►
From YouTube: Kubernetes UG VMware 20200806
Description
August 6 2020 meeting of the Kubernetes VMware User Group - covered new features + bug fixed in the recent update release to the vSphere Cloud Provider and kube-vip, a relatively new open source load balancer that may be useful for on premises Kubernetes deployments
A
A
You
can
see
the
agenda
there.
I'd
also
encourage
people
to
put
their
name
in
the
attendees
list.
Sometimes
people
talk
and
would
like
to
interact
later
and
sometimes
with
zoom.
It
isn't
clear
who
was
speaking
so
if
you
put
your
name
there,
it
helps
others
connect
with
you
later.
Should
they
choose
to
the
agenda
today
we're
going
to
start.
We
originally
had
planned
with
on
having
a
speaker
who
couldn't
make
it
today.
A
Dan
finnerman
but
miles
did
a
a
last-minute
attempt
to
learn
the
topic
and
he's
going
to
talk
to
us
in
a
way
I'm
looking
for
miles
and
I
were
chatting
over
slack
and
he
seemed
a
little
nervous.
But
I
think
in
a
way
it's
great
to
see
somebody
take
something
on
for
the
first
time
and
learn
it
because
they
can
provide
good
feedback
to
the
rest
of
us
about
what
the
experience
might
be.
Like.
So
he's
going
to
talk
about
cube,
vip,
a
load,
balancer
and
maybe
some
a
few
other
things.
A
Then
david
von
fenin
is
here
to
talk
about
the
new
version
of
the
vsphere
cloud
provider
that
just
came
out
gee,
I
think
about
a
week
or
two
ago,
but
he'll
give
you
the
exact
details,
then.
Finally,
we're
going
to
cover
some
of
the
recent
cncf
conference
activity
just
quickly,
we'll
give
you
links
to
a
few
things,
but
we
we
gave
a
user
group
session
at
the
cloud
native
china
just
a
few
days
ago,
and
then
we
prepped
our
recorded
talk
for
the
upcoming
kubecon
europe
event.
A
B
So
I'm
just
going
to
share
the
cuvette
website,
because
why
not
like
steve,
said
we
originally
had
dan
fernan,
who
actually
wrote
this
schedule
to
come
on,
but
due
to
some
other
circumstances,
he
could
make
it
today.
So
I've
made
an
attempt
over
the
last
five
six
hours
to
learn
this
technology
and
try
and
deploy
it
so
that
I
could
tell
you
how
it
works,
and
I
figured
the
easiest
way
to
do
that
would
be
via
cluster
api
for
vsphere.
B
B
Oh
apparently
they
do
you
get
an
error
too
many
requests
and
it
just
stops
working
okay.
So,
basically,
what
cube
vip
is?
Is
you
can
think
of
it
as
kind
of
like
metal
load
balancer
if
you're
familiar
with
bare
metal
kubernetes,
it's
kind
of
like
metal,
lb
except
metal
lb
is
just
for
services
within
the
kubernetes
cluster.
B
Cubevip
is
designed
to
do
that
and
also
to
provide
a
load
balancer
in
front
of
the
control
plane
node,
so
you
can
think
of
it
as
a
replacement
for
hi
proxy
as
well.
So
currently,
in
cap
v,
we
package
h
a
proxy
if
you
have
multiple
control
plane
nodes
to
make
sure
that
the
api
server
front
end
ip
is
load
balanced
between
those
control,
plane
nodes
and
in
cap
v
0.7.0,
which
is
in
beta
at
the
minute.
It's
not
fully
released.
We've
integrated,
cube,
vip,
so
cube.
The
the
reason
for
that
is.
B
It
removes
the
dependency
on
hj
proxy.
So
it's
another
thing
that
we
don't
need
to
deploy
and
it
can
do
a
dual
purpose.
You
know
act
as
the
front
end
load
balancer
for
your
control,
plane,
node
and
act
as
your
load
balancer
for
your
services
as
well,
so
playing
with
it.
It
is
actually
really
nice.
It's
pretty
slick,
some
of
the
features
in
there
compared
to
metal,
lobe
balancer
are
lagging
a
bit.
For
example,
there's
no
bgp
support.
B
B
It's
a
really
nice
project
to
go
and
have
a
look
at,
particularly
because
you
can
use
it
inside
and
outside
the
cluster
and,
like
I
say,
it's
being
integrated
into
cluster
api
vsphere
and
the
0.7.0
release,
which
is
coming
out.
I
think
next
week
or
the
week
after
sometime
like
that,
it's
currently
in
beta,
so
you
can
try
it
now.
B
If
you
want
to
it
requires
that
you
give
it
a
static
ip
address,
but
other
than
that
really
quite
simple
to
set
up,
but,
like
I
said,
I've
had
some
challenges
because
I
didn't
know
what
I
was
deploying.
I
was
trying
to
deploy
something
with
hh
proxy
and
thought
it
was
cube
vip
and
basically
I
just
didn't
read
the
manual
correctly.
So
that's
about
all.
A
I
had
a
few
questions.
Maybe
you
know
these
or
not,
and
one
comment
I
guess
where
you
talked
about
middle
I'll,
be
not
being
usable
outside
the
kubernetes
cluster.
Well,
obviously,
the
whole
point
of
the
load
balancer
is
to
stand
stuff
up
to
use
from
outside
the
cluster,
but
you're,
meaning
that
things
get
hosted
behind.
A
You
know
the
other
side
of
the
load
balancer
right,
and
I
know
that
because
I've
done
this
myself,
if
you
take
the
load
balancer
and
use
it
in
conjunction
with
a
service
mesh
or
at
least
with
istio
istio,
is
capable
of
providing
mesh
connections
for
it's
called
mesh
expansion
for
legacy
vm
or
they
don't
necessarily
even
have
to
be
legacy
apps,
but
new
vm
services
and
it's
actually
istio
acting
as
an
intermediary.
But
it
is
possible
to
use
pretty
much
any
load
balancer
to
connect
non-kubernetes
cluster
things.
A
B
New,
it's
pretty
it's
pretty
fresh
at
the
minute.
I
think
it's
only
got
100
or
something
stars
on
github.
The
main
contributor
obviously,
is
dan.
It
was
his
project.
There
are
a
couple
of
other
people
that
are
contributing
to
it
now
and
seem
to
be
quite
active.
So
it
is
something
that's
growing,
but
I
think
it
was
only
released
a
couple
of
months
ago,
so
it
is
still
pretty
fresh
as
far
as
projects
go.
A
B
Yeah,
well
and
and
metal
lb
has
essentially
come
to
a
standstill
as
well,
because
the
main
contributors
stopped
contributing
to
it
so
yeah
it's
a
valid
concern
to
have.
But
that
said
you
know,
metal
lb
is
largely
functional
for
most
things
that
you
want
to
use
it
for
and-
and
it's
you
know
quote
unquote
good
enough.
A
A
If
you
are
on
premises,
if
you're
big,
you
might
have
like
an
f5,
but
what
do
you
do
when
you're
moving
out
there
to
edge
or
remote
offices
where
you're,
perhaps
not
willing
to
spend
the
money
to
put
the
big
hardware-based
load
balancer
in
place
when
your
needs
are
light?
And
you
know
you'd
like
a
software-based
solution?
They're,
it's
almost
like
it
strikes
me
that
there
isn't
a
critical
mass
of
everybody's
gotten
behind
one
solution
and
there's
kind
of
a
I
don't
know.
A
Maybe
I've
come
across
up
to
five
of
these
things
that,
like
I
say,
they're
used
here
and
there,
but
don't
have
a
massive
foundation,
support
behind
them
with
a
huge
number
of
users
and
developer
contributors.
But
I'm
going
to
take
a
look
at
this
because
I
find
myself
playing
with
these
different
load
balancers.
Just
as
I
try
to
learn
this
check,
so
maybe
if
it's
of
interest
leave
us
word
or
comment
on
the
slack
channel,
because
this
keeps
coming
up
for
people
using
kubernetes
on
premises,
so
anything
anybody
discovers
out
there
in
load
balancer.
A
A
C
Yeah
sure
so
yeah,
just
as
an
fyi
cloud,
the
cloud
provider
vsphere
was
we
had
a
release.
Was
it
well
technically
on
friday
and
yeah?
So
it's
now
1.2.
I
just
wanted
to
like
cover
like
four
kind
of
like
major
notable
things.
There's
a
bunch
of
other
bug
fixes
documentation
updates,
which
are
all
great,
but
the
four
one
I
wanted
to
really
touch
on.
Probably
one
of
the
more
important
ones
is
node
deletion.
C
So
the
cpi
now
kind
of
follows
convention
with
the
other
cloud
providers
such
that
when,
like
the
vsphere
vm,
is
no
longer
accessible
or
present
according
to
vcenter,
it's
going
to
trigger
a
no
deletion
and
remove
it
from
kubernetes.
So
so
that's
definitely
something
to
be
aware
of.
There
are
flags
if
anyone
needs
the
old
behavior
to
re-enable.
C
Like
the
old,
you
know
the
old
behavior,
which
is
basically
leave,
leave
the
the
kubernetes
node
like
registered
there.
You
can
anyone's
interested.
You
know
there
is
a
flag
for
it,
just
as
an
fyi
that
no
deletion
is
the
default
behavior.
So
just
be
aware
of
it.
There's
experimental
nsxt
support
now
for
load
balancers,
it's
experimental.
C
The
next
thing
is
that
there's
yaml
support
for
the
cloud
config.
So
previously
the
cloud
config
was
using
like
the
old,
any
style
configuration
which
most
other
cloud
providers
use,
and
now
we
have
support
for
yaml
and
just
just
to
let
everyone
know
that
the
the
yaml
or
the
any
functionality
is
still
there
by
default.
It's
going
to
try
to
whatever
cloud
config.
C
C
A
C
We
had
to
do
like
hacks
like
in
like,
if
you
had
the
old,
any
configuration
which
had
like
bracket
vcenters
or
there's
a
vcenter,
and
then
it
had
the
ip
address
in
order
to
support
multi-tenancy,
because
you
could
potentially
have
multiple
accounts
for
a
given
vcenter
server.
So
in
order
to
kind
of
like
and
in
that
configuration
you'd
have
a
vcenter
server,
two
entries
with
the
exact
same
ip
address
and
in
when
you
actually
parse
any
it.
That's
not
allowed,
because
the
key
is
exactly
the
same.
So
we
had
to
do.
C
Is
we
had
to
do
things
like
overwrite
that
property
replace
it
with
like
a
tenant
id
instead
and
then
have
a
field
within
the
vcenter
any
configuration?
That
said,
this
is
the
actual
server
just
in
order
to
get
uniqueness
in
that
that
any
tag,
so
things
like
that
were
very,
very
difficult
to
express
in
using
any,
and
so
that's
why
we
moved
towards
that
yaml
configuration
there
are
the
I
forgot,
the
gentleman's
name
who
the
few
people
who
implemented
the
the
nsxt
load
balances
support.
C
They
actually
had
all
their
configuration
based
in
yaml
first
and
because
we
didn't
have
a
yaml
configuration
that
we
could
like
plop
into
the
cloud
config.
They
made
efforts
to
like
convert
everything
to
any,
and
so
this
basically
it's
like
it's
a
more
expressive
way
of
like
like
writing,
down
configuration,
and
it
also
happens
to
fall
in
line
with
like
the
way
we
deploy,
pods
and
stuff
using
ammo
configurations.
So
that's
kind
of
like
the
big
reason
and
motivation
for
doing
it.
Yeah
sounds
like
it.
A
Also,
maybe
lets
you
avoid.
You
know
mental
mistakes
and
things
because
you're
kind
of
in
the
mindset
of
using
yaml
all
day-
it's
just
like,
if
I
don't
know
about
the
rest
of
you,
but
if
you're
a
developer
and
you
work
a
job
in
java
for
a
year
and
then
go
back
to
c
plus
plus
that
that
is
kind
of
a
context
shift
that
you
end
up
really
screwing
up
and
you.
I
would
never
wanted
a
job
where
you
really
did
that
flipping
back
between
languages
every
hour.
C
C
If
you're
using
the
ammo
that's
available
in
the
repo
the
manifests
folder,
that's
there
it
by
default,
now
deploys
as
non-route.
So
that's
basically
for
security
reasons
to
avoid
like
privilege,
escalation
and
and
that
stuff
for
for
security
and
that's
kind
of
about
those
are
the
kind
of
the
big
highlights
that
I
just
wanted
to
point
out
and
there's
a
bunch
of
other
fixes
and
some
other
features
but
yeah.
Those
are
like
the
main
ones
that
really
stand
out
and
that's
it.
A
Okay,
thanks
david
and
I
just
post,
put
a
link
to
the
actual
release
notes
for
that
release
in
the
chat,
so
that
has
the
full
list.
I
think
it's
a
pretty
big
list
so
go
have
a
look
if
yeah,
if
you're
dependent
on
that.
So
moving
on
the
next
item
on
the
agenda.
Is
these
online
cncf
conferences
so
miles
and
I
just
finished
a
presentation
at
what
they
called:
they
didn't
call
it
kubecon.
A
So
far
I
haven't
seen
a
publicly
available
recording
link
and
I
think
the
cncf
is
perhaps
trying
to
fund
the
foundation
partially
through
charging.
You
know
for
registration
at
these
conferences,
so
even
though
I
think
miles-
and
I
do
have
a
recording
of
that-
I'm
in
respect
to
the
cncf-
I'm
not
going
to
put
it
out
there-
I'd
expect
if
they
behave
like
they
have
in
the
past.
A
Maybe
a
month
from
now
they're
going
to
upload
that
presentation
to
youtube,
but
in
the
meantime
the
deck
is
there
in
the
agenda
notes
and
let
me
just
copy
it
into
the
chat.
If
for
people
who,
perhaps
don't
have
the
notes
open
the
content
that
we
covered
in
that
deck
was
the
migration
from
entry
to
auto
tree
both
for
the
cloud
provider
and
for
storage.
A
A
A
C
Yeah
sure
so,
basically,
the
talk
covers
mainly
how
to
use
the
multi-tenancy
support
that
was
introduced
in.
I
think
it
was
1.1,
but
there
was
a
bunch
of
other
features
that
were
added
to
enable
like
further
enable
the
multi-tenancy
story,
and
so
the
my
talk
basically
covers
like
cpi
and
csi
and
like
how
you
can
potentially
deploy
both
the
cpi
and
csi
driver
in
ways
that
you
know
could
enable
you
know
better
efficient
use
like
sharing
of
resources
in
vsphere
like
and
as
they
pertain
to
kubernetes
clusters.
A
B
A
Okay,
that
sounds
like
great
material
and
then
second,
the
second
half
of
that
talk
is
something
a
little
bit
different
from
what
we've
been
talking
about
so
far
and
that's
running
kubernetes
on
vmware's
desktop
hypervisors.
That
would
be
fusion
for
the
mac
and
workstation,
which
has
versions
for
running
on
linux
and
on
microsoft
windows,
and
I
give
kind
of
an
attempt
at
a
live
demo
of
standing
up
minicube
on
top
of
the
windows
desktop
hypervisor.
A
I
had
potentially
chose
windows
because
it's
the
more
difficult
and
the
one
that
seems
to
be
very
infrequently
covered
in
blog
posts
and
things,
but
I
got
kubernetes
running
on
my
windows
laptop.
I
didn't
do
a
video
recording
of
the
desktop,
because
in
these
new
online
conferences
the
bandwidth
is
challenged.
So
I
did
it
more
screenshot
style,
but
I
cover
a
step
by
step
and
I
think
if
you
watch
that
get
the
recording
and
then
pause
it
you,
it
should
be
reproducible.
A
A
Finally,
a
couple
other
sessions:
we
don't
have
speakers
for
these
in
this
meeting
today,
but
they
sounded
interesting.
So
there's
a
user
panel
session
which
had
about
four
different
kubernetes
end
users.
This
is
not
really
vmware
specific,
but
I
mean
this
is
a
user
group,
so
it's
kubernetes
users
talking
about
their
experiences
and
lessons
learned
and
then
finally,
there's
a
user.
A
A
Also,
there
might
be
of
some
first-timers
here,
so
we
often
would
open
the
floor
to
people
introducing
themselves
and
just
give
us
a
flavor
if
you
choose
to
of
you
know
why
you
came
here
or
what
you'd
like
to
get
out
of
the
group,
and
maybe
we
can
do
something
about
that
or
just
floor
is
open
to
anybody.
So
just
unmute
yourself
and
speak
up.
If
it's
it's
a
small
and
a
girl
off
group,
we'll
try
free
for
all.
Instead
of
making
you
raise
hands
and
see
how
it.
A
D
A
Yeah
sure
the
a
couple
of
ground
rules
that
we've
had
problems
with
these
zoom
meetings
with
trolls
coming
in
disrupting
meetings,
so
the
directives
coming
down
from
the
kubernetes
community
leadership
are
that
we
shouldn't
be
posting
the
zoom
link.
We
make
people
at
least
join
the
the
list
and
then
the
the
the
zoom
link
to
meetings
is
posted
in
there,
so
it's
slight
inconvenience
but
believe
it
or
not.
There
are
people
using
google
searches
just
to
find
zoom
links
and
dropping
in
for
their
own
amusement.
A
I
think
it's
kind
of
the
digital
age
equivalent
to
graffiti
tagging,
a
wall
or
a
railroad
car
or
something
where
I
don't
know
who
has
enough
time.
Apparently
some
people
do
but
other
than
that
yeah
anything
you
want
to
do
to
promote
the
existence
of
the
group
how
to
join
that.
That's
all
open!
You
can
certainly
even
post
meeting
agendas
things
like
that,
but
the
only
limitation
is
on
the
zoom
link
itself
that
I'm
aware
of
the
other
thing.
That
is
a
typical
thing.
A
People
run
into
the
agenda
notes
thing
document
for
this,
and
essentially
all
documents
in
the
kubernetes
project
are
gated
by
group
membership.
So
you
know
in
the
interest
of
participation
we
let
everybody
go
and
have
edit
rights
to
this
document,
but
it
isn't
literally
any
random
anonymous
person.
The
docs
are
kept,
you
know
as
a
google
doc
and
you
have
to
be
logged
in
as
that
account
an
fyi.
A
If
you're
new
to
this
group,
a
frequent
complaint
is
someone
will
register
their
group
membership
in
a
personal
or
a
corporate
email
and
then
flip
over
to
the
other.
One
try
to
open
the
dock
and
report
that
hey,
I
don't
have
access
right,
so
just
join,
join
the
group
and
you
should
have
access
to
it.
Assuming
you
are
logged
into
google
using
that
same
account,
you
used
to
join
the
group.
A
A
It
should
be
just
general
topics,
like
you
know,
serving
all
users
with
regardless
of
whether
they're
on
pure
open
source
or
some
commercial
product
talking
about
just
kind
of
the
general
requirements,
is
fine
or
best
practices
and
you've
crossed
the
line.
If
you're,
quoting
product
numbers
and
prices,
you
know
general
topics,
yeah
we're
we'd,
even
welcome
them
for
future
meetings.
D
Yeah,
so
that
that
is
actually
a
follow-on
question
in
that
we
are
working
with
open
source
software
vendors
to
enable
their
their
offerings
to
work
with
vsphere.
But
we
do
have
some
closed
source
software
vendors
that
we
work
with.
I
think
it's
like
at
least
100
different
people
who
sell
proprietary,
backup
and
recovery
software.
D
A
A
If
you
cross
into
the
area
of
standing
up
stateful
workloads
on
kubernetes,
there's
people,
I
swear,
I
you
know
it's
new
to
some
people,
so
I've
heard
people
speculate
with
crazy
talk
as
crazy
as
gee.
It's
in
my
docker
image
and
I
recorded
something
well.
This
docker
image
was
running,
so
it
must
get
saved
in
the
image
registry
right,
and
I
mean
obviously
anybody
who
who
knows
how
this
works
knows
that
isn't
true,
but
there
are
people
who
just
don't
understand,
fundamentals
like
that
and
have
without
knowing
it
made
wrongful
assumptions.
A
D
You
know,
first
of
all,
like
how
the
vmware
storage
stack
works
with
kubernetes,
so
people
can
start
to
you
know
understand
you
know,
what's
going
on
there
and
then
we
have
some
partners
we're
working
with
who
are
open
source
providers
and
we
might
invite
them
to
start
to
come
into
these
calls
and
talk
about
what
they're
doing
with
vmware
and
how
they're
going
about
providing
their
services.
On
top
of
our
platform.
A
Yeah,
I
think
it's
even
fair
to
invite
you
know
I
I
think
users
is
often
interpreted
as
the
end
user.
You
know
somebody
who
is
standing
this
up,
but
often
there's
an
where
they
have
vendors,
who
do
things
on
their
behalf.
I
mean
when
it
comes
to
backup
in
dr
they'll,
often
engage
organization
wide
with
somebody.
A
I
don't
know
I'll,
throw
a
few
names
out
there
like
tivoli
or
symantec
other
names,
that
kind
of
started
in
the
bare
metal
world,
but
they've
got
an
ongoing
relationship,
and
I
know
having
worked
in
back
up
in
dr
you
don't
change
vendors
there
overnight,
even
in
the
course
of
a
year
or
a
decade,
because
they're
very
sticky
simply
because,
for
even
legal
reasons,
you've
got
to
keep
these
backups
online
and
available.
Sometimes
for
four
years.
You
know,
there's
hipaa
things
sometimes
for
10.
A
Some
financials
want
to
keep
it
essentially
forever
and
flipping
a
vendor
means
that,
since
I
still
have
to
keep
my
legacy
inventory
of
these
backups,
what
you're
really
asking
me
to
do
to
flip
vendors
is
to
take
on
two
vendors
simultaneously,
so
they
they
tend
to
not
go
away
and,
as
an
organization
moves
out
of
bare
metal
and
virtualization
into
containerized,
they
they
aren't
going
to
be
able
to
get
a
divorce
from
that
vendor.
Who
is
keeping
the
massive
inventory
of
their
legacy?
A
Stateful
apps,
so
they're
tending
to
want
to
stick
the
backhand
stick
with
whatever
back-end
storage
they
still
had.
So
this
becomes
a
deal
of
transitioning
rob.
If
you
can
bottom
line
is
if
you
can
get
those
kinds
of
people
invited
to
hear
the
presentation,
I
think
it's
an
interaction
between
the
users
and
that
class
of
service
providers
and
I'd
welcome
that
at
meetings
of
this
user
group,
as
long
as
it
doesn't
turn
into
product
pages,.
B
B
A
A
A
Please
last
call
otherwise
I'm
gonna
end
this
early
and
let
you
do
whatever
okay.
Well,
thanks
for
attending
the
meeting
and
we'll
see
you
first
thursday
of
next
month
and
final
reminder
there.
There
are
some
good
sessions
at
this
upcoming
kubecon
europe
online
conference,
but
one
other
thing
that
the
agenda
notes
has
a
link
to
register
for
that
conference.
A
You
have
about
24
hours
to
get
a
free
registration
for
at
least
the
keynote,
and
I
think
what
they
call
the
equivalent
of
the
vendor
hall
or
something
you
can
get
that
for
free
in
the
next
24
hours.
After
that
you
have
to
pay
75
bucks,
and
if
you
want
to
go
to
the
sessions,
I
think
you
pay
75
bucks
anyway.