►
From YouTube: CAPZ Office Hours - 6/12/2020
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
And
for
anyone
that's
new
today,
if
you
want
to
introduce
yourself
the
first,
it's
June
12
2020.
This
is
the
cap,
C
office,
hours,
yep,
and
so,
if
we
have
any
new
members,
if
you
want
to
go
ahead
and
introduce
yourself
feel
free
to
do
that.
C
A
A
Well,
thank
you
all
right.
So
our
normal
structure,
we're
gonna,
go
through
our
open
discussion
items
and
we
don't
have
that
many
so
feel
free
to
jump
in
maybe
after
these
top
three.
If
you
have
any
things
that
you
remember
and
then
we'll
go
through
our
project
board
and
see,
what's
going
on
there
and
try
and
trigger
some
new
stuff.
So
we'll
do
that
Ernest!
Do
you
want
to
take
it
away
with
the
testing
provider,
a
jar
with
a
cubsy
demo
or.
A
D
Item
I,
usually
just
call
you
out
yeah
sorry
I,
put
that
one,
oh,
that
was
my
bed
I
put
that
a
few
days
ago.
Thinking
because
I
wanted
to
show
him
Wesley
Steven
and
later,
but
we
can
even
still
show
it
if
you
want
and
maybe
asked
shirt
again
next
time
where
we
can
just
do
it
next
time
way.
Up
to
you
he's
your
oh.
B
C
C
All
right,
can
you
guys
see
my
screen:
yeah
cool,
so
yeah
I've
been
working
with
Cecile
and
other
folks
on
improving
the
testing
strategy
for
cap
Z.
So
for
the
past
month,
I've
been
working
on
enhancing
the
enhancing
a
couple
scripts
in
a
capsule
repo
to
allow
flexibility
when
testing.
C
So
if
you
go
to
our
Docs,
there's
a
some
instruction
on
how
you
can
execute
some
of
that
test,
yeah
so
more
noticeably
on
this
scripts.
Ci
entry
point
is
where
you
can
call
the
script,
and
you
can
run
upstream
kubernetes
end-to-end
tests,
for
example,
conformance
tests
and
any
upstream
tests
that
lives
in
the
creepy
Nettie's
repo.
C
This
group
also
help
you
set
the
Kip
config
to
the
correct
cluster
to
your
workload
cluster.
So
all
you
have
to
do
is
just
go
to
your
project
directory
and
call
whatever
the
script
you
want
to
call.
So
the
intent
test
will
run
on
top
of
the
workload,
not
casi
clusters,
and
currently
we
are
utilizing
this
script
to
run
a
bunch
of
tests
to
improve
our
test
coverage.
So
currently
we
we're
currently
running
conformance
tests
with
that
script
on
both
machine
deployment,
cluster
and
also
machine
pool
cluster.
C
D
So
it's
actually
very
related
to
what
Ernest
was
just
talking
about,
so
what
Ernest
was
showing
was
mostly
testing
upstream
with
cabs
nice,
so
using
cabs
me
as
the
provisioning
tool
to
test
cloud
provider
and
Azure
disk
as
your
file
etc.
In
parallel
we're
also
working
to
improve
how
we
test
cabs
in
itself,
because
right
now
it's
very
bare
minimum.
The
end-to-end
test
suite
doesn't
do
much.
D
It
just
creates
a
cluster
and
deletes
it
and
doesn't
test
anything
in
between
so
Ernest
Nader
and
I
got
together
and
drafted
a
little
testing
plan
document,
which
is
supposed
to
be
what
we
hoped
testing
should
look
like
and
where
we
want
to
be
so
feel
free
to
take
a
look.
It's
shared
in
the
meeting
notes.
It's
that
link
right
there
and
let
us
know
if
you
have
any
comments
or
feedback
right
now.
It's
as
I
said
it's
mostly
your
drafts
and
we
got
together
some.
D
D
Yeah,
so
we
so
there's
a
few
categories
of
testing
that
we
want
to
focus
on
so
unit
test
and
two-man
tests,
conformance
tests
and
periodic
jobs.
So
right
now
we're
focusing
most
of
the
work
on
end-to-end
tests,
because
that's
where
we
think
we're
most
behind
for
unit
tests,
we
want
to
have
test
coverage
in
PR
gates
so
that
that
enforces,
you
know
new
code
to
be
covered
within
the
tests.
That's
an
easier
one
to
get
done.
D
The
end-to-end
is
where
there's
the
most
work
needed
and
most
of
the
action
items
you'll
see
at
the
bottom
are
actually
related
to
n
to
n
tests.
So
if
anyone
wants
to
help
out
with
that
efforts
feel
free
to
reach
out
to
me
or
earnest
or
nadir
or
post
in
the
sectional,
we
have
action
items
at
the
bottom
that
don't
have
any
assignees
on
them.
D
We've
also
started
opening
issues
on
the
repo
based
on
those
action
items,
because
the
action
items
are
mostly
define
what
this
work
should
be,
and
so
once
they're
into
github
issues,
it's
easier
for
someone
to
pick
up,
because
there's
better
definition
of
what
needs
to
be
done
so
yeah.
Any
questions
on
that.
E
D
Sounds
great,
you
can
just
comment,
slash
a
sign
and
I
will
assign
you
if
you
want.
Oh
yeah
just.
D
So
in
this
context,
when
I
say
conformance
I
mean
the
test,
suites
are
defined
by
other
projects
such
as
kubernetes
like
upstream
projects,
so
kubernetes
has
its
own
conformance
test
suite
it's
when
you
run
the
queries
and
to
entice
with
focus
conformance
and
that
will
check
that
any
kubernetes
cluster
it
has.
You
know
it's
capable
of
doing
certain
things
and
it
will
run
functional
tests
on
the
kubernetes
cluster.
So
we
build
a
cavity
cluster
and
then
we
run
Korea's
conformance
on
it
to
make
sure
that
caps
and
clusters
are
conformance.
D
Kubernetes,
wise
I
think
it's
different
from
compliance,
because
when
you
talk
about
compliance
you're,
often
talking
about
laws
that
are,
you
know,
put
in
place
defined
like,
for
example,
the
center
of
Internal
Security
like
the
CIS
rules.
Those
are
compliance,
so
I
think
it's
slightly
different.
But
right
now
we're
mostly
focusing
on
conformance
tech,
startups,
yep.
A
Great,
let's
move
on
I
think
I
have
the
next
thing
so
I'm
curious,
so
I
had
someone
ask
and
I'm
also
curious
about
this.
So
how
do
we
enable
customers
and
users
to
basically
provision
worker
nodes
faster
and
so
today,
I'm,
not
sure,
is
it
just
a
singly
it
does
it
do
it
one
after
the
other,
but
I
think
there
was
some
no
I'm
singing
happy
that
we
were
able
to
turn
but
I'm
just
wondering.
Do
we
have
a
full
picture?
D
You
know
how
many
missions,
machines,
work
or
missions
can
deploy
at
the
same
time
and
those
are
current
in
that
by
default,
they're
being
deployed
concurrently
and
then
there's
the
whole
like
cluster,
like
how
do
we
paralyze
and
optimize
cluster
provisioning?
Because
right
now
you
have
to
provision,
you
know
one
control
plane
at
a
time
and
then
the
worker
machines
and
before
you
even
start
provisioning
control
planes.
D
We
also
creates
the
azure
cluster
infra,
which
includes
the
V
net,
the
subnets,
the
network,
interface
or
sorry,
not
network
interface,
the
network
security
groups
throughout
tables,
and
so
for
those
they're
being
done
serially,
and
so
that's,
where
I'm
hoping
we
can
optimize
and
use
them.
That's
not
all
of
them
are
dependent
on
each
other.
Of
course,
some
infrastructure
is
going
to
have
dependencies
where
you
can't
provision
this.
Some
that's
before
you
provision
the
be
net,
but
there
are
others
where
you
can.
A
H
H
So
the
time
like,
let's
imagine
you
run
through
with
the
maximum
number
of
parallel
deployments
at
the
same
time,
your
your
time
span
is
the
longest
is
how
long
your
deployment
will
last
so
your
longest
deployment
there
if
they
all
start
at
the
same
time,
the
time
to
completion
is
just
your
one
longest
deployment.
So
each
wheeling
number
two
number
two
is
how
long
that
single
deployment
takes.
I
see
right.
A
Cuz,
that's
what
I'm.
Basically,
the
user
is
saying
that
they
tried
to
dial
those
knobs
for
concurrency,
but
they
weren't
seeing
a
difference,
so
that
leads
yeah.
That
leads
me
to
believe
that
what
you're
saying
about
the
longest
appointed
time
might
be
the
case
that
might
be
okay,
although
that
more
details.
D
D
A
A
A
Okay,
I'm
sure
will
it
be
back
to
that
kind
of
talk
anyway,
in
the
future,
we're
working
on
our
machine
pool
deployed
the
control
points
of
that.
A
I
A
I
A
I
D
So
you
could
always
change
the
config
before
you
do:
cluster
cattle
like
config
or
like
after
you
before
you
apply
your
llamo
or
even
a
sir
that's
what's
caught
by
a
cluster
API,
you
don't
need
to
know
what
your
cluster
will
look
like
at
creation
time.
You
can
always
create
it
without
like
with
just
one
mission
bowl
and
then
add
one
later
so
just
you
just
need
a
template
that
defines
what
your
mission
pool
will
look
like
so
you'll
need.
Add
your
mission
below
mission
pool
and
cube
ATM
config
and
then
there's
actually
an
example.
D
I
D
So
right
now,
they'll
end
up
in
the
notes
on
it.
At
the
moment,
there's
only
two
subnets
being
created,
one
for
control,
plane,
node,
someone
for
worker
nodes.
There
is
an
issue
open
in
the
repo
to
support
multiple
subnets.
There
are
use
cases
where
you
want
each
worker
each
pulling
a
different
subjects.
That's
not
supported!
Yes,.
A
A
A
D
F
A
Talks
for
a
manager,
cluster
woo
I,
get
to
review
in
progress
then
put
that
here
what
an
alien
root,
creation
and
cloud
provider.
It's
cannot
be
created,
kind
bug.
Someone
help
me
understand
that
this
is
something
that
we
should
be
doing
sooner
really
yeah.
J
A
But
the
mouth
sodas
be
next
I'll
put
it
in
the
backlog.
For
now,
as
your
resource
names
do
not
follow
conventions
priority
soon.
A
F
A
If
any
of
those,
however
V
as
far
as
some
that
I
didn't
seem
like
this
guy
is,
will
do
that
in
a
second
multiple
subnets
of
role
mode,
so
you're
doings
for
a
drum
machine
pools.
D
A
D
A
Don't
thing
you
didn't
do
I
guess
they
automatically
goes
over
okay.
Is
there
anything
that's
in
or
to
do
that
should
either
not
be
these
guys
should
be
in
backlog
or
something
that
you
want
to
know.
This
is
the
one
I
get
tricked
by
every
time.
Okay,
right.
F
A
A
Multi-Tenancy
with
one
manager
and
since
has
priority
important,
says:
he'll.
Never
ever
just
do
sorry,
which
one
the
multi-tenancy
with.
I
I
D
I
think
those
are
two
different
things.
The
multi-tenancy
one
is
really
being
able
to
have
separate
credentials
for
multiple
clusters.
If
you
want
to
use
the
same
management
cluster,
to
create
cluster
work,
Lloyd
clusters
with
different
credentials
and
the
other
one
was
I
think
being
able
to
pass
the
credentials
as
a
secret,
so
that
they're
not
plaintext.
I
D
So
there's
a
design
doc
link
to
the
issue
if
you're
interested
I,
think
it
explains
I'm
not
saying
I
wouldn't
be
using
a
secret,
is
just
that
the
one
the
other
one
I'm
talking
about
was
specifically
further
cube,
ATM
config
spec,
which
had
like
an
adjacent
file
in
it,
which
contains
secrets
and
clean
text.
So
we
wanted
to
get
rid
of
that
face
up.
G
A
Yeah,
alright,
we
can
fix
that
after
maybe
take
off
that
label
anything
close
it.
So
then
it'll
get
off
the
board.
Okay,
everything
else
looks
okay
to
me
there
anything
that
anyone
wants
a
call-out
from
the
backlog
that
they
want
to
move
over.
So.
D
Some
of
them
might
not
be
labeled
as
good
first
issue,
just
because
there's
like
a
little
bit
more
complexity,
and
it
might
not
be
something
that's
an
easy
one
line
change
but
they're
still
like
good.
You
know
good
places
to
get
started
and
every
one
of
us
here
can
always
help.
If
you
get
stuck
so
there's
a
lot
of
comment
on
an
issue.