►
Description
Project Eirini update / demo by Julian Skupnjak of IBM and Mario Nitchev of SAP
Introducing EiriniX by Vlad Iovanov of SUSE
Agenda here: https://docs.google.com/document/d/1SCOlAquyUmNM-AQnekCOXiwhLs6gveTxAcduvDcW_xI
A
A
B
B
Good
to
see
everyone
as
well
I
don't
have
a
whole
lot
of
updates
other
than
yes.
We
are
counting
down
to
summit
in
a
few
weeks.
The
only
thing
I
wanted
to
mention
or
if
you
still
are
not
registered,
and
if
you
are
a
current
or
a
past
contributor.
Let
me
know
I
can
share
the
code
again
so
that
you
can
register
as
a
contributor.
We
have
pretty
many
days
leader
activities
with
contributor
summit
user
day,
hackathon
all
the
efficient
of
awesome
events.
B
C
D
B
Unconference
organizers,
we
also
have
a
project
specific
slide
on
our
online
store.
So
if
you
go
to
the
foundry
online
colette
and
slide
store,
you
will
see
project
specific
t-shirts
and
stickers
like
bosch,
ireenie
and
quarks.
So
if
you
are
interested
in
ordering
any
of
those
and
get
them
shipped
through
the
venue
to
the
event
venue,
you
can
do
that
just
drop
it
and
drop
me
a
note
privately
on
the
instructions.
A
Excellent,
thank
you
going
up.
Yes,
we
have,
as
SWANA
mentioned,
there's
also
the
hackathon
we'll
be
running
this
search.
You
have
ideas
and
you
want
to
help
please,
as
you'll
see
in
the
arena
extension
there's
probably
tons
of
potential
hacks
to
be
done
to
be
on
and
to
be
worked
on
there.
So
just
seeding
ideas
here.
Ok,
any
question
for
anybody:
community
you
get
a
chance
to
pings.
Wanna,
live!
A
D
Dr.
max
thanks
see
ya
so
I
guess
it's
a
little
bit
quiet
but
I
want
to
point
out
a
few
things
see
if
deployments
released,
I
think
a
couple
of
major
versions
since
last
Capcom,
first
10
and
then
version
11.
Shortly
thereafter,
the
CLI
they
recently
released
version
6
about
46
and
they're
working
on
a
patch
release
with
some
loggin
refactoring
changes
in
there
as
well.
D
A
When
it
comes
to
the
updates,
I'll
try
to
get
some
info
from
the
first
factor,
we'll
get
there
for
extensions.
We
have
one
kind
of
big
announcement
and
it's
that
Stratus
finally,
graduated
seems
like
it
was
a
late
bloomer.
Actually,
it's
just
the
process.
I
think
Neal
was
hesitant
to
put
it
there
and
he
of
course,
had
some
vacation,
but
it's
finally
there
it's
graduated
us
in
an
email.
E
A
Excellent
Thank,
You
Neil
I
think
there
are
a
few
people
with
background
noise,
so
if
you
know
it
might
be
good
to
to
go
on
a
background,
a
little
bit
all
right
cool.
Thank
you
very
much,
alright,
so
and
then,
besides
that,
we
have
I
think
of
few
other
projects
that
are
retiring
and
also
I've,
graduated
I
listed
them
so
check
those
out.
A
F
A
Yeah
we
have
two
talks,
one
by
air,
Jules
and
Mo.
You
who
are
doing
the
nitty-gritty
details
on
Irina
and
they'll,
give
an
update
on
essentially
here
in
itself
and
where
it's
going,
where
it's
at
and
what
to
expect
and
so
on,
and
then
we'll
have
a
talk
by
Vlad
Souza,
who
is
gonna,
give
us
a
little
bit
on
a
weenie
X,
which
is
an
extension
project
for
you
really.
So,
let's
get
started
with
Jews.
G
H
G
He
will
share
runs,
will
switch
awesome,
all
right,
stow,
welcome
to
the
arena,
update
talk
or
the
pre
update
talk
before
the
actual
update
talked.
It
will
be
on
the
next
summit,
so
my
name
is
Jules
I'm
working
on
Irina,
together
with
my
colleague,
Mario
and
we'll
first
start
with
a
recap
just
to
let
you
know.
What
do
you
mean
e
is
what
it
does
what
we
did
before
the
last
arena
update.
H
So
I'm
gonna
keep
this
free,
quick,
basically,
I'm
gonna
tell
you
what
I
really
look
like
looked
like
up
till
the
previous
summit
in
Philadelphia
and
if
you
happen
to
be
completely
oblivious
to
what
everything
isn't
more
like
what
it
does,
it's
basically
bringing
the
CF
Bush
experience
to
Cadiz,
and
how
do
we
do
this
well?
This
is
how
CF
kind
of
looks
like
today.
You
have
Diego
on
the
right,
the
big
goof
bit,
and
what
we
want
to
do.
H
What
every
knee
is,
basically
this
just
but
kubernetes
instead
of
Diego
and
have
IV
knee,
be
a
small
little
mapping
layer
between
tile
controller
and
kubernetes.
So
what
did
we
actually
do
up
to
that
point?
Well,
if
you
do
a
CF
Bush,
then
IV
knee
will
create
a
stateful
set
in
kubernetes,
which
is
gonna,
represent
your
app
and
then
communities
will
actually
schedule
a
instance
of
of
that
app.
H
If
you
see
F
scale
the
same
thing
we
just
changed,
replicas
of
that
stateful
set
and
communities
will
bring
up
all
the
instances
that
you
want,
and
if
one
of
that
those
instances
happens,
do
you
crash,
then
it
open
it
back
up.
So
how
do
we
actually
do
the
transition
from
droplets
to
images?
Well,
the
truth
is
that
we
actually
have
everything
whether
to
drop.
That's
exactly
the
same,
so
staging
is
the
same
and
the
differences
that
we
have
bit
service
serving
as
a
registry
which
uses
the
droplets
to
actually
serve
images
to
Granny's.
H
So
how
do
we
do
this?
Well,
you
do
the
cs.
Bush
I
mean
guess
the
staging
request
and
what
it
does
is
use
that
staging
for
us
to
create
a
job
in
communities
which
will
create
a
pod
and
that
board
is
actually
running
the
staging
process,
so
it
will
produce
droplets
and
when
it's
done
it
will
put
it
back
in
control
or
so,
but
service
can
have
access
to
it,
but
how
this
actually
with
service
use
the
droplets
you
play
the
image.
H
Well,
we
just
use
the
droplet
as
a
the
topmost
layer
of
the
image,
and
the
rest
of
the
image
is
what
we
call
tearing
FS,
which
is
a
small
little
wrapper
of
CF
Linux
FS
3
this
time
around.
It
was
to
back
at
Summit,
but
we
upgrade
it,
and
this
will
also
contain
a
little
binary
which
will
actually
show
later,
which
will
actually
run
the
droplets
the
same
way
that
Diego
does
it.
So
it's
the
same,
absolutely
same
code
and
how
does
communities
actually
reference
this
image?
H
Well,
it
gets
URL
and
it's
running
two-bit
serviced,
so
it's
it
will
actually
treated
there's
any
other
duck
registry
out
there
cool.
So
next
we,
the
logs
we
used,
flew
in
D,
which
was
data
collection
to
we
deploy
it
on
every
single
node
and
in
the
kubernetes
cluster.
It's
actually
look
at
the
container
logs
from
the
runtime
and
send
them
over
to
logger
Gators.
H
Well,
do
you
register
that
route?
So
it's
to
go
rather
doesn't
route
traffic
to
apps
that
are
that
can't
handle
it?
We
did
events
I
mean
crash
events,
the
in
a
very
similar
way
again,
if
an
app
crashes,
it
will
generate
an
event
and
using
that
event
we
create
and
a
crash
event
in
cloud
controller.
We
do
this
because,
when
you
do
CF
app
Dora,
for
example,
it
lists
instances-
and
this
will
tell
you
which
instances
actually
crashed.
H
Lastly,
we
did
metrics
and
we
used
to
use
heap
store
service
for
this,
but
that
got
deprecated.
So
now
use
metrics
server.
It
works.
Similarly,
you
just
get
the
metrics
from
that
server
and
emit
them
to
Doppler.
So
when
you
do
CF
African,
you
can
see
how
much
your
application
is
consuming
and
I
think
that
was
pretty
fast.
So,
let's
move
on
to
new
things.
G
Alright,
so
a
big
topic
and
what
we
focused
on
was
security
in
the
last
month
since
the
less
updates.
So
let's
take
a
look.
What
we
did
there
first
thing
was
connections
so
basically
having
secure
connections
to
all
the
components
you
really
communicates
to
you.
So
this
is
basically
the
cloud
controller
and
the
bit
service
the
registry
part
in
that
case,
so
we
had
a
TLS
from
the
cloud
controller,
araignee
and
also
the
other
way
around,
and
also
we
had
a
TLS
to
the
bit
service
registry
and
we
also
added
basic
authentication.
G
So
you
need
a
user
and
a
password
to
actually
pull
images
from
the
bit
service
registry
part,
but
we
also
focused
on
the
CF
application
side
so
previously
we're
able
to
reach
other
applications
within
the
arena,
namespace
directly
from
your
application
that
you
pushed
and
also
you
were
able
to
reach
the
kubernetes
master
and
API
and
everything-
and
we
just
don't
allow
that
anymore.
So
you
cannot
reach
any
other
application
nor
the
kubernetes
master
or
the
API
or
anything
anymore.
G
G
Staging
we
had
one
staging
job
and
which
was
performing
the
whole
staging
process
which
is
download,
builds
so
download
the
app
it's
built,
a
droplet
and
upload
the
droplet
back
to
the
blobstore,
and
that
happened
all
in
one
pot
and
a
problem
was
that
this
pot
had
all
the
secrets,
and
so
what
we
did
to
improve
that
we
split
up
the
the
staging
in
in
three
different
steps.
Each
step
is
happening
in
a
separate
container,
so
first
container
downloads
the
app
it's
from
the
blobstore.
G
G
Next
part
is
the
root
of
s
patching
now
you're
able
to
dynamically
patch
the
root
of
s,
let's
say
you're
the
operator
of
your
Cloud
Foundry
and
Irina
kubernetes
cluster,
and
you
want
to
use
a
specific
with
the
best
version
and
what
you
would
do
as
an
operator.
You
would
update
your
home
chart
with
the
for
the
best
version
of
your
choice
and
you
would
perform
the
helm,
upgrades
and
helm
would
up
would
update
the
rooted,
best
version
label
on
rainy
and
the
bit
service,
which
will
cause
a
restart
of
these
components.
G
And
now
all
applications
that
are
newly
pushed
to
kubernetes
will
have
to
knew
where
the
desperation,
but
there,
of
course,
also
the
applications
that
are
already
scheduled
by
Rini
and
there's.
The
third
component
that
comes
in
here.
That's
the
route
as
patcher,
so
it
waits
till
on
upgrade
till
till
the
bit
service
is
back
up
again
and
then
it
updates
all
the
applications
with
the
right
word,
the
best
version
and
make
sure
to
restart
them,
so
that
all
applications
pull
the
right
image
from
the
bit
service.
G
H
So
here
we
have
Cloud
Foundry
deployed
with
the
CF
and
emini
on
the
left
and
on
the
right.
We
actually
have
some
applications
already
pushed
and
what
I'm
gonna
do
now
is
show
you
the
route
fast,
that's
actually
being
deployed
when
I
get
all
the
zoom
things
out
of
the
way.
So
if
we
edit
this
spot
or
I
mean
we
can
see
that
in
in
the
labels,
we
have
the
first
version,
which
is
version
80
1.00,
and
just
so.
We
can
compare
the
actual
route
fast
when
we
swap
it
I'm
gonna
say
to
that
container.
H
I'm
gonna
do
sha-256
some
on
this
file
and
we
can
see
that
it's
one
CC
159
something
something
something.
This
is
actually
the
launcher
which
launches
the
droplet.
So
let's
say
we're
an
operator
and
we
decide
that
version.
81
is
insecure.
Now
it
has
some
phone,
but
if
T-
and
we
want
to
actually
change
that
version.
So
how
do
we
do
that?
Well,
we're
gonna
open
up
the
values
file
that
we
use
for
helm
for
deploying
irony
and
I've
put
here.
The
root
first
version
up
top,
so
this
81
is
now
insecure.
H
Let's
change
it
with
84,
which
we
know
is
a
good
version.
So
if
I
helm
upgrades
my
deployment,
it
will
actually
do
the
things
that
you
said
if
it
works
today
it
does
so
it's
gonna
restart
that
service.
We
can
see
it's
restarting
it's
restarting
ireenie
and
you
can
also
notice
this
root
fast,
but
your
job
over
here
virtually
for
which
we
started
our
containers
on
the
right,
so
they
already
started.
H
H
Now
there
is
a
small
bug
that
we
actually
caught
two
hours
ago
before
before
this
kepco,
and
you
actually
need
to
get
you
actually
need
to
manually
restart
the
applications
to
get
director.
Rufus
bonus
points.
If
you
actually
spot
the
bug,
we
all
go
into
it,
but
yeah
we're
working
on
fixing
that
so
back
to
the
presentation.
G
Okay,
next
part
is
Container
security.
What
we
wanted
to
do
is
to
have
the
same
security
container
security
configuration
as
garden
has
today
in
Diego,
and
so
that's
we
had
a
table
with
parameters
that
garden
already
compared
to
kubernetes
and
based
on
that,
we
had
to
applied
for
a
meet
you,
and
we
had
like
three
parameters
we
had
to
focused
on
so
the
first
one
was
the
second
filtering
enable
it,
which
is
basically
a
white
listing
for
syscalls.
G
Then
the
next
part
I
want
to
talk
about
is
the
resource
limits.
We
just
mentioned
the
disk
quotas.
So
far
we
hard-coded
to
disk
quota
to
2
gigabytes,
but
we
just
today
started
the
story
to
make
that
configurable
we
are
CFC,
Li
and
CPU,
and
memory
is
already
configured
by
the
CFC
Li.
So
you
can
set
it.
However,
you
want
previously.
We
had
kubernetes
default,
which
is
limitless
so
next
part
scalability.
We
did
some
scalability
testing
and
actually
we
had
free
deployment
scenarios.
G
Two
of
them
were
performed
by
us,
the
arena
team
and
the
third
one
was
performed
by
a
guy
from
sa
P
and
he
actually
did
the
biggest
environments
with
250
nodes
and
40,000
app
that
he
perhaps
that
he
pushed
on
SCF
based
deployment
with
vallini
on
gke,
and
we
had
one
scenario
where
the
both
deployed
CF.
That
was
pointing
to
an
external
kubernetes
cluster
on
gke,
and
we
also
did
some
performance
testing
with
a
SEF
based
deployment
with
Iranian
Ike.
G
Yes,
for
the
boss
deployed
scenario,
we
had
thousand
apps
that
we
pushed
on
21
nodes
and
for
the
IKS
scenario
we
had
200
apps
that
we
pushed
on
10
nodes.
So,
of
course
we
had
some
issues.
We
expected
that
and
some
issues
we
fixed
right
away
was
routing.
We
had
problems
that
the
routing
was
scaling
and
many
routes
were
dropped.
You
were
not
able
to
reach
your
apps
anymore.
Also,
metrics
did
not
show
up
and
the
lock
related
kept
crashing
so
that
no
locks
showed
up.
G
But
this
three
problems
were
fixed
and
the
remainings
in
the
remaining
issues.
You
basically
have
how
upgrades
or
up
grades
on
community
communities
itself,
if
you
update
master
or
worker
notes,
so
you
have
some
larger
downtime
on
your
applications
and
also
the
routing
for
some
apps
might
be
gone.
So
we
have
to
investigate
there
a
little
bit
more
second
one
which
was
actually
fixed
this
week,
I
think
yesterday
we
had
a
short
period
of
500
two's
on
upgrades.
G
So
when
you
try
to
reach
up,
you
saw
a
502
for
a
few
seconds
and
then
you
were
able
to
reach
up
again,
and
that
was
because
of
the
asynchronous
update
that,
with
the
that
do
we
need
us
with
the
routes.
Then
I'm,
not
arena,
related
issue.
We
had,
but
it's
a
community's
issue,
that
you
cannot
push
more
than
14-thousand
stateful
sets
to
a
Goodwin
8's
cluster,
so
kubernetes
prevents
you
from
doing
that.
And
finally,
we
discovered
this
that
bit
service
was
the
weakest
part
in
performance
testing.
So
it
has
this
high
load.
G
H
So
ireenie
recently
had
its
first
inception
and
one
of
the
things
that
we
decided
we
wanted
to
move
forward
to
is
migration,
so
getting
people
migrated
from
Diego
to
I
mean
II,
and
this
is
something
that
we
haven't
done
yet.
We
like
Eric,
said
previously
that
we
have
been
in
talks
with
Cappy
to
make
this
a
smooth
transition,
and
we
have
several
ideas
on
how
this
could
happen,
for
example
using
a
flag
in
in
the
CF
CLI
to
tell
traffic
controller.
H
This
app
is
for
I
mean
this
app
is
for
Diego
or
using
isolation
segments
for
ebony.
But
again
this
is
something
that
we
are
towards
and
something
that
you
would
have
to
look
forward
to
another
thing
that
we're
working
on
is
making
Bosch
and
I.
We
work
together
more
nicely.
Now
there
is
a
community
maintain
bas-reliefs
for
every
knee.
We
rather
better
make
a
really
kind
of
independent
from
how
you
deploy
CF
so
just
plug
in
into
your
kubernetes
cluster,
and
it
wouldn't
matter
if
it's
abortion,
Floyd
or
containerized
CF.
H
It
would
just
work,
but
this
is
something
we're
still
spiking
and
working
towards
and
that's
pretty
much
it
I
just
wanted
to
mention
that
we
have
regular
releases
on
our
Irene
release,
github
repo
they
they're,
mostly
weekly,
but
not
always
so
that's
a
place
where
you
should
go.
If
you
want
to
actually
try
out
the
Rini
and
there's,
of
course,
instructions
on
how
to
do
that.
H
We
have
also
office
hours
happening
every
first
Monday
of
the
month
at
10:30
a.m.
Eastern,
Time
I'll
post
a
link
to
this
blog
post.
So
you
can
get
the
zoom
link
or
you
can
just
go
over
to
I
mean
def
in
the
cloud
control.
Cloud
Foundry
stack,
so
you
can
get
that
link
from
there.
So
if
you
have
any
questions,
you
want
to
tell
us
how
much
you
like
a
weenie
or
hate
it
or
you
just
want
to
chat
with
the
team,
but
we
have
to
have
you
and
that's
the
of
course.
H
A
C
E
H
A
Thank
You
Smith
all
right
any
anything
else,
because
we
have
a.
We
have
a
good
agenda.
Other
things
to
get
to,
but
I
want
to
thank
you
again,
Mario
and
Jules
for
a
great
update,
I,
really
like
the
first
few
slides,
where
you
kind
of
showed
how
it
actually
works,
but
with
pictures
so
that's
kind
of
nice
for
everybody
and
and
good
update.
So
I
guess
you
guys,
as
you
mentioned,
you
will
have
an
even
deeper
dive
at
summit.
A
So
if
you
want
one
more
reason
to
go
to
summit
this
is
it
and
in
that
way
you
can
actually
spend
some
time
and
talk
to
them
and
see
them
in
person
so
more
opportunities.
Ok,
so
last
call
for
questions,
okay,
excellent!
So
thank
you
guys.
Let
me
pass
it
to
Morgan.
He
actually
joined
a
little
bit
late,
but
he
pasted
some
updates
on
what's
going
on
in
Bosh.
Let's
give
him
maybe
two
or
three
minutes
to
to
give
those
highlights
and
maybe
field
some
questions.
I
Sure,
sorry,
good
morning,
I
think
the
most
interesting
thing
is:
we've
cut
a
new
director
so
to
70.5
and
CEO
I
version
six,
one
of
the
big
things
which
we
heard
a
lot
is
folks,
have
a
deployment,
and
then
they
try
to
upgrade
that
deployment
and
something
fails
and
then
they
want
to
do
recreate
or
start
or
stop.
And
then
their
entire
deployment
gets
rolled
back
and
then
they're
like
hey
bashed.
I
What
the
hell
would
she
just
undo
all
the
stuff
I
did
so
as
of
those
versions,
there's
a
new
flag
on
the
CLI,
which
is
no
conversion,
and
if
you
add
that
it
will
actually
only
touch
the
thing
you're
trying
to
touch
and
not
roll
back
to
the
entire
deployment.
So
we
really
want
to
get
a
lot
of
feedback
on
that
and
see
where
it
breaks
and
where
it
doesn't
break.
I
Thank
you,
yeah.
It's
been
a
long
requested
one.
The
other
interesting
one
is
a
lot
of
development
teams.
Will
upload
bosch,
DNS
or
something
in
a
runtime
config,
and
then
their
pipelines
will
run
clean
up
all
and
then
the
next
time
they
run
they
will
break
because
we
cleaned
up
releases
that
were
specified
in
a
runtime
configuration
at
appointment.
I
So
as
of
that
release
releases
that
are
included
in
a
runtime
config
but
might
not
be
included
in
a
deployment
will
be
marked
as
in
use,
so
you
won't
actually
delete
them
when
you're
cleaning
up
all
your
old
blobs
and
then
the
last
kind
of
exciting
one
that
folks
have
been
asking
for
is
being
able
to
configure
the
length
of
your
certificate
expiration
when
you're
setting
Bosch
create
ends.
So
in
the
directors
manifest.
I
A
I
A
So
if
you
have
any
question
for
Morgan,
this
is
your
chance.
Otherwise
we're
gonna
pass
it
glad.
Let
me
see
if
there's
anything
on
the
chat.
Okay,
nothing
on
the
chat.
Thank
you
again,
Morgan
all
right.
So
the
next
and
the
last
call
the
last
talk
I
guess,
but
on
the
beast,
is
vlad
yuna,
I'm
gonna
mispronounce
your
last
name.
So
sorry
about
that.
But
Vlad
is
from
Essie.
A
C
We're
just
going
to
quickly
talk
about
irony,
X,
I'm,
sorry,
I,
don't
have
a
demo
at
or
who
works
much
more
on
irony
X
than
I.
Do
you
had
a
laptop
emergency?
Something
died
with
the
thing,
and
now
you
need
to
fix
it.
So
I'm
just
going
to
quickly
tell
you
what
I
really
X
is
and
then
point
you
in
the
right
direction.
C
C
C
C
So,
while
trying
to
build
these
extensions,
we
realized
that
there's
a
lot
of
common
functionality
that
we
could
put
inside
of
a
library
a
go
package.
So
this
is
what
irony
exits.
It's
a
bunch
of
code
that
helps
you
write
extensions
specifically
for
irony
by
by
writing
a
lot
of
the
code
that
you'd
have
to
repeat
so
the
main
mechanism
we
use
to
extend
what
I
really
does
are
kubernetes
web
hooks.
C
They
use
these
to
mutate,
the
pods
that
I
really
create
creates,
and
we
also
have
Watchers.
So
you
get
notified
when
an
application
is
being
created
or
deleted.
So
this
common
functionality
is
very
useful
when
you're
trying
to
extend
right
so,
for
example,
the
admission
web
hooks,
we
use
them
for
Percy
support.
So
what
happens
as
mario
and
jules
describe?
Cloud
Foundry
the
controller
and
all
of
the
internals
will
tell
Irena
to
create
an
application.
I
really
will
go
to
kubernetes
it'll
create
a
pixel
set,
and
then
a
pod
will
be
created
in
kubernetes.
C
What
happens
with
the
extension
is
that
we
tell
kubernetes
that
we
want
to
register
a
web
hook
so
that
whenever
a
pod
in
a
specific
name,
space
is
created,
we
want
to
be
notified,
so
well,
not
just
notified.
We
actually
want
to
mutate
that
definition.
So
once
the
pod
wants
to
be
created,
but
before
any
containers
are
created,
kubernetes
will
call
the
extension
first
and
we
have
a
chance
to
mutate
the
pause
before
any
container
gets
created.
So
in
the
case
of
Percy,
the
mutation
is
that
we
add
volume
mounts
to
the
pot.
C
So
we
have
this
I,
really
the
broker
that
actually
worked
with
kubernetes
storage
classes,
so
ucf
create
a
service
which
results
in
a
BBC
being
created
for
you,
you
bind
that
to
the
application
and
then
this
extension
will
mutate.
The
application,
pods
and
add
volume
mounts
so
this
is
this
is
how
it
worked,
and
the
library
kind
of
does
everything
for
you.
It
sets
up
an
HTTP
server
generates
certificates.
It
creates
those
certificates
as
secrets
in
kubernetes.
It
creates
the
web.
The
mutation
web
hook
configuration
that
kubernetes
needs.
All
of
that.
C
C
The
library
can
be
found
that,
to
the
slash
I
really
X
on
github,
we
have
three
extensions
so
far,
Percy
and
logging.
They
are
already
done
so
you
could
actually
cut
them
out.
Percy,
like
I
said,
is
for
supporting
with
persistent
storage
in
your
apps.
It
worked
hand
in
hand
with
I
really
Percy
broker.
Then
we
have
an
extension
for
logging.
This
replaces
the
fluency
solution.
That's
in
irony
core.
C
What
we
do
here
is
we
watch
for
the
application
pods
that
get
created
by
a
greeny,
and
then
we
use
the
coordinating
API
that
stream
logs
and
then
forward
them
directly
to
to
the
logger
Gator.
So
we
kind
of
act
like
the
Alegre
Gator
agent
that
specifically
understands
how
to
read
ordinary
spots
from
from
coordinates
and
then
finally
we're
working
on
at
the
page.
This
is
in
progress
right
now,
so
you
know
if
you
want
to
help
out
we'd
love
that
and
two
more
things.
C
C
A
A
Thank
You
Vlad
and
pronounce
your
last
name
you've
an
of
exclusive.
That's
perfect!
Apologies
again!
Alright!
So,
let's
see,
if
you
haven't,
you
have
any
question.
I
have
one
question
for
you,
but
let's
see,
if
other
people
have
questions
for
you
any
questions
for
glad
so
Simon
murdered
you,
you
can't
speak
or
do
you
want
me
to
say
for
you
didn't
you
hear
me
yeah,
we
can
hear
you
Simon
go
for
it.
I.
E
I
was
trying
to
speak
earlier
to
by
the
sounding:
what's
broken,
okay,
so
the
question
for
flood
that
I'm,
having
is
so.
This
is
great
I,
mean
certainly
super
helps
to
speed
up.
You
know
all
the
features
into
the
Irene
that
we
need,
but
do
you
guys
have
any
kind
of
plans
to
eventually
move
that
back
to
mainland
ireenie
at
one
point
in
time,
or
is
this
going
to
become
a
permanent
solution?
As
you
see
fit,.
C
C
First
and
then
you
know
after
they
get
some
mileage,
we
can
move
pieces
over
so
I
think
the
long-term
vision
is
that
yes,
I
at
least
I
really
X
would
become
part
of
core
and
then
the
all
of
these
side
projects
can
stay
as
side
projects,
for
example,
Percy
and
SSH.
Perhaps
logging
can
actually
replace
fluency
in
core.
F
Other
thing
I'd
add
to
that
I
think
that's
a
perfect
answer.
One
thing:
one
of
the
reasons
it
weenie
X
I
think
is
really
really
good
and
important
is.
It
means
all
the
different
teams,
potentially
in
compounds.
You
can
ship
their
things
in
a
way
that
actually,
even
though
it's
called
arena
X
nothing
really
to
do
with
the
really
way
it's
like
it
will
work
with
any
application.
F
However,
you
push
it
whether
it
was
a
weenie
or
not.
So
if
you
use
like
the
percy
broker
that
they
have,
you
now
have
the
percy
functionality.
Whether
or
not
you
use
your
weenie
so
now,
Percy
can
be
in
front
of
deep
cut,
hold
useful
component,
so
I
think
I
think
what
we're
hoping
is
all
these
lay
at
the
moment.
We've
got
these
projects
to
fill
the
gaps,
I'm,
hoping
that
all
the
various
kind
of
foundry
teams
over
time
will
start
to
ship
kind
of
on
top
of
something
like
in
really
X
and
probably
iridium.
E
F
E
F
So
yeah,
this
is
exactly
what
I
mean
at
the
moment
before
it
really
SSA
just
a
separate
project,
while
I'm
hoping
will
happen
and
we've
started
some
discussions
about
this
is
that
Diego
will
ship
an
arene
e,
ssh
right
and
so
will
be.
The
core
teams
will
expose
the
core
features
in
this
way
and
then,
hopefully,
what
will
happen
is
we'll
have
basically
a
CF
deployment
like
thing
which
will
give
you
the
tested
set
of
extensions
that
all
work
together
in
one
package.
F
C
A
I
I
think
I
think
this
is.
This
is
interesting,
I
think
it's
big
dreams,
but
there's
only
a
an
aspect
of
it
that
could
probably
move
fast.
These
are
extensions
to
the
run
time.
So
maybe
you
know
people
like
you
know
Eric
and
Yui,
or
the
right
person
to
talk
to,
but
some
of
those
that
could
be
extensions
outside
of
the
run
time.
I'd
be
happy
to
consider
them
and
PMC
extensions,
especially
if
they
can
be
self-contained
and
we
can
maybe
move
them
to
incubator
and
so
on.
A
He
might
have
dropped
to
so
I.
Don't
know,
let's
see
yeah,
he
might
have
dropped
all
right,
but
I
guess
summit
is
gonna,
be
a
good
place
to
figure
out
how
to
move
forward
on
this
any
other
questions.
I
have
I.
Have
a
one
question:
that's
more!
That's
a
little
bit
different
than
the
discussion
having
here,
but
any
question
or
comment
on
this
particular
discussion
about
extensions
and
how
to
organize
them
and
move
forward.
A
No
okay
cool.
So
my
question,
I
guess,
Vlad
and
Jules
is
with
this
extension
mechanism.
Is
there
any
part
of
you
Eenie?
That
would
not
be
extension
because
it
seemed
very
generic
where
you
could
essentially
enter
like
you
could
put
your
extension
in
between
kubernetes
and
uranium
jam
things.
So
what
are
the
kind
of
things
that
you
would
not
want
to
have
extensions
for
I?
Think.
F
F
F
F
We
had
to
kind
of
all
these
different
things
that
are
developed
independently
and
like
garden.
For
example,
we
don't
run
camps
in
our
CI
at
all.
We
just
have
tests
for
garden,
but
then
every
time
you
push
your
new
garden
release,
it
gets
picked
up
by
Relan
makes
its
way
into
new
CF
employment
as
being
its
tested
together.
So
yeah
I
think
there's
lots
to
talk
about
there
and
I
want
one
thing:
that's
one
thing
we're
trying
not
to
do
in
arena.
It's
kind
of
owed
all
of
transparency,
but
on
kubernetes
cuz
I.
F
We
can't
buy
it.
It's
too
big.
We
can't
do
that,
and
so
we
want
to
just
own
like
a
core
thing
kind
of
like
bein
Alec.
The
analogy
is
DFO
a
gaggle
doesn't
test
all
the
different
releases,
just
as
the
AVO
and
then
r
ellen
has
small
together
so
I'm,
hoping
that
melons
we'll
be
able
to
step
up
and
take
on
some
of
that
kind
of
cross
component
testing.