►
From YouTube: Kubernetes SIG Testing 2019-06-04
Description
A
B
B
So
it's
gonna
say
I
feel
like
since
we
don't
really
have
an
agenda
for
today.
I
personally
have
nothing
prepped,
so
I
feel
like
it
would
be
more
respectful
of
everybody's
time.
If
we
call
this
meeting
cancelled
unless
there's
anything
pressing
that
we
want
to
discuss
so
typically
the
way
I
do.
This
is
I.
Wait.
You
know
five
ish
minutes.
B
C
I'm
working
on
kima
I'm,
an
architect
and
kima
project
and
we're
using
prowl
to
build
kima,
basically
we're
right
now
trying
to
to
monitor
the
space
and
to
see
what
happens
with
Prato.
Where
does
it
go
we're
eagerly
waiting
for
the
first
release?
I
know
it's
somewhere
on
the
roadmap,
where
we're
right
now
planning
so
because
we
just
use
a
little
subset
of
the
services
that
you
use.
C
We
have
a
request
from
our
team
to,
for
example,
to
start
using
test
grid,
so
we
talked
to
some
some
guys,
trout
next,
that
we
could
be
allowed
to
just
publish
our
results
since
we're
open
source
onto
your
tests.
Credit,
the
dashboard
by
adding
a
pull
request
and
just
pushing
our
results
there,
so
I
think
we're
going
to
do
that
pretty
soon
we're
also
looking
into
using
tied
very
soon.
C
B
Just
in
case,
we
do
end
up
using
this
recording,
I'm
going
to
say
much
feel
now,
but
I
reserve
the
right
to
decide
or
not.
This
is
just
sort
of
an
informal
thing
and
maybe
post
it
just
in
case
hi.
Welcome
to
the
weekly
suggested
meeting
today
is
Tuesday
June
4th
American
burger
city
beard,
and
you
are
all
here
adhering
to
the
kubernetes
code
of
conduct
by
not
being
jerks.
You
can
go
wash
yourselves
doing
so
because
this
is
a
publicly
recorded
medium
and
this
will
be
posted
to
YouTube
later
on.
B
B
So
anyway,
I
think
it's
cool
that
kima
is
using
prowl,
where,
like
we
haven't,
necessarily
been
super
pushy
about
you
don't
have
like
a
product
manager,
so
it's
been
really
encouraging
to
see
people
pick
this
up
and
use
it
on
their
own.
How
would
you
feel
about
like
maybe
giving
a
presentation
to
this
group
sometime
later
about
sort
of
what
your
setup
looks
like?
B
B
C
Also
happy
with
a
point:
one
I
just
want
an
official
release
in
the
github.
That's
what
we're
looking
for,
because
right
now
we're
cherry-picking
on
like
we
go
to
the
announcement
MD
and
check
for
the
feature
that
we
want.
We
check
for
the
date
that
is
in
front
of
it
and
then
check
with
that
date
cross
her
and
reference.
One
of
the
commits,
like
possibly
the
last
commit
of
that
day
or
the
next
day,
and
use
that
as
a
base
for
our
new
deployment
of
the
update.
C
Basically,
okay,
so
I
would
I
would
be
happy
if,
if,
for
example,
you,
instead
of
just
using
the
announcement
empty
like
make
test
change,
log
empty,
maybe
and
use
instead
of
May
13th
2019
just
make
it
version
point
zero,
zero
one
right,
I
would
be
nice
for
us,
because
then
we
can
say
like.
Oh,
this
is
the
latest
technic.
We
just
deployed
its
version
zooms
or
one.
That
makes
a
lot.
C
B
C
I
mean
this.
That
is
a
big
thing
for
us
as
well,
since
I'm
I'm
from
Mississippi
I
mean
chemo
ASAP,
so
we're
we're
heavily
looking
into
what
else
Google
is
doing
as
well.
So,
for
example,
I
moved
one
of
our
development
teams
just
last
year
from
a
Jenkins
installation
to
Google
cloud
built.
For
example,
I
know
that
Aaron
from
from
and
and
Christy
have
been
in
the
carpool
team
or
are
still
I,
don't
know
about
the
current
set
up,
but
but
they're
now
doing
Tecton
yeah.
A
C
What
is
gonna
be
pushed
as
maybe
the
sole
CI
CD
are
they
gonna
be
merged
at
one
point,
because
I
know
Chris,
we
talked
to
Christi
about
Tecton,
maybe
leveraging
prowl
in
the
background,
or
probably
yeah,
probably
in
the
backend
and
Fisk
Tecton,
just
being
the
front
end
of
that
and
I
I,
just
like
it's
very
interesting
to
see
that
I'm
in
all
three
spaces,
kind
of
and
I'm
looking
where
things
are
going,
and
it's
very
interesting
because
also
but
not
having
release
its
kind
of
I.
Don't
I,
don't
know
where
we're
recovered.
C
B
Like
I
said
this
was
this
was
not
something
we
ever
built
with
a
plan
to
turn
into
its
own
project
or
product
yeah.
It's
only
like.
Oh
my
god.
Managing
a
couple
hundred
Jenkins
instances
is
really
terrible
and
painful.
We
need
to
do
something
better,
and
so
it's
largely
evolved
to
to
meet
our
needs,
but
the
fact
that
other
people
have
picked
it
up
organically
and
adopted
it
makes
me
think
we
should
be
more
mindful
of
all
of
that.
Yes,
there
is
some
overlap
with
tacked
on
I
there's
also
like
overlap
with
github
actions.
B
So
I
know
from
a
for
tacked
on,
for
example,
like
I.
Don't
think
we
want
to
turn
prowl
into
this
crazy,
complicated
scheduling
thing
and
tacked
on
doesn't
Tecton
pipelines
seem
to
do
a
good
job
of
expressing
a
directed
acyclic
graph
of
different
jobs
that
work,
clothes
and
stuff?
So
maybe
you
could
just
have
like
prowl
trigger
Tecton.
If
you
have
a
lot
of
complicated
words
on
this
I
think
you
have
a
a
pipeline
controller
I
saw
a
line
of
PR
for
it
anyway,
I
don't
know
where
we're
at
and
actually
like
landing.
A
B
B
C
Have
a
lot
of
lot
of
plugins
attached
to
it
and
we're
typically
not
going
for
the
latest
version.
So
if
we,
if
we
see
announcement,
MV
and
I,
think
the
last
one
was
some
some,
how
mid-may
my
13th
or
something
I'm,
not
sure,
but
I
I
already
told
someone
today
because
we're
like.
Maybe
we
should
update
prowl
again
and
like
yeah,
but
never
picked.
The
the
latest
line
and
the
announcement
empty
picks.
C
Something
before
that,
because
we've
seen
typically
that
you
do
up,
there's
a
Frau
and
then
you
roll
it
back,
maybe
the
next
day
because
he
figured
out
it
didn't
work.
The
way
that
you
want
it
and,
like
you,
do
another
update
the
day
after
so
we're
always
waiting
for
a
more
stable,
safe
up
version.
So
like
two
three
days
without
any
changes
in
the
announcement
MB
and
that's
that's
a
good
good
release
for
us
to
use
it's
fun
using.
B
One
of
the
most
active,
most
chaotic
open-source
projects
in
the
world
as
your
guinea
pig
as
your
staging
environment
is
I.
Guess
we
tests
live
in
production
or
something
we'd
love
to
find
a
way
to
not
but
I.
Don't
I
gather
you're
sort
of
you're
using
us
as
your
staging
environment,
and
then
you
sort
of
do
the
same
thing.
I'm
guessing
we're
like
you
pick
something
you
try
it
out
and
if
it
doesn't
blow
up
great
yeah.
B
C
C
The
jobs
so
that
I
can
push
them
to
prowl
and
have
like
an
ad
hoc
run
off
of
a
job
definition.
So
that
has
helped
us
a
little
bit
as
well
and
that's
pretty
cool
but
I.
Think
there's
a
lot
of
stuff
in
testing.
Far
that
we
haven't
touched
that
or
because
also
I
inherited
kind
of
the
ownership
of
the
prowl
thing.
From
from
from
the
team.
That
basically
said:
hey
we
to
move
away
from
Jenkins
and
quickly
move
on
to
something
else,
and
they
decided
for
prowl
and.
C
Some
of
the
things
I
think
they
they're
just
not
there
because
they
ran
out
of
time
in
in
in,
in
terms
of
what
else
can
we
add
that
makes
our
life
easier,
so
we're
we're
looking
into
ways
of
what
can
we
get
from
from
the
proudest
infra,
ladies?
That
makes
our
life
easier
and
so
we're
cherry
picking
right
now
test
grid
as
one
tide
as
another
one,
just
seeing
what
else
we
can
use
from
you
guys,
okay,.
B
A
bad
test
grade
so
like
we
are
in
the
process
of
like
trying
to
turn
the
gears
to
get
approval
to
open
source
parts
of
that
or
rewrite
parts
of
it.
There
are
big
chunks
of
it
that
are
tied
up
in
a
bunch
of
magical
google
stuff
that
I
don't
understand.
I
I
tend
to
like
avoid
all
the
Google
internal
stuff,
so
that
I
can
talk
about
everything
that
I
work
on.
That's
just
sort
of
how
I
operate
and
working
on
that.
B
B
B
Maybe
keep
flow
that
like
they're,
not
actually
part
of
the
kubernetes
community,
but
since
the
only
people
who
can
support
prow
and
all
the
other
testing
for
right
now,
our
Googlers
like
will
carry
the
torch,
but
we
have
stood
up
like
a
Google
crowd.
Instance:
that's
for
google,
it
open
source,
I
think
we're
looking
at
things
over
there.
It's
unclear
to
me
whether
we
might
do
something
similar
for
test
periods
as
well.
B
B
C
We
when
we
talk
to
the
Googlers
at
Google
next,
they
were
saying
we
don't
need
an
open-source
test
grid
at
the
moment
for
chemo
to
be
able
to
publish
results
on
the
Gaydos
test
right
side.
So
apparently
they
said,
then
it
should
be
a
matter
of
adding
our
upstream
in
a
pull
request.
I
think
so
that
it
displays
on
your
website
and
the
only
requirement
that
you
guys
have
is
that
our
project
is
open
source,
which
keema
is
that's.
That's.
C
B
I
feel,
like
I've,
heard
a
lot
about
your
interest
in
trout
from
a
road
map
and
planning
perspective
and
I'd
like
to
figure
out
how
to
turn
prouds
more
formalized
sub-project
that
we
can
talk
about
and
include
you
as
contributors
over
there.
Caslen.
If
you
don't
mind
me
putting
it
on
the
spot,
you've
mentioned
you've
sort
of
wanted
to
show
up
as
a
newer
contributor
interested
to
help
out
sort
of.
Where
are
you
coming
from?
What
are
you
interested
in
and
how
could
we
help
you
out?
Yeah.
A
So
hello,
everyone,
I'm
caslen
I,
went
to
the
contributor
summit
at
cube.
Connie
you
and
I've
been
interested
in
contributing
to
communities
for
a
while
and
I've
been
considering
what
which
sake.
I
should
start
looking
into
I
have
a
background
in
testing,
because
I
was
a
Quality
Assurance
engineer
at
one
point
and
I
think
that
testing
is
really
cool
and
interesting.
So
I
thought
I
would
ask
about
cig
testing
and
try
to
learn
more
about
what
you
all
are
working
on
and
how
I
could
contribute.
B
So
I
feel
like
the
answer
I
want
to
give.
You
is
like
take
a
look
at
our
good
first
issue
and
Help
Wanted
issues
that
are
lived
in
the
testing
for
repo
and
see,
if
anything
in
particular
interests
you
there
we
and
I
can't
I.
Honestly
I
can't
speak
to
exactly
like
what
it
is
we're
doing
this
quarter
off
the
top
of
my
head,
just
because
Connie
you
is
I'm
still
kind
of
in
recovery
from
that,
and
also.
B
Yeah,
it's
working
on
a
lot
of
different
things,
but
I
think
like
to
Daniel's
point
about
sort
of
proud
becoming
its
own
project.
I
have
sort
of
a
longer
term
interest
in
getting
all
of
the
prow
stuff
into
its
own
sub
project,
or
it's
like
its
own
repo
and
getting
the
community
to
manage
like
the
prow
instance
and
the
community
to
own
all
the
infrastructure
instance.
There's.
B
A
bits
that
need
to
happen
for
that
another
place
that
I
find
new
contributors
often
have
a
fun
time
contributing
to
is
some
of
the
plugins
and
automation
related
to
like
workflow
on
this
project.
So
Prowse
plug-in
model
is
relatively
simple.
It
seems
like
it's,
maybe
a
nicer
area
of
code
to
try
comprehending
than
the
entirety
of
the
kubernetes
codebase,
and
it
could
be
that
you
have
some
itches.
B
You
want
to
scratch
as
a
result
of
interacting
with
pull
requests
and
stuff
on
the
communities
project,
and
you
may
find
that
there
are
contributions
you
feel
like
you
want
to
make
there
we're.
Also
like
I
know,
one
of
our
interns
just
updated
the
cat
plugin.
So
that
I
think,
if
you
slash
note,
you
get
a
grumpy
cat,
we're
like
totally
down
for
fun
stuff
like
that,
so
just
sort
of
get
your
feet
wet
and
understand
how
all
that
works
as
well.
B
B
So
browse
the
front
of
the
ship
and
then
like,
like
I,
was
alluding
to
it's,
not
exactly
a
cohesive
product,
but
it's
just
a
bunch
of
like
small
distributed
services,
many
of
which
are
kind
of
designed
to
work
together
to
do
sort
of
an
if
this
and
that
style
workflow
for
kubernetes
and
then
have
some
like
emergent.
Things.
Come
out
of
that,
and
so
a
common
way
of
this
happening
is
we
have
a
component
called
hook
which
takes
in
things
from
github.
A
B
Register
like
what
kinds
of
events
that
they
can
respond
to
and
some
of
those
plugins
included
things
like
so
the
cat
plug
and
I
alluded
to
earlier,
or
somebody
comment
slash
me
out.
It
will
go
talk
to
the
cat,
API
find
an
image
and
then
post
a
github
comment
to
the
appropriate
issue
with
pull
requests
with
that
image
awesome
we
have
a
dog
plug-in
as
well,
because,
of
course,
we
do
anecdote.
B
B
A
B
Which
will
go
off
and
trigger
jobs
somewhere
when
we
originally
wrote
it?
It
was
like
go
trigger
a
Jenkins
job
over
here
and
then
wait
for
Jenkins
to
report
back
that
it's
done
and
these
days
we
now
like
go
trigger
a
job
in
a
kubernetes
cluster,
so
yeah
like
if
this
than
that
but
forget
hub.
We
also
have
folks
from
cloud
bees
working
on
abstracting
away
that
48
subpart.
We
also
internally
run
part
of
an
installation
of
proud
that
runs
against
Garrett
or
Jared
I.
B
B
But
if
had
interest
to
you,
there's
that
and
then
like,
if
you're
interested
in
just
trying
to
go
after,
like
pretty
easy,
mechanical
checklist
type,
stuff
I
feel
like
I
need
to
file
an
issue
for
us
to
convert
our
Python
script
to
Python
3,
because
I
thought
to
as
end-of-life
fitting
at
the
end
of
this
year,
and
so
we
could
work
through
what
those
scripts
look
like
and
which
ones
you
want
to
modify
things
of
things
of
that
nature.
We
generally
are
responsive
in
the
sig
testing
channel.
A
A
B
B
A
B
B
C
B
C
One
more
thing
maybe
or
we're
looking
into
monitoring
of
prowl.
How
do
you
guys
do
that?
That
would
be
interesting
for
us
because
currently
a
crowd
crashes
and
it
doesn't
recover
sometimes-
and
we
only
find
out
because
it
doesn't
work
anymore.
So
if
someone
goes
in
and
restarts
it,
but
how
do
you
monitor
prowl
crashing.
B
B
B
C
B
I
mean
that
that
often
like
the
bot
doesn't
seem
to
be
doing
anything
from
a
random
community.
Member
is
our
first
heads-up
that
something
is
wrong.
I
do
think
that
we
have
like
health
Z
endpoints
on
all
of
our
crowd,
components
and
health.
C
I
understand
to
be
just
like
a
common
API
endpoint
the
cheating
the
hit
and
get
back
a
200
okay,
the
thing
is
up
and
running
I,
don't
know
if
I
feel
like
we
have
that
exposed
to
a
Prometheus
thing
at
one
point:
yeah.
C
We
we
don't
use
tide
and,
as
you're
saying,
metrics
are
pushed
by
tide.
That
might
be
one
of
the
things
why
we
don't
have
proper
monitoring
like
we
have
a
story
in
the
back
of
actually
tool
to
use
tide,
and
then
maybe
we
can
look
into
what
tide
does
for
Prometheus.
So
we
can
scrape
that
and
get
a
graph
and
a
dashboard
or
some
stuff.
B
But
I
feel
like
what
you
were
asking
is
pretty
bare-bones
like
is
it?
Is
it
alive?
Is
it
up
and
so
someplace
you
could
look
to
see
if
or
how
we
implement,
that
is.
Velodrome
is
sort
of
a
complicated
codebase,
but
if
you
go
to
Belgium
that
cait's
thought
I,
oh
really,
all
we
use
that
for
now
is
the
fact
that
it
has
Agra
fauna
instance.
That's
pointed
at
an
influx
DB
instance,
and
we
push
data
into
that
influx.
B
B
They
may
not
be
there
right
now,
I'm,
not
a
hundred
percent
short,
so
I
can
I
can
link
to
the
monitoring
stack
issue
that
Steve
brought
my
attention
to
as
place
for
you
to
take
a
look,
but
I
feel
like
tide
is
not
so
much
about
the
health
of
prow.
The
tide
metrics
are
literally
just
like
how
many
PRS
are
we.
B
C
Well,
we'll
have
a
look.
We
haven't
started.
Looking
too
it's
just
one
of
the
requests
that
we
have
like
hey.
We
we
noticed
prowl,
sometimes
doesn't
work,
and
we
know
that
it's
our
fault,
that
it
crashes,
but
we
don't
know
when
it
doesn't
work,
and
we
just
need
to
see
that
we
get
a
monitoring
stack
for
that.
Ok,
I've
gotta
run
I'm
losing
my
room.
Thank
you
all
for.