently
current
way.
We
hacking.
That
is,
we
add
some
beauty
in
labels,
because.
A
Some
of
the
public
cloud
already
announced
support
with
virtual
collet,
so
virtual
cool
ladder
is
actually
a
a
also.
You
can
think
it's
kind
of
a
remote
node
format
right
the
they
don't
have
a
a
real
world
virtual
machine
in
the
in
the
back
in
the
back
end,
so
we
added
the
same
built-in
labels
with
virtual
equivalent
and
also
we
added
qh
specific
building
built-in
labels.
A
But
probably
we
need
to
if
this
one,
if
the
virtual
corvlet
labels
not
work,
maybe
we
also
need
to
do
the
same
way
so
reach
out
the
reach
out
the
the
manage
the
kubernetes
vendors
ask
if
they
are
okay
with
special
casing
the
remote
nodes,
otherwise,
otherwise
it
would
be
hard
or
we
can
replace
the
node
controller,
but
that's
kind
of
dangerous,
because
you
know
that
node
node
controller
is
managing
the
whole.
A
The
the
the
audi
notes
are
inside
the
cluster
and
it's
very
sensitive,
especially
when
any
disaster
happen
so
yeah.
So
if,
if
the,
if
the
vendors
has
no
customization
of
the
node
controller,
maybe
we
can
just
replace
it
with
the
upstream
node
controller
and
also
special
casing
the
remote
nodes.
D
Okay,
so
custom
label
could
work
and
probably
also
better
to
manually
register
the
edge
node.
A
A
So
I
think
currently
we
we
do
need
some
input
information.
G
A
We
we
probably
need
the
others,
help
to
verify
on
the
other
public
clouds
to
check
whether
currently
implementation,
how
many
managed
kubernetes
can
work
with
cubah
or
not
yeah,.
D
Discuss
do
you
have
a
number
of
which
scale
cube
edge
should
work
with
the
kubernetes
caster
like
having
one
one,
kubernetes
master,
and
what's
the
number
of
edge
devices,
it's
capable
of
managing
it's
not
limited
by
kubernetes
right.
A
I
need
to
double
check
that
I
think
currently
there
are.
There
are
some
discussion
ongoing.
So
so,
basically
we
have
the
model
that
you
know
the.
Not
all
the
nodes
are
directly
connected
to
the
api
server
and
and
our
our
model
actually
did
some
optimization
to
kubernetes
master
analog
communication.
A
Well,
so
one
thing
would
affect
the
scalability
is
actually
how
many
the
model
of
I
mean
the
number
of
nodes,
pods
and
also
devices
so
like
upstream
kubernetes.
They
have
some
so
they
tested
basically
50
parts
per
node
and
it's
able
to
manage
5
000
right.
If
I
remember
correctly
in
the
age
cloud
in
the
age
computing
scenario,
the
node
would
be
definitely
smaller
than
the
y
in
the
cloud.
A
So
in
theory,
actually,
we
might
be
able
to
manage
more
nodes,
but
another
thing
I
mean
another.
Another
thing
is
about
how
many
iot
devices
you
want
to
manage,
so
that
would
affect
the
number
of
scalability.
A
G
A
A
A
A
All
right,
if
not
maybe
we
can
end
the
meeting
early
today,.
A
Okay,
so
thank
you,
everybody
for
joining
the
meeting
next.
One
next
time
will
be
the
pacific
friendly
time
it's
july.
The
first
see
you
all
next
week
and
take
care.