►
From YouTube: TGI Kubernetes 183: Sonobuoy
Description
Join @johnschnake and @eleanor-millman as they walk through how to run Sonobuoy to run the upstream, Kubernetes end-to-end tests and show how easy it is to write/edit our own plugins with the new plugin-skeleton and plugin-helpers.
A
So
we're
here
representing
sanibui-
I
am
the
kind
of
main
engineer
working
on
sanibui
right
now
and
so
yeah.
So
we
want
to
go
through
today
and
talk
about
the
project.
What
we
do,
how
you
can
use
it
and
and
kind
of
where
we
might
go
from
there.
One
of
the
big
things
we're
going
to
do
during
today
is
we're
going
to
make
our
own
little
testing
plug-in.
Hopefully
that
will
inspire
people
to
to
use
sonaboo
to
integrate
their
own
tests,
but
we'll
we'll
get
more
into
that
eleanor.
B
A
B
A
B
A
From
the
eastern
province
of
saudi,
this
is
one
of
the
cool
things
whenever
being
on
tgik,
seeing
people
from
all
over
the
world
legitimately
right.
We
sometimes
are
in
this
little
bubble
or
kind
of
imagine
people
being
in
a
small
bubble,
but
it
really
is:
is
global.
Finland,
amsterdam,
hi
everybody
thanks
for
joining
us
today,
I
am
in
the
southern
united
states,
I'm
in
louisiana.
A
A
B
A
A
A
A
A
A
B
B
B
A
Well,
so
the
kubernetes
api
is
really
big,
but
so
there's
whole
classes
of
endpoints,
but
some
of
those
classes
right.
If
you
look
in
the
path
of
the
url
for
the
api,
endpoint
it'll
say
something
like
v1
beta1
slash.
You
know
some
pod
rotation
object.
So
as
that
feature
whatever
it
is,
goes
on
in
life
and
gets
more
testing
more.
A
That,
like
this,
is
going
into
all
the
clusters
we're
going
to
put
this
as
part
of
conformance
it'll
evolve
through
that
beta
from
alpha,
then
to
beta
and
then
eventually
it'll
change,
that
pathing
to
where
it'll
be
a
generally
available
endpoint
at
some
other
api
group.
Okay,
does
that
answer
the
question.
B
Mostly
I'm
used
to
think
of
kubernetes
apis
as
like
custom
resources
or
like
creating
a
resource
like
that
kind
of
declarative
api.
But
it
sounds
like
when
you're
talking
about
like
with
the
path
it
sounds
more
rest
like
so
I'm
trying
to.
I
I'm
even
asking
a
more
basic
question
than
you
think.
I'm
asking.
A
A
B
A
B
A
I
didn't
realize
that
was
in
github
I
had
scrolled
down
before,
and
I
thought
it
looked
like
a
nice
little
white
page
white
paper
website,
but
it's
all
in
github,
so
yeah,
so
anybody
who's
interested
can
go
through,
but
especially
if
it's
got
the
the
sig
and
the
working
group
both
being
behind
it.
You
know
that
you're
getting
the
the
best
level
of
information.
B
B
A
A
So
if
something
says
like
it's
certified
kubernetes,
you
have
to
submit
the
results
of
these
end-to-end
tests.
That
kubernetes
maintains
sauna.
Buoy
doesn't
maintain
those
tests,
but
you
submit
these
test
results
running
those
tests
can
be
complicated
if
you
just
have
to
start
from
the
source
code
and
the
kubernetes
repo,
so
they
made
you
know
they,
they
containerize
it
all.
There's
a
couple
different
ways
to
run
them.
The
cncf
recommends
sonaboo.
It's
just.
A
Go
over
to
the
show
notes,
we're
going
to
talk
about
like
running
these
end
to
end
tasks,
but
eventually
we're
going
to
talk
about
making
our
own
plugin.
So
how
can
you
still
leverage
a
bunch
of
the
nice
things
about
sony
buoy
but
run
your
own
tests
and
and
we're
going
to
talk
about
some
future
ideas?
And
this
is
really
where
I'll
definitely
hand
over
to
you
eleanor
to
talk
more.
B
B
And
just
to
give
you
a
teaser
now
so
as
you're
kind
of
seeing
the
tech
you
can
be
thinking
about
how
to
extend
it.
As
john
has
said,
exactly
alexanderbu
is
so
solid,
stable
running
the
conformance
test,
but
john
and
vlad
already
did
a
lot
of
work
to
maybe
make
it
easier
for
others
to
make
plug-ins
and
folks
have
started
to
develop
other
uses,
and
we
just
the
more
we
thought
about
it.
B
The
more
we
realized
like
automated
cluster
validation,
there's
so
many
use
cases,
for
example,
you're
a
support
engineer,
and
you
want
to
make
a
really
quick
diagnosis.
Initial
diagnosis
of
what's
wrong
with
the
cluster,
you
can
run
a
suite
of
tests.
Basically
the
performance
test
and
say
networking
is
broken
because
the
networking
tests
are
failing
or
maybe
you've
done,
a
restore
you've
done
a
backup.
Then
you
restore
something
how
the
heck
do
you
know
if
what
you
restored
is
working?
You
could
use
automated
tests
to
see
whether
things
are
functioning
as
it
should.
A
Of
running
the
e
to
e
test,
one
of
the
big
values
of
sonogui
is
just
that
kind
of.
You
can
leverage
all
the
ui
and
results
reporting
structure-
that's
already
there.
So
if
you
already
have
it
in
your
ci
system
to
run
these
certified
kubernetes
tests
to
make
sure
your
cluster
is
operating
the
right
way,
you
can
change
just
one
line
in
your
scripts
and
you
can
potentially
be
testing.
You
know
two
three,
four
other
test
suites.
A
At
the
same
time,
some
of
them
are
maybe
proprietary
closed
source
things
that
you
developed,
maybe
other
ones,
are
open
source
plug-ins,
there's
a
few.
We
can
go
over
those
towards
the
end.
I
mean
we
have
a
sun
buoys,
plug-in
repo.
Some
of
them
are
security-based.
Some
of
them
are
performance-based
and
anything
that
you
can
write.
You
know
if
you
can
containerize
it
sound
booty
can
run
it.
So
that's.
B
A
A
A
A
A
One
of
the
main
commands
you'll
you'll
do
is
sanibuy
run
right,
run
a
plugin,
that's
how
we
we
describe
it
when
you
say
to
run
a
plug-in.
It's
going
to
leverage
your
cube,
config
file
locally,
to
talk
to
your
cluster,
so
you
don't
have
to
give
sonaboui
any
extra
permissions
or
anything
like
that.
You
don't
have
to
tell
it
where
your
cluster
is.
It's
just
going
to
be
looking
at
what
your
current
targeted
cluster
is
through
that
q
config.
A
Then
it
will
launch
an
aggregator
pod.
This
is
when
it's
launched.
The
name
of
it
is
just
sonibuy.
This
is
the
thing.
That's
going
to
gather
up
all
the
results
from
whatever
you
decide
to
test
because,
like
I
said
you
can
test
multiple
test
suites
all
at
the
same
time,
you
can
be
gathered
by
default.
We
gather
logs
from
the
nodes
themselves
and
we
run
the
ede
test
suite.
A
So
you
know
the
aggregator
then
launches
all
the
the
plug-ins,
and
we
have
this
little
worker
here
sitting
next
to
it.
That's
also
sanibui
just
running
in
a
different
mode.
It's
the
same
sonaboo
cli
though,
but
it's
working
with
this
plugin
to
communicate
those
results
back
to
the
aggregator
and
then
from
your
side
all
the
way.
At
the
client
side,
you
can
use
other
commands
and
we'll
see
that
how
to
get
those
results
back
out
and.
A
A
A
Yeah,
that's
the
pattern
that
we
use,
so
it's
considered
a
side
car.
So
your
plug-in
definition
is
really
just
a
yaml
file
that
tells
us
what
image
to
run
as
a
container
how
to
run
it.
What
are
some
you
know
pod
requirements
you
want.
We
take
that
and
then
we
add,
we
inject
our
own
little
sona
movie
sidecar,
so
that
it
will
redirect
the
results
to
the
aggregator.
So
each
individual
plugin
doesn't
have
to
do
that.
A
You
know
a
single
pod,
but
it
could
be
a
single
pod
over
every
single
node,
okay,
okay.
So
now,
let's
pull
this
up.
Let's
actually
see
it
in
action
a
little
bit,
so
if
we
do
sanibori
run
what
this
is
going
to
do
is
run
the
default
conformance
suite,
there's
actually
a
few
tests
in
the
upstream
that
are
a
little
bit
destructive
to
production
clusters.
B
A
B
A
A
We
don't
want
to
continue
until
it's
done.
So
as
it's
waiting,
it's
going
to
be
sending
out
a
bunch
of
information
just
so
you
can
kind
of
get
a
sense.
Are
things
progressing,
especially
namespaces
have
a
tendency
to
get
stuck
deleting.
I
don't
really
know
why
that
is.
I
think
there
have
been
a
number
of
issues
about
why
namespaces
take
so
long
to
delete
if
anybody
in
the
the
comments
has
an
idea,
yeah.
B
A
Yeah
there's
sometimes
so,
I'm
using
a
kind
cluster,
that's
kind
of
what
we
recommend
for
local
development
and
there's
times
when
a
namespace
just
will
not
delete
what
it
is.
Is
these
finalizers?
You
can
look
through
this
json
and
you
can
see
there's
certain
finalizers.
It's
waiting
for
just
something
you.
B
A
B
B
A
Thank
you
so
now
we're
going
to
go
back,
we're
going
to
run
in
quick
mode,
so
we're
just
running
one
test
and
we're
saying
to
wait.
One
of
the
improvements
we
made
lately
it
used
to
be
that
it
would
just
sit
here
and
wait,
but
a
lot
of
people
would
say
I
don't
is
it
hanging?
Is
it
sitting
there
for
an
hour?
Is
it
done?
Is
it
progressing
at
all,
so
we
added
more
feedback.
Just
like
in
the
delete.
You
got
to
see
the
actual
json.
Here
we
have
different,
more
user-friendly
input
or
output.
A
A
A
So
you
can
do
like
sono
buoy
status
at
any
point,
if
you,
you
know,
cut
off
that
command
in
the
first
place,
so
that
you're
not
waiting
on
it
anymore.
You
can
always
just
do
something
status
and
get
that
same
information
and
the
the
super
pro
tip.
If
you
add
the
the
json
flag
here,
I'm
going
to
pipe
this
through
jq,
so
it's
nicer
to
read:
you
can
get
a
little
bit
of
extra
information,
especially
if
there's
like
test
failures.
A
A
Itself
has
a
bunch
of
options
if
you
just
do
retrieve
it.
If
you
just
do
retrieve
it's
to
download
to
you
know
a
time
stamped
tarball,
you
can
capture
that
or
you
can
just
refer
to
it
if
you
want
copy
paste,
but
you
can
also
specify
like
what
file
you
want
it
to
be,
or
this
this
is
maybe
my
favorite
one
is.
You
can
just
extract
everything
yourself,
so
it's
going
to
download
the
tarble
and
extract
it
into
the
output
folder.
B
A
Absolutely
yeah,
so
this
is
one
thing
with
sanibe.
The
sonaboo
itself
doesn't
have
like
an
api,
it
exposes,
and
that's
really
for
the
security
concerns.
We
don't
want
to
have
to
worry
about
authentication
in
any
other
way,
so
the
only
way
that
the
cli
is
operating
with
the
aggregator
is
just
through
the
kubernetes
api
itself.
So
all.
B
A
From
you
know,
some
like
aggregator
exposed
endpoint,
that's
actually
on
the
the
pod
itself,
so
that
this
aggregator
is
updating
its
own
metadata
as
we
go.
So,
let's
see
key
controls
get
pods
and
sonography,
and
you
can.
You
can
always
bounce
back
and
forth
between
cube
control
and
sauna
buoy
for
for
some.
A
Labels
here
showing
that
it
is
the
aggregator
all
these
things
about
progress
and
the
status
are
just
annotations
on
the
pod.
But
it's
doing
that
so
the
worker
is
getting
that
information
and
then
securely
talking
to
the
aggregator.
They
agree
on
like
a
key
and
they.
So
it's
all
https,
I
think,
while
it's
in
the
cluster
but
other
than
that,
once
exposed
out
into
the
cluster,
it's
just
through
the
normal
cube,
kubernetes.
B
A
Well,
there
is
a
a
kubernetes
api
endpoint
that
allows
you,
I
think,
to
like
run
remote
commands
and,
and
so
what
you're
doing
is
you're
like
telling
it
to
run
a
tar
command
that
just
spits
out
the
tar
data
of
this
file
that
you
expect
to
be
there
cool
and
and
so
yeah.
So
you
kind
of
get
away
with
a
little
nifty
bit
of
cheating
there
right.
A
So
once
you
have
so
we
have
that
output,
car
or
out
dot
car
okay.
So
if
you
have
that
tarball,
you
could
open
it
up
and
explore
there's
tons
of
data
in
there,
but
you
don't
have
to
do
that.
You
can
also,
if
you
just
hand
that
car
ball
to
a
co-worker
or
you're
looking
at
it
in
your
ci
system,
you
can
do
sanibuy.
B
A
And
it
will
automatically
look
into
the
tarball
and
see
what's
going
on
so,
at
least
as
far
as
the
plug-ins
are
concerned,
we
actually
have
an
open
issue
about.
Can
we
surface
some
more
useful
information
here,
especially
it
would
be
useful
in
like
a
broken
cluster
situation.
To
know
was
the
api
server
responding?
Were
these
nodes?
Responding
are
there,
you
know,
failed
pods,
but
here
you
can
see
the
e
to
e
plug-in
ran
it
passed.
There
was
one
test
that
passed
all
these
other
5666
were
skipped.
A
We
actually
have
a
blog
post
on
this
with
sonibu,
because
that
was
our
like
one
of
our
most
common
questions.
Why
are
there
so
many
tests
skipped?
Even
when
you
do
the
full
conformance
suite
you
run
300
or
so,
but
you
still
skip
5200
5400
and
that's
just
because
so
many
of
these
tests
are
like
feature
specific
they're,
not
related
to
how
you
expect
every
cluster
to
behave.
You
can
always
run
those.
You
can
always
specify
what
you
want
to
run.
B
A
A
A
So
we
just
added
an
e
to
e
command,
which
is
a
way
to
get
those
test
lists
and
it's
checking
what
version
of
kubernetes
I
have,
because
these
test
lists
change
every
version
and
in
default
it
will
just
spit
out
the
test
list.
You
could
scan
through
that,
but
there's
a
lot
of
data
there.
So
you
can
also
do
this
mode.
A
A
Some
tests
check
privileged
behavior
sigs
have
their
own
tags.
So
you
can
play
with
this,
though
the
nice
thing
about
the
ede
mode.
If
you
were
really
trying
to
figure
out
what
you
or
your
company
wants
to
test,
you
can
double
check
that
these
things
look
right
because
you
can
say
focus,
let's
say
conformance.
A
A
B
B
A
A
A
A
A
B
And,
dare
I
say
like,
of
course,
with
a
new
feature,
we're
kind
of
playing
around
we're
working
with
a
support
team
to
try
to
figure
out
ways
to
help
them
with
that
support,
diagnosis,
use
case
and
where
john
has
done
an
excellent
job
of
making
a
poc
here,
sling
try
it
out,
but
of
course,
if
folks
are
using
it
and
you
have
a
ui
request-
we
are
if
there
are
users
using
it
for
specific
use
cases.
We
are
delighted
to
tweak
the
ui
if
it
makes
it
easier
to
use.
A
Yeah
absolutely
yeah.
We
were
trying
to
solve
a
problem
and
you
know
I'm
assigned
a
movie
dev
now,
but
before
I
work
directly
on
sanobui,
I
was
just
on
a
team
at
another
company
using
sonibui
to
try
and
run
all
these
conformance
tests
constantly.
So
I
do
at
least
understand
some
of
those
those
concerns
about
trying
to
run
things
faster,
trying
to
run
get
the
results
out
in
a
predictable
way.
So
we've
tried
to
add
those
things
over
the
last
couple
years,
always
room
for
improvement,
though
well
speaking
of
those.
B
A
Where
you
want
to
know,
if
you're
passing
conformance
but
you're
not
trying
to
get
certified
kubernetes,
you
kind
of
are
being
resource
conscious,
we
just
added
a
new
mode,
there's
actually
a
sanibui
modes
command.
Now
too,
if
you're
curious,
what
those
modes
mean,
sunbuoy
mode
spits
out
all
that
data,
so
first.
A
A
The
defaults
here
is
it's
non-disruptive
conformance,
so
we
exclude
a
couple
tests,
but
otherwise
we're
just
focusing
on
conformance
tests
and
they
are
run
in
series
because
there
are
some
tests
that
have
to
be
run
in
series
there,
but
we
just
added
this
one.
This
conformance
light
excuse
me
and
the
focus
performance.
Oh
so
we're
running
all
the
conformance
tests,
but
we
skip
a
whole
bunch
of
tests.
It's
truncated
here,
there's
a
verbose
flag,
but
we're
skipping
like
30
tests,
and
we.
B
A
A
And
we
chose
the
fastest
test
so
in
the
happy
path,
if
you're
running
this
in
ci-
and
you
expect
it
to
pass,
but
you
still
want
to
run
it
you
can
get
instead
of
like
a
30
minute
to
an
hour
run.
It's
like
six
minutes
is
what
I
had
locally.
So
this
can
be
really
useful.
I
think
for
for
saving
people
some
ci
time.
A
Pretty
close
so
there
it
definitely
can
run
on
all
these
different
platforms.
There's
a
sometimes
there's
a
known
issue
so
like
on
google
cloud
platform,
there
was
like
a
permission,
a
role
binding.
That's
just
not
present.
We
just
tell
you
how
to
resolve
that.
If
you
find
a
problem
with
a
particular
provider,
let
us
know
so.
There
was
an
issue
with
running
on
docker
desktop
because
of
the
way
docker
desktop
works,
but
otherwise.
A
A
A
A
Set
up,
they
give
instructions
for
how
they
set
up
their
clusters,
how
they
run
the
conformance
test
yeah.
So
this
person
is
using
sonibui,
they're
kind
of
doing
it.
Interestingly
they're
pointing
out
the
fact
like
the
cli
is
really
for
you
as
the
user.
You
know
at
your
local
machine,
but
really
what
powers
the
run
itself
are.
The
images
that
create
those
pods.
So
all
of
that's
controlled
by
this
just
a
yaml
file.
You
can
actually
see
that
if
you
do
like
sanibuji,
let's
do
just
like
our
command
before
mode.
B
A
There's
an
environment
variable
set
on
that
container
for
e
to
e
focus,
for
which
test
to
focus
on
so
and
there's
a
lot
of
extra
information
here,
99
percent
of
the
time
you
don't
need
it.
Some
power
user
cases
really
do
want
to
tweak
something
there.
In
the
most
common
situations,
though,
we
just
add
a
cli
flag,
if
it's
really
useful
for
enough
people,
but.
B
A
A
I
think
there's
a
few
other
things
like
cube
test.
I
don't
have
a
lot
of
experience.
I
know
that
kubernetes
test
grid,
I
think,
uses
cubetest
a
lot,
because
cubetest
has
some
other
functionality
to
create
the
test.
Cluster
then
run
tests,
then
tear
down
the
test.
Cluster
sonobuoy
is
just
about
getting
those
results
and
trying
to
make
things
really
easy
to
integrate
with
other
systems
and
other
plugins.
B
A
A
Here's,
let's
see
I've
lost
a
little
track,
so
we
saw
while
a
test
was
running.
You
could
see
progress
messages
like
I
said
here.
We
had
a
happy
path.
Nothing
was
failing,
but
that
the
progress
was
really
useful
because
of
the
fact
that
I
said
a
full
conformance
run
can
take
an
hour
if
you
have
a
large
cluster,
it
can
take
longer
than
that.
So
it's
really
nice
to
see
whether
or
not
you're
having
failures
as
you
go.
A
That
was
something
that
was
added
close
to
maybe
about
a
year
ago,
oh,
I
will
say,
while
something's
running
there's
also
a
logs
command,
so
you
can
do.
Let's
do
this.
You
can
launch
on
a
buoy
into
other
name
spaces
too.
So
I'm
not
going
to
do
a
delete
here,
I'm
going
to
say,
like
sony,
bui
run,
I'm
going
to
do
the
full
conformance
run,
but
I'm
going
to
run
it
in
another
namespace
called
called
full
conformance,
I'm
going
to
run
it
asynchronously.
A
A
A
A
A
The
actual
ede
plugin
looks
like
right.
There's
some
pod
spec.
We
want
to
make
sure
to
run
on
a
linux
node,
that's
just
a
requirement
of
the
ede
container.
It's
spitting
out
results
in
junit
format.
A
lot
of
these
things
you
don't
have
to
worry
about,
but
you
can
see
the
plugin
is
really
a
a
pod
spec
and
a
container
spec.
That's
telling
you
how
to
run
it
and
down
here
the
image.
Okay,
so.
B
B
A
Very
careful
with
that
command,
so
nothing's
in
here.
What
I'm
going
to
do
is
I'm
going
to
copy
all
the
stuff
from
what
we
had
in
the
sonically
plugins
reboot.
There's
an
example,
ede
skeleton
and
again
it's
early
days
on
that.
I
want
to
make
it
better,
but
you
can
see
that
you
get
a
bunch
of
stuff
for
free
here.
This
temp
file
is
just
something
locally,
that's
junk,
but
otherwise
you
get
a
docker
file
because
we
have
to
have
an
image.
A
You
get
a
basic
build
script
right
if
you're
really
developing
something
you'll
modify
that,
but
this
is
a
way
to
get.
You
know
off
the
ground
really
really
fast
and
you
get
a
plugin.yaml
file.
That
is
a
plug-in
specification
and
then
the
actual
go
code
right.
So
this
is
a
plug-in
example
for
go.
We
do
have
a
plug-in
helper
and
the
skeleton
is
primarily
around
go
since
that's
a
lot
of
what
the
kubernetes
ecosystem
uses.
A
A
A
A
Now
you,
if
you
want
to
look
at
the
docker
file,
you
can
it's
adding
a
few
useful
things
in
here,
depending
on
what
your
tests
are.
I
had
a
jq
cube
controls
in
there
you
can
run.
You
can
do
a
lot
of
pretty
powerful
things
with
just
having
a
basic
plugin
that
has
access
to
cube
control
right.
You
can
script
all
sorts
of
cube
control
commands
or
you
can
be
piping,
the
actual
json
data
through
jq,
and
then
you
know
making
decisions
on
that.
B
A
So
test
main,
if
you
know
about
go
testing,
this
isn't
a
test
itself.
It's
just
the
stuff
that
kind
of
sets
up
all
the
tests-
and
we
added
some
nice
nice
to
have
features
here
so
that
you
get
them
automatically.
When
you
start
your
plugin,
that
includes
progress
through
sonibui
so
that,
if
you
add
a
bunch
of
long
tests
or
you
have
100
200
tests,
you're
going
to
get
progress
messages
as
they
complete.
A
B
A
A
We're
going
to
test
pod
listing
from
cube
system,
for
instance,
so
this
is
what
I'm
talking
about
getting
the
kubernetes
client
for
free.
You
know
it's
just
so
easy
you
just
in
this
function.
You
happen
to
have
a
config
object
and
you
call
client
on
it
automatically.
You
have
a
kubernetes
client,
so
here.
A
A
A
Yes,
absolutely
so
you
know,
and
there
is
a
client
go.
You
know
a
library
you
can
use
to
talk
to
the
api
server.
If
you've
done
a
lot
of
projects
using
clientgo,
it's
definitely
doable
and
it's
very
strong
client,
but
there's
so
many
api
endpoints.
It
can
be
a
little
frustrating
and
all
the
different
versioning
and
all
the
different
dependencies,
and
you
have
to
set
it
up
because
it
needs
to
know
how
to
talk
to
the
api
server
if
you
can
avoid
it
like
I,
I
appreciate
how
much
easier.
B
A
A
A
B
A
Just
test
hey
make
sure
none
of
those
pods
in
cube
system
are
in
a
failed
state
right,
make
sure
they're
all
running.
Those
are
all
all
things
you
could
do.
You
have
the
pod
list
now
just
iterate
over
it.
However,
you
want,
and
then
we
threw
another
test
in
here,
just
as
an
example,
same
sort
of
thing
we're
in
a
pod
list
from
cube
system.
Here
I
wanted
to
make
it.
A
Every
test
has
excess,
so
if
you
have
some
useful
helper
object
that
you
want
for
a
test
or
a
series
of
tests,
you
can
always
pass
that
in
here.
I'm
using
that
to
send
async
messages
that
just
every
five
seconds
mention
yeah,
I'm
still
waiting
still
waiting
all
right,
but
so
all
we've
done.
We
edited
the
plugin.yaml.
We
really
just
changed
the
name
right.
That's
the
only
thing
we
did,
but
now,
if
we
do
have
to
delete
sunbury
delete
and
it's
going
to
be
delete,
namespace.
A
Is
the
cluster
still
even
alive?
Maybe
it
got
destroyed
by
ci.
Sonobui,
though,
is
constantly.
It
always
gets
more
information,
so
that
in
that
tar
ball,
you
can
actually
debug
the
problem.
Even
if
the
cluster
doesn't
exist
anymore,
it
doesn't
obviously
have
everything
you
can't.
You
can
configure
it
to
get
a
lot
more,
but
we'll
see
what
they
have.
It's
kind.
B
Of
about
that
last
little
reference
in
the
last
little
thing
and
of
course,
soundably
totally
cleans
itself
up
afterwards.
It
deletes
all
the
namespaces
it
deletes
whatever
it
creates
to
do
testing
right,
you
have
the
destructive
tests
that
are
like
you
can
see.
Things
happen
to
your
cluster,
but
beyond
that
it
doesn't
touch
anything.
A
A
A
That
the
yaml
file
does,
is
it
tells
what
image
to
run
and
how
to
run
it
like
the
container
and
the
pod
configuration
excuse
me
what's
in
it,
okay,
so
we're
going
to
see
status
here
pretty
soon.
I
know
I
wrote
it
so
that
long
test
would
slow
things
down,
so
we'll
actually
see
some
amount
of
progress
here.
A
Oh,
I
didn't
build
the
image.
Oh
yeah
see
silly
demo
right.
So
it's
it's
saying
it's
running,
it's
going
to
be
waiting
for
a
while
and
this
would
actually
fail,
but
this
is
another
area
we
want
to
improve
things
if
we
enter
an
image
air
pull
state,
which
is
what
we've
done
here.
It
would
be
nice
to
surface
that
information,
a
little
faster
and
a
little
bit
better.
So
let's
do
this
on
a
booty
delete,
so
the
full
demo
is
remember.
B
A
B
B
A
A
When
someone
pushes
a
change,
you
don't
want
it
to
reference,
your
old
plug-in,
that's
already
been
pushed.
You
know
to
docker
hub,
you
just
build
it
locally
in
the
ci
system,
load
it
into
your
kind
cluster,
that's
also
in
the
ci
system
and
then
run
your
test,
and
unless
you
set
the
image
pool
policy
to
always
it'll
just
use
that
one
that's
already
pulled
pre-pulled
onto
the
node.
So
that's
how
you
can
keep
testing
in
ci
without
having
to
constantly
put
like
dev
tags
on
all
your
images.
A
A
This
is
all
the
output
from
like
go
test.
This
is
what
it
looks
like
when
you
do
out
go
test
with
that.
Json
flag
and
sonibuis
supports
that
type
of
output.
So
what
the
plugin
is
doing?
It's
running
the
go
tests,
it's
outputting
it
to
standard
out
so
that
we
can
see
it
here
in
the
logs,
but
it's
also
saving
it
to
a
file
which
sanibui
will
process
as
the
results.
A
A
A
Now
it
already
is
integrated
into
the
cli
to
where
you
can
see
how
many
tests
pass
failed.
It
gives
you
a
list
of
failures
there,
and
the
same
thing
was
true:
if
you
had
done
sonibury
status,
we're
gonna
type
that
through
jason
jq,
it's
all
here
as
well,
so
as
it's
running
through
either
the
status
command
or
the
results
command.
Your
ci
system
can
hinge
on
whether
or
not
it's
seeing
failures
there
or
getting
that
list
of
failure
and
logging
them.
B
A
A
A
It
lists
the
definition
here.
This
is
good
to
have,
especially
again.
If
you're
handing
this
tarball
to
somebody
and
say
my
plugin
failed.
What
was
the
problem?
They
are
going
to
ask
questions
like
well,
how
was
the
plugin
configured
they
can
go
through?
All
of
that,
they
can
also,
though,
ask
questions
about
what
the
status
of
the
cluster
was
like
right.
So
that's
where
you
can
look
at
all
this
extra
data
here.
B
B
A
A
You
were
working
with
what
the
pull
policy
was
be
careful
with
the
image
pool
policy.
That
would
be
a
a
pro
tip
as
well.
If
you
are
iterating
over
and
over
on
the
same
project,
make
sure
if
you're
not
bumping
the
tag
of
your
image,
you're
saying
image
pool
policy.
Always
there
have
been
a
lot
of
wasted
hours
on
my
part
because
ran
the
plug-in
fixed,
the
found,
a
bug
fixed
it
and
just
didn't
seem
like
the
fix
was
working
and
I
was
driving
myself
crazy.
A
To
pull
the
fix
so
and
then
you
can
also
get
pod
logs
by
default.
Sonobuoy
will
get
the
pod
logs
from
sonibui
namespace
itself.
Those
I
mean
those
are
important.
You
can
definitely
see
what
the
aggregator
that's
what
these
are
just
all
the
aggregator
logs
right
now,
but
you
can
also
see
inside
of
cube
system
by
default.
A
A
You
can
look
at
the
files
here,
they're
named
api
services
or
the
component
statuses,
the
name
spaces
the
nodes
and
then
in
the
namespace.
You
can
see
the
namespace
objects
too.
So
you
can
see,
then,
if
you're
having
an
error
with
a
pod,
you
can
check
the
pod
status.
You
can
actually
look
at
what
the
api
would
look
like
and
that's
because
it
may.
A
B
A
The
power
of
sonicway,
I
think
that
we
get
this
stuff
automatically
and
we
integrate
we're
so
easy
to
integrate
with
all
these
different
systems.
Okay,
so
yeah,
then
you
just
go
from
here.
You
add
whatever
tests
you
want,
you
have
access
to
a
client,
you're
automatically
getting
name
spaces
you're
automatically
sending
results
to
sonobui.
A
You
kind
of
got
all
the
things
and
you
just
start
writing
your
code
right
and
I
just
copied
the
code
over
there.
That's
how
that
worked.
So
I
think
it's
it's
really
nice
way
to
get
set
up,
and
especially
if
you're,
writing
pocs
or
something
like
that.
Don't
waste
your
time
writing
you
know
the
docker
file
from
scratch
and
a
build
script
from
scratch
and
a
plug-in
definition
from
scratch,
just
copy
all
that
we're
trying
to
make
it
easier,
because
we
would
love
for
people
to
make
more
plugins.
A
B
B
Yet
it
felt
like
folks,
didn't
really
know
that
so
our
strategy,
the
more
we
thought
about
it,
it
just
it
feels
like
so
for
for
non-kubernetes
cluster
testing,
it's
very
common
to
write
automated
tests.
If
you
want
to
make
sure
that
your
app
works,
you
write
an
automated
test
to
kind
of
validate
that
your
app
is
working
in
a
certain
situation.
B
B
B
We
see
so
much
potential
for
this,
so
we're
hoping
that
if
folks
hear
how
easy
we're
trying
to
make
it
and
how
interested
we
are
on
in
iterating
and
improving
the
experience,
we're
just
hoping
we'll
see
a
big
explosion
of
these
plugins
being
created.
John,
you
know
we're
very
responsive
on
the
sonubui
channel.
So
come
see
us
there.
If
you
have
any
questions
about
this,
john
already
is
her.
You
know
folks
have
seen
this
john's
heard.
A
Yeah,
you
mentioned
the
plug-ins
and
hoping
for
more.
I
I
had
mentioned.
I
would
come
back
to
this
at
some
point.
There
is
so
there
the
vmware
tons
of
sonabooi
is
the
main
repo,
but
we
have
a
sonicboy
plugins
repo
with
a
bunch
of
different
every
other
plugins.
You
can
take
a
look
at
there
are
various
states
of
maturity
or
support.
A
So
there's
a
bunch
of
tools
that
already
exist
out
in
the
world
for
kubernetes,
who
can
was
one
of
those
and
bridget
had
done
this,
and
you
can
see
like
she's
just
using
the
who
can
binary
it's
a
a
tool
that
checks
based
on
kubernetes
policy
and
like
the
our
back
policy
who
can
list
resources
of
different
types
or
things
to
that
extent,
and
but
she
just
containerized
it.
She
scripted
effectively
what
data
she
wants
and
then
put
it
into
this
plugin
that
way.
A
Anybody,
she
could
just
say:
hey,
run
this
yaml
file
and
then
all
of
a
sudden
you
just
get
all
that
output
for
free
and
all
you
would
really
have
to
do
on
your
end.
If
you're
already
set
up
for
like
the
ede
test,
you
would
just
have
to
also
add
you
know
you
can
or
you
can.
You
know
you
can
run
multiple
plugins,
so
you
can
run
the
ede
plug-in
and
you
can
and
you
can
run
your
other
custom
tests.
A
A
Have
we
have
cube
hunter,
so
that's
a
security
plug-in
from
aqua
security.
Again
they
had.
I
think
they
even
have
the
image
already
made.
That's
why
there's
not
a
lot
in
here,
so
we
took
their
open
source
repo.
I
think
and
just
built
it
and
that's
why
it's
in
the
sonogui
repository
here
or
registry,
but
really
it's
all
their
open
source
code.
A
I
was
just
going
to
say
in
the
slack
channel
someone
was
just
mentioning.
We
put
it
in
this
examples,
because
they're
they're
really
kind
of
like
minimal
plug-ins.
They
were
really
supposed
to
be
more
inspiration,
but
this
command
runner.
I
made
this
one
so
long
ago
because
I
kept
getting
the
question
of
like
oh.
Could
you
do
this,
or
could
you
do
that?
So
I
made
this
in
like
an
afternoon
to
point
out.
A
You
can
do
anything,
just
put
it
in
a
container
and
what's
funny
about
this
one,
it's
just
a
bash
script
and
what
it
does
is
it
takes
each
like
argument
as
a
command
runs.
The
command
on
the
shell
saves
the
output
to
a
file
and
then
like
keeps
going.
So
it
doesn't
say
like
pass
fail
right
now,
but
I
made
an
example
here
so
like
command
runner,
and
I
made
sure
the
image
had
like
jq
and
cube
control
in
it
and
stuff
like
that.
A
So
you
know,
like
echo,
hello,
world
cube
control
cluster
info,
but
you
could
also
do
things
about
like
checking
the
ip
status
of
of
the
node
or
checking
journal
config.
All
sorts
of
you
just
do
anything
right.
So
if,
if
you
can
have
like
you
were
saying
eleanor,
if
you
could
have
a
kubernetes
expert
run
their
custom
script,
they've
developed,
you
can
put
it
into
a
plugin
and
then
it's
way
easier
to
pass
around
and
know
like
it's
very
standardized.
How.
B
And
again,
the
idea
is,
you
know
you
have
this
kubernetes
expert
they're
already
doing
something?
Why
not
automate
what
they
do?
You
you
get
to
use
their
time
on
on
something
maybe
more
specialized
and
then
they're.
The
script
is
more
reliable,
doesn't
matter
if
they're,
sleepy
or
not
alex
asked
a
really
great
question
about
whether
we're
working
with
open
source
teams
or
commercial
companies
to
have
them
develop
plug-ins.
So
our
our
strategy
on
this
is
we're
both
a
vmware,
so
there's
enough
vmware
teams
that
are
excited
to
kind
of
do
things
like
this.
B
I
think
I
mentioned
the
support
team,
we're
working
with
our
support
team
to
do
like
a
support,
diagnostic
tool
and
we're
thinking
of
trying
to
do
like
a
post,
restore
kind
of
confirmation
that
some
of
our
tan
zoo
clusters
work.
So
our
strategy
is
kind
of
the
work
we're
going
to
do
we're
going
to
basically
be
using
to
improve
our
open
source,
product
and
kind
of
they'll
be
our
first
guinea
pigs,
basically
john,
and
I
will
learn
how
to
make
the
best
experience
for
everyone
again.
B
B
A
Yeah
speaking
of
this
good
segue
into
this
other
plugin
that
we're
starting
to
make
this
requirements
check,
I
kind
of
really
struggled
with
the
name,
because
this
is
where
I
think
a
lot
of
other
teams
or
other
companies
would
still
find
this
really
useful,
and
our
hope
was
so.
I
do
comment
that
it's
a
work
in
progress,
we're
trying
to
make
this
like
a
library
of
really
useful
things
that
anybody
might
want
to
check,
and
the
kind
of
the
inspiration
for
this
was
eleanor
had
mentioned.
Different
products
have
a
whole
bunch
of
checks.
A
A
We
don't
want
you
to
do
that,
but
maybe
you
run
it
automatically,
but
you
still
had
to
somebody
at
some
point
had
to
write
it,
and
it
was
probably
manual
we're
trying
to
write
all
of
those
little
bits
that
you
might
want
to
check
into
this
library
so
that
you
can
just
say
things
like
well.
I
need
this
version
of
this
tool
to
at
least
be
this
right.
A
I
need
at
least
version
1.2
of
foo,
and
I
also
want
to
make
sure
that
I
have
three
nodes
and
I
have
at
least
this
much
memory,
and
you
know
just
take
that
list
of
things
you
want
to
check.
They
can
be
in
logs.
They
can
be
in
the
api
data,
whatever
it
is.
If
we
put
those
kind
of
common
checks
in
here,
then
it
is
a
lot
easier
for
like
building
blocks
for
people
to
make
their
own
plug-in
and
their
testing
right.
That
skeleton
I
showed
you.
A
But
as
far
as
actually
running
checks,
you
do
still
have
to
go
and
get
the
pod
and
you
still
have
to
go
and
check
what
the
version
is
manually
and
know
the
field
for
it.
So
this
is
kind
of
that
last
step
of
trying
to
make
whatever
check
you
want
to
run
even
easier
to
do
so
that
you
can
just
kind
of
plug
and
play
your
test
is
just
say:
hey
yeah
use
this
one
for
me,
use
this
one
for
me,
I
need
four
instead
of
three
nodes,
that
sort
of
thing.
B
Yeah,
exactly
and
because
already
like
it's
so
the
the
most
the
plugins
were
created,
I
think
before
we
were
saying
how
can
we
make
it
as
easy
as
possible
there?
There
were
some
of
that.
I
think
then
john
made
his
ede
skeleton,
which
now
we're
that's
what
we're
going
to
use
with
the
vmware
teams
and
anyone
who
wants
to
now
who
wants
to
be
kind
of
more
of
an
alpha
tester
that
we'll
do
that.
B
But
then,
as
we
see
these
alpha
testers,
making
these
plug-ins
exactly
as
john's
saying,
we
want
to
pull
out
those
common
components
to
make
it
as
easy
as
possible
and
I'd
love
to
see
everyone
in
the
kubernetes
community,
using
saniboy
to
do
checks
to
build
on
by
the
way
those
common
components.
As
john
said,
it
might
be
a
version
check,
but
imagine
even
like
knowing
whether
there's
a
load
balancer
in
your
cluster.
A
Yeah
I
wanted
to
just
mention
windows
as
well.
I
don't
know
how
many
people
in
the
chat
might
be
concerned
about
windows.
The
kubernetes
testing
of
windows
was
always
kind
of
difficult,
there's
a
whole
repo
about
how
they
they
test
it
and
again,
sonography
tried
to
make
those
steps
as
trivial
as
possible.
Right,
try
to
front
load
all
the
difficult
stuff,
and
so
there
are.
A
You
can
run
the
e
to
e
plug-in
or
the
kubernetes
end-to-end
test
against
windows
nodes,
there's
a
blog
post
on
it
on
the
sonobui
site,
and
we
kind
of
go
through
like
how
we
built
this
and
what
complications
there
are
and
why
running
on
windows
is
different,
but
yeah.
If
windows
is
something
that
interests.
You
know
that
sonibus
supports
that
as
well.
B
B
A
B
A
A
B
A
Develop
their
own
tools
for
a
multitude
of
reasons.
Maybe
someone
just
had
more
experience
in
one
area
or
another,
and
they
didn't
mind.
You
know,
writing
their
own
tool
or
they
already
had
it
for
some
other
reason.
So
yeah
you
can
flip
through
here.
If
you're
curious,
it's
also
just
kind
of
neat
to
see
who
is
conformant.
B
A
It
is
a
conformant
cluster,
it
it
will
run
sonic
movie,
just
fine
right.
It's
we're
really
just
running
containers,
that's
all
we're
doing,
and
the
aggregator
and
the
the
worker
they're
the
same
image.
They're
just
running
with
different
commands,
there's
really
no
magic
sauce.
That
makes
it
difficult
to
use
the
value
of
it
is
just
trying
to
standardize
these
things
so
that
the
bits
talk
to
each
other
in
the
same
way,
so
that
it's
it's
easy
to
plug
into
ci
or
it's
easy
to
run
locally.
It's
easy
to
script
off
of
that.
A
B
A
Yeah
and
in
the
sonibui
github
people
make
issues
with
questions
all
the
time.
That's
not
a
problem
I'll
get
to
you
there
for
sure.
I
go
through
that
pretty
much
every
day
and
yeah.
You
know
it
will
try
and
help
you,
even
if
it's
not
per
se
a
sonobu
issue.
I
remember
somebody
was
trying
to
run
a
windows
cluster
through
aws
and
they
were
having
a
problem
right.