►
From YouTube: Keptn Developer Meeting - June 15, 2022
Description
Meeting notes: https://docs.google.com/document/d/1y7a6uaN8fwFJ7IRnvtxSfgz-OGFq6u7bKN6F7NDxKPg/edit
Learn more: https://keptn.sh
Get started with tutorials: https://tutorials.keptn.sh
Join us in Slack: https://slack.keptn.sh
Star us on Github: https://github.com/keptn/keptn
Follow us on Twitter: https://twitter.com/keptnProject
Sign up to our newsletter: https://bit.ly/KeptnNews
A
A
A
A
A
Then
we
work
a
lot
on
improving
our
testing
pipeline,
specifically
changing
the
integration
test
and
moved
them
into
component
tests.
This
had
a
big
boost
in
terms
of
run
time
for
the
pipelines
now,
instead
of
requiring
20
minutes
to
run
this
test
or
10
minutes
for
running
this
test.
We
are
down
to
two
minutes
and
also
here
we
plan
to
do
the
same
things
for
the
shipping
controller,
but
for
other
parts
of
the
for
other
tests.
Basically,
is
there
any
question
for
this
frame.
A
A
A
B
B
So
this
sprint
I
worked
of
a
few
fixes
and
to
our
integration
test
and
some
cleanup
on
all
deprecated
things.
So,
let's
see.
B
B
Which
was
functioning
as
a
proxy,
but
we
are
ready
for
other
tests
used
mock
server.
So
what
I've
done
here
was
just
adding
a
few
line
of
code
to
make
sure
that
we
can
use
the
mox
servers
as
the
proxy
in
this
test
and
yeah
verify
that
everything
works.
Fine,
nothing
much
here
for
the
cleanup
pr!
This
was
about
removing
the
old
subscription
from
the
uniform.
B
Yeah
here
in
the
subscription
we
used
to
have
subscriptions
and
subscription
where
there
was
a
list
of
of
the
subscription
in
the
model.
Maybe
I
can
find
the
model
here
somewhere.
B
And
now
this
has
been
removed
by
all
parts,
and
I
can
show
you
in
the
pr
here.
This
has
been
changed
in
the
core
utils
library.
Just
eliminating
this
part
and
then
what
has
to
be
done
was
to
in
shipyard
make
sure
to
remove
all
the
filters
where
we
were
working
with
that
in
the
handlers
and
in
the
database.
B
Nothing
complicated,
just
removing
something
which
was
deprecated
since
three
or
four
version.
B
Then
there
was
a
tiny
little
bug
in
the
scheduled
pipeline.
Doesn't
really
even
worth
speaking
about
that.
The
point
is
that
only
when
the
pipeline
is
scheduled,
this
branch
will
be
checked
and
the
curly
brac
were
missing.
So
this
part
was
empty
and
scheduled
pipeline
was
failing
to
bet
as
the
final
work
in
the
new
library
we
have
the
cp
connector.
B
We
used
to
have
nets
connector,
which
was
initialized
by
the
user
and
passed
to
the
library,
and
this
was
a
problem
because
then
we
would
have
a
time
in
which
the
nats
is
receiving
events,
but
the
library
is
not
actively
using
it
yet
because
it's
not
started
so
to
avoid
that
to
just
changed
this
in
a
way
that
when
we
create
new
nuts,
we
just
store
the
data
relative
to
it,
and
then
every
time
we
use
the
nuts
connection,
we
will
call
a
function
where
we
only
once
connect
to
nuts.
B
C
C
Due
to
some
reason,
I
cannot
share
my
screen.
Maybe
we
can.
C
Great
yes,
so
last
print,
I
worked
mostly
on
the
two
issues.
We
had
some
leftovers
from
the
zero
downtime
testing,
which
was
mostly
about
fixing
the
bugs
we
discovered
polishing
the
tests
and
yeah
really
testing
and
that
we
have
zero
downtime.
C
C
C
Was
not
able
to
get
the
subscriptions
from
the
older
and
was
waiting,
so
this
was
fixed
by
this
pull
request,
where
we
actually
immediately
after
the
start
of
the
subscription
source.
We're
trying.
C
Yes,
yes,
immediately
after
starting
the
subscription
source,
which
kind
of
thing.
A
The
problem
was
that
we
were
catching
the
subscription
only
after
this
timer
was
started,
so
we
didn't
fetch
immediately
subscription
as
soon
as
the
integration
started,
but
only
after
a
specific
amount
of
time.
Instead,
now
the
first
thing
we
do
when
the
subscription
starts,
then
we
also
fetch
the
information,
and
then
we
start
the
timer.
A
D
Okay
cross
your
fingers,
I'm
gonna
share
my
screen.
E
D
Are
gonna
briefly
discuss
is
also
in
the
part
of
the
cp
connector
library,
where
I
just
enhanced
the
synchronization
during
shutdown
of
the
thing.
Basically,
what
me,
what
that
means
is
the
library
is,
in
the
background,
kicking
off
some
core
routines,
which
we
want
to
shut
down,
gracefully
like
we
do
for
other
parts
or
other
components
in
captain
as
well,
and
also
here.
I
enhanced
it
a
little
bit
with
the
usual
things
you
do,
namely
with
a
weight
group
on
the
outside.
D
So
each
components
that's
gonna
be
started,
for
example,
the
event
source
the
subscriptions.
So
as
we
receive
that
weight
group-
and
I
think
you're
gonna
see
here-
there
will
be
a
weight
group
of
size
two,
because
we
have
two
components
to
be
synchronized
and
later
on,
when
the
things
are
going
to
shut
down.
D
We
synchronize
on
this
weight
group
to
be
finished,
and
only
after
that
we
set
the
state
to
registered
force,
because
only
that
we
know
that
each
component
was
able
to
gracefully
shut
down,
for
example,
disconnect
from
a
message
broker
disconnects
from
subscription
api
and
so
on.
So
that
was
the
first
thing,
which
should
even
hardening
the
shutdown
a
little
bit
more.
D
Although
we
didn't
saw
issues
there,
it
was
not
completely
correct
and
now
it's
synchronized
and
just
you
can
now
trust
on
this
boolean
value,
which
says
it's
registered
or
not
because
it's
properly
synchronized
it
was
the
first
one.
Next
thing
always
the
cfp
connector.
Now
we
are
able
to
inject
a
log
implementation.
D
Kind
of
yeah
you
see
it
here
default
logo,
which
is
just
yeah
fulfilling
the
interface
of
all
our
loggers
and
just
putting
the
stuff
on
standard
out
and
nothing
more.
Now,
you
are
able
to
use
the
with
logger
function
for
each
component
to
inject
a
proper
logger,
for
example,
logras
or
what
else
so
that
we
can
actually
use,
or
the
user
can
actually
inject
his
favorite
logo
and
use
all
the
lock
level
features
and
so
on
with
this
library.
D
D
D
You
can
see
it
here:
control,
plane,
package,
event,
matcher
event,
source
a
fake
package
lock
for
water,
so
basically
for
each
com,
each
aspect
of
the
library
we
will
have
now
our
nice
little
package,
which
makes
it
yeah
way
more
clean
on
my
mind,
and
we
don't
need
this
strange.
D
That's
it
for
this.
I
guess
nothing
too
complicated
there
and
one
little
thing
I
just
explored
where
we
are
using
swagger
documentation.
This
wac
command
we
are
using
for
it
offers
a
neat
little
thing,
which
is
called
smack.
Swag
swag
format,
which
I
just
executed
for
the
whole
captain
code
base
and
what
it's
basically
doing
is
just
reformatting
the
swagger
commands
to
have
yeah
to
be
aligned,
and
so
on.
So
just
a
formatting
thing.
D
I
think
we
can
put
that
in
a
pipeline
or
I
don't
know
what
useful
useful
thing
I
think
other
than
that.
One
thing
I
wanted
to
mention
about
anna's
ticket
about
the
connection.
Thingy,
don't
worry,
I
think
it
was
not
a
problem
that
you
will
lose
events
because
of
the
connection
already
established.
It
was
because
the
subscription
and
ultimately
the
the
subscription
and
net
on
the
message
broker.
D
So
the
problem
here
was
more
like
you
need
a
already
established
connection
to
create
the
control
plane
now
with
honors
changes,
it's
way
more
easier,
because
you
can
just
configure
your
control
plane
component
this
one
here
and
don't
need
at
this
moment
already
connection
to
knots.
This
one
will
just
later
be
established
when,
when
the
connection
is
when
the
control
band
is
started,
so
with
the
current
release
of
captain,
we
don't
lose
events
of,
because
of
that,
it's
just
nicer
to
use.
D
Good,
that's
it
from
my
side.
I
think
I
don't
know
andre
it's
working
then
I'll
hand
over
to
you
and
stop
my
sharing.
If
there's
no
question.
C
Yeah
already
the
screen
sharing
works
as
well.
Yes,
so
I
think
giovanni
already
mentioned
the
zero
downtime
testing
and
issues
binded
with
it.
So
I
will
move
to
the
new
git
credentials
model,
which
was
the
main
issue
for
me
in
the
last
sprint.
C
Recently,
we
decided
that
we
will
make
it
breaking
change
and
refactor
the
api
model
for
storing
the
git
credentials.
The
previous
model
for
the
project
looked
like
this.
It
had
a
flat
ear
here
and
actually
was
not
that
nice
and
not
that
easy
to
use
and
the
test.
So
we
decided
to
move
to
a
more
sub,
more
structured
hierarchy
with
where
the
type
of
the
authentication
has
its
own
sub
structure
and
yeah.
C
It's
kind
of
easier
nicer
to
use
more
effective
as
part
of
this
pull
request
is
part
of
this
issue
and
also
pull
requests.
C
We
decided
to
kind
of
normalize
the
used
models,
because
actually
we
were
using
for
each
service
a
separate
model
which
was
not
that
kind
of
convenient
to
use,
because
if
we
needed
to
make
the
change,
we
need
to
make
the
changes
in
all
of
these
models.
So
now
we
have
just
one
model
for
all
services
which
are
used
widely
and
therefore
it's
pretty
easy.
C
For
now.
This
change
is
not
yet
merged
because
there
still
need
to
be
some
adaptations
in
the
bridge
side
to
actually
make
it
usable,
but
the
changes
are
prepared
for
the
model
is
placed
in
go
utils,
where
we
adapted
the
models
for
creating
the
project
for
retrieving
the
project
and
also
model
for
storing
the
project,
and
also
the
adaptations
for
the
captain
core
are
also
done.
C
There
were
a
lot
of
places
which
needed
to
be
adapted
as
well
as
there
was
a
need
to
implement
a
migrator
which
will
migrate
the
database
and
also
the
secret
data
from
the
older
model
to
the
newer
model.
So
this
all
is
implemented
and
tested,
and
hopefully,
in
the
next
sprint
it
will
be
merged
and
part
of
the
captain
core.
A
Okay,
I'm
also
having
issues
with
screen
sharing
seems
like.
Oh
no,
it
seems
to
work
yeah.
I
plan
to
present
a
rather
small
improvement
for
the
bridge.
It's
about
showing
a
loading
indicator
for
sequences
before
yeah
the
filters
are
applied
and
the
result
set
is
displayed.
A
So
we
had
there
the
issue
that,
when
you
switched
in
the
bridge,
this
is
an
old
image
with
the
issue
between
tabs
every
time
when
you
switch
to
the
sequence
screen,
you
see
for
a
short
while
that
there
are
no
sequences
found,
yeah
and
yeah.
This
is
now
fixed
so
that
when
you
jump
to
the
sequence
screen,
you
see
a
loading
indicator
first
and
then,
of
course,
as
soon
as
you
have
the
sequences,
you
see
the
sequences
there
yeah.
F
Okay,
so
for
me,
a
lot
of
bug,
fixes
or
yeah
only
bug
fixes.
So,
let's
jump
to
the
first
one.
There
was
the
case
that
not
all
evaluations
were
shown
in
the
environment
screen.
This
was
this
one:
the
relations
on
the
right
side
only
to
a
maximum
of
20
were
shown
here
and
not
all
of
them
and
yeah.
This
was
due
to
a
missing
limit
parameter.
It
was
not
sent
to
the
back
end
and
now
it's
correctly
set
to
to
100
or
to
the
polling
or
to
everything.
F
F
The
next
thing
is
we
had
a
broken
ui
if
the
connection
was
lost,
so
if
in
between
here
or
in
some
screen,
the
call
to
the
project's
endpoint
project
endpoint
did
not
work.
The
whole
ui
was
not
usable
anymore.
F
F
Next
thing
then
yeah
we
had
some
kind
of
flickering
in
the
sequence
screen
for
the
filter.
If
you
selected
something
here
here
and
selected
a
sequence
or
something
every
time
there
was
a
bit
of
a
flickering
in
the
filter,
and
this
is
now
fixed
that
it
only
updates
or
refreshes
when
needed.
F
Next
to
that,
we
had
the
missing
relations
card
in
environment
screen.
Also,
this
was
this
one
here
that
is
quite
missing.
This
was
related
only
here
if
the
latest
event,
in
that
sequence
was
from
the
web
hook,
service
and
yeah.
Now
it
is
fixed
that
is
correctly
fall
back
to
the
right
evaluation
operation
here
with
the
red
square
yeah.
This
was
this
bug,
then
the
next
bug
there
was
also
the
case.
F
If
you
select
a
sequence
or
a
stage
and
refresh
it,
it
did
not
select
the
right
stage,
but
only
for
a
sequence
that
wasn't
initially
loaded
in
the
sequence
screen,
so
everything
that
first
25
worked
fine,
but
if
something
out
of
the
25
had
another
stage
selected,
it
was
not
working
and
just
selected
the
latest
one.
F
So,
for
refresh
now,
the
right
stage
is
selected
instead
of
the
last
one
yeah
yeah.
This
is
also
fixed.
Then
the
next
fix
yeah.
We
there
were
some
cases
where
the
slo
file
was
invalid,
encoded
and
then
the
whole
heat
map
wasn't
shown
anymore
and
yeah.
You
couldn't
see
any
evaluation
if
one
evolution
had
an
emulsion
effect
and
yeah.
This
is
also
fixed.
F
So
if
there
is
an
invalid
slope
right
now,
you
can
see
the
button
is
disabled
and
there's
an
overlay
that
you
have
an
invalid
encoded,
slo
file
yeah
and
if
there
is
no
sro
file,
this
button
is
just
not
here:
it's
not
then
removed.
F
This
was
this
pack,
then
another
bug
we
had
for
the
approver.
So
if
you
had
an
approval
or
something
and
accepted
it
or
disliked
it,
that
was
the
case
that
two
evaluation
finished
were
sent
by
bridge
server
by
the
webbridge,
client
and
yeah.
This
is
now
also
fixed
that
the
bridge
only
sends
one
finished
event
instead
of
two.
F
This
was
this
bug
and
another
case
yeah
the
project
bot.
So
that
was
the
case
that
if
you
go
to
dashboard
the
projects
weren't
updated
correctly
only
if
the
page
is
initially
loaded,
but
not
if
you
go
to
the
dashboard
again,
it
showed
the
old
state
instead
of
the
current
one-
and
this
is
also
fixed
by
now
yeah
and
that's
everything
from
my
site.
Are
there
any
questions.
A
I
think
we
lost
you
meg
today.
This
portal
is
not
the
most
stable
one.
E
Yes,
okay,
currently,
bigger
task,
I'm
working
on
is
the
introduction
of
estate
management
in
the
bridge
angular
project.
Therefore,
I
made
some
pro
preliminary
work,
for
example
renaming
the
dashboard
and
we
want
to.
E
E
E
So
it's
basically
doing
the
same
stuff
for
over
and
over
for
this,
for
these
classes
so
having
a
module
and
then
instead
of
importing
instead
of
importing
the
module
in
the
app
modules
you
just
import
the
module,
and
so
all
these
components
will
be
removed
and
it
will
be
more
modernized
which
is
good
for.
E
Reducing
the
size,
the
browser
has
to
load
so
not
load
the
whole
application
at
once,
but
just
in
smaller
chunks,
then
there
was
a
fix.
As
you
know,
if
you
enable
the
version
checking,
then
there
is
a
version
loaded
from
the
official
captain
page
something
like
this,
and
if
it's
not
retrievable,
then
it
errors
and
so
there's
now
a
default
version
chasing
returned
and
the
ui
works
as
expected.
E
So
does
not
crash
and
another
fix
yeah,
some
small
one
when
you
had
a
filter
in
this
in
this
filter
bar
and
you
would
reload
the
page.
E
All
right
off
was
the
case
that
you
had
the
old,
even
though
you
cleared
the
filter
before
and
the
old
values
were
inside,
and
this
request
fixes
this
and
the
ui
as
always
the
state
you
expect
it
to
have
so
yeah
thanks.
That's
one
from
my
side.
G
Yeah,
can
you
still
hear
me
yep,
I'm
getting.
We
can't
access
your
screen,
no
big
deal.
You
all
have
the
link,
I
think
the
I
don't
know
how
much
you
heard
me
say
this
new
tutorial.
I
think
it's
pretty
good.
Adam
and
afsal
did
some
beautiful
work
on
it,
but
it's
right
now
you
make
one
click.
G
And
I'm
having
to
sign
in
for
killer
coda
that
if
we
get
a
if
we
officially
sign
up
for
a
membership,
apparently
we're
free
to
use
it
as
much
as
we
want
to.
But
we
can
get
past
where
they
will
have
to
sign
in.
But
then
otherwise
you
make
one
click
and
wait
about
three
minutes
and
everything
is
installed
and
set
up
which
the
old
tutorial
you
just
went
nuts,
because
I
had
to
do
all
this
stuff
to
get
it
installed.
G
G
Yeah
it
was,
it
sounds
like
it's
okay
or
is
there
anything
I
need
to
know
for
the
future.
A
I
Exactly
so,
it's
just
a
bug,
basically
yeah.
So
we
ran
this
auto
generation
of
the
cli
documentation
based
on
when
attack
was
pushed,
regardless
of
if
it
was
a
prerelease
or
a
full
release
tag.
I
So
basically,
just
the
filter
needs
to
be
set
correctly,
so
that
pre-releases
won't
trigger
that
anymore.
And
then
you
should
only
get
a
pr
for
full
releases
and
that
should
be.
G
A
Thank
you
very
much
so
see
you
at
the
next
developer
meeting
then
in
two
weeks
it
will
be
wherever
on
thursday,
not
on
wednesday
anymore,
so
normal
time
then
have
a
great
day.
Everyone.
Thank
you.
Bye-Bye,
bye-bye.