►
From YouTube: Post Deployment Monitoring - Think Big White-boarding
Description
Post Deployment Monitoring - Think Big White-boarding
https://gitlab.com/groups/gitlab-org/-/epics/3088
B
A
A
And
focus
our
direction
towards
so
in
terms
of
how
I
see
this
meeting
going.
Is
that,
instead
of
going
through
all
of
that
past
material,
again,
I
included
as
a
little
bit
of
homework
to
briefly
check
out
the
recording
of
last
week's
meeting,
as
those
things
were
already
discussed
there,
so
this
meeting
can
be
dedicated
towards
us,
I
da
ting
and
solutionizing
and
brainstorming
on
you
know
what
things
are
eventually
possible
and
with
that
there's
this
little
introduction
and
I
think
it's
kind
of
interesting.
A
C
A
Like
like
a
common
diagram
to
kind
of
help,
us
would
you
know
align
on
as
to
where
we
are,
and
this
first
big
block
has
actually
already
been
done
by
Oriole
like
she
has
done.
You
know,
like
she's
gone
out
there.
She
was,
she
said
like
all
right,
we're
gonna,
scored
in
post-employment
monitoring.
They
were
problem,
validation
that
we're
gonna
converge
on
those
insights
or
those
those
research
results
and
create
insights,
and
then.
A
But
you
know
some
of
them.
We
will
be
certain
on
while
others
we
will
be
uncertain.
I
mean
idea
here
is
for
us
to
to
make
trackable
improvements
right.
So
with
that,
you
will
notice
that,
as
we
start
to
evaluate
those
ideas,
there
might
be
additional
ideas
that
come
up
and,
as
you
can
see,
that's
also
of
this
kind
of
diverging
part
of
the
triangle
as
well.
Let
me
see
quick
highlight
of
this
discussion
of
each
of
the
current
existing
idea
of
an
opportunity.
A
I
would
like
to
invite
orient
for
that
and
as
well
to
first
all
to
move
towards
the
whiteboard,
so
here's
a
white
board,
meeting
and
I
think
that
is
going
to
be
the
agenda
going
forward.
This
is
more
of
like
a
script
going
for
it
now.
The
mural
here
is
the
place
where
we
discuss
further,
as
it
allows
us
to
kind
of
put
our
thoughts
and
ideas
on
the
map.
I
kind
of
move
them
around
them.
A
That
helps
us
to
kind
of
make
things
visual
where
we
need
it
to
be,
and
basically
it's
a
Google
Doc
just
put
your
notes
in
a
little
document.
You
can
move
around
instead
of
like
one
huge
doc
which
were
constricts
constricted
to
the
outline
of
formatting
of
a
Google
document.
So
it's
the
same
thing
you
can
just
move
things
around
and
I
would
like
to
invite
or
briefly
to
talk
about
each
one
of
those
insights
and
not
the
idea
so
much
and
basically
keep
it.
A
A
D
Okay,
so
in
no
particular
order,
I
will
briefly
go
over
over
them.
So
first
one
is
users
want
to
get
notification
throughout
the
callback
process,
so
we're
we're
talking
about
post-deployment,
yrn
and
one
of
the
issues
there.
It
says
that
we
need
to
support
row
lock
and
basically,
what
this
incentive
says
is
that
users
want
to
know
everything
going
on
with
the
process
they
want
to
know
when
the
rollback
started.
They
want
to
know
the
rest
during
the
rollback.
They
want
to
know
that
it's
completed.
A
B
A
D
C
D
Want
to
see
the
alerts
on
the
environment
page,
you
can
want
to
see
them
on
the
pipeline
view.
Do
we
want
to
see
them
on
the
alerts
page
itself,
Tetra,
etc?
So
I
am
NOT
going
to
be
biased
and,
let
me
say
anything,
but
some
of
our
users
wanted
to
see
it
on
the
pipeline.
Okay,
so
I
wanted
to
see
alerts
their
users
required
those
auto
and
manual
rollback
methods.
So.
D
Putting
it
really
simply
not
everyone
has
automatic,
see
ICD.
So
what
a
manic
rollback
doesn't
always
make
sense,
and
for
some
people
who
have
automatic
pipelines,
they
just
don't
trust
us
enough
to
do
an
automatic
olmec.
They
want
to
make
that
decision
alone
and
some
are
really
cool
about
having
everything
automatically.
So
we
need
to
support
all
of
them.
D
D
Moving
yo
user
confer
on
a
roll
back
to
manual
for
specific
thresholds,
so
for
those
users
who
do
want
to
use
automatic
roll
back,
they
don't
want
to
do
it
always
like
there
can
be
numerous
alerts
on
the
pipeline,
but
not
every
alert
is
created
equal,
so
they
want
to
be
able
to
configure
such
manage,
which
is
the
specific
threshold.
That's
really
interesting.
In
order
to
trigger
an
automatic
rule
like,
for
example,
let's
say:
I
have
CPU
usage
over
90%,
that's
really
bad.
It
means
something
really.
D
Bad
is
gonna
happen
to
the
system
and
yeah
I
just
want
to
roll
back.
I
don't
want
to
get
in
the
notification.
I
don't
want
to
delay
anything
I,
don't
want
to
do
it
manually,
I
just
want
it
on
a
micro
back.
In
that
specific
scenario.
Now
this
is
a
very
customer
oriented.
Not
every
customer
thinks
that
every
threshold
is
the
same.
So
we
need
to
make
sure
that
there's
the
users
have
a
way
to
manage
that
next
is
users
for
first
stop
roll
out.
D
D
And
okay,
so
they
want
to
know
it
in
that
at
that
point
and
they
want
to
make
the
conscious
decision
whether
to
make
the
rollback
or
not
they
don't
want
us
to
do
it.
Just
wants
a
snap
roll
off
users
prefer
the
finding
metrics
through
UI.
So
since
these
are
metrics
that
define
actions
on
the
deployment,
whether
we're
gonna
stop
the
pointment
or
do
a
rollback,
they
want
to
see
it
visual.
They
don't
really
want
to
do
this
with
scripts
and
stuff
like
that.
D
D
So
when
we
started
with
post-deployment,
the
MVC
is
around
metrics
that
already
exist
and
get
them
so
Prometheus
built-in
metrics
off
the
top
of
my
head,
there's
error
rate
throughput
and
latency,
but
for
our
kubernetes
use
users,
health
checks
and
pods
are
just
as
important,
and
they
want
to
know
that
their
deployment
is
healthy.
So
that's
something
that
we
need
to
think
about,
and.
D
Users
want
to
roll
back
to
the
latest,
stable
version,
so
vinegar
allows
you
to
configure
what
your
roll
back.
What
version
you're
going
to
roll
back
to
I
mean
some
has
to
be
the
latest.
It
can
be
go
to
back
three
back
whatever
you
can,
you
can
decide
which
version
and
for
us
for
all
the
users
that
I
had
interviewed.
They
always
go
to
the
latest
stable
version
and
not
something
that
you
configure.
D
Just
don't
lay
this
one
that
was
Green
so
as
an
NBC
or
even
you
even
forward
from
that
I
think
we're
gonna
stick
to
that.
Go
to
the
latest.
Successful
woman
without
having
the
users
be
able
to
configure
them,
will
see
with
demand
if
they
ask
for
it
next
user
and
wants
a
roll
back
in
case
of
application
failures.
D
D
D
Basically,
no
one
cares
about
what
goes
on
from
the
dev
environment,
I'm,
sorry
to
say,
but
for
staging
and
for
production
people
want
to
know,
but
it's
not
necessarily
the
same
people
that
need
to
be
notified,
so
production.
It's
usually
like
see
like
sea
level
that
want
to
know
that
there's
some
harm
in
their
business
and
for
for
staging
it
still
in
the
devs
hands.
So
we
need
to
make
sure
that
you
can
configure
different
audiences
for
different
environments
for
the
notifications.
A
Perfect
thanks
for
that,
they're
younger
to
it,
like
the
other
two
opportunities
here
are
coming
out
of
technical
research
and,
like
I've,
documented
based
upon
the
discussions
that
we
have
gone
through
the
last
time.
So
now
we
know
a
little
bit
about
those
opportunities.
I
would
like
to
invite
us
to
now
get
interactive
into
the
second
part,
and
because
of
that,
because
I
don't
want
us
to
focus
too
much
on
the
existing
ideas
out
there.
I
would
like
us
to.
A
So
we
are
here
in
this
view,
and
this
is
now
a
figure.
The
idea
from
here
on
out
and
I
will
say
here:
we're
gonna
brainstorm,
some
ideas,
first
by
writing
down
our
thoughts
ourselves
by
applying
them
on
the
board.
Sticky
notes
this
assurance
ideas
are
written
down,
it's
easy
and
quick
to
read,
notes
and
are
not
so
much
influenced
by
each
other's
ideas,
because
this
is
a
diverging
ideation
action.
A
So
each
every
one
needs
to
be
on
the
board
and
create
their
sticky
notes
by
themselves
in
a
specific
order
in
this
empty
space,
and
then
we're
going
to
briefly
discuss
it
together,
probably
come
up
with
some
of
the
same
ideas,
but
probably
also
going
to
come
up
with
some
ideas
that
we
haven't
thought
about
before.
So
is
everyone
ready?
Is
anybody
not
clear
what
they
need
to
do,
because
if
so
I'll,
just
an
explanation,
I
would.
A
C
A
Think,
or
do
you
have
done
with
some
existing
work
already
when
you
had
created
the
inside
issues,
you
have
created
additional
feature
implementation
issues
as
well.
That
is
I
mean
it
would
have
been
ideal
if
this
would
have
done
been
done
before
that,
because
that
we've
all
chipped
in
and
we
wouldn't
be-
you
know
pre
biased
by
any
existing
ideas.
It
would
be
free
to
roam
around,
however,
we're
still
gonna,
try
and
I
think
there
might.
A
A
A
A
A
C
C
A
A
Is
good
I
think
this
is
kind
of
like
this
is
one
of
those
those
opportunity
like
those
phases
in
this
kind
of
like
aviation,
whiteboarding
practice
that
you
have
to
time
box,
because
it's
like
on
our
own
right
and
I
think
the
most
valuable
conversations
we're
gonna
have
is
is
right.
Now
we
gotta
give
a
short
presentation,
each
of
us
towards
the
others
and
explain
their
ideas.
If
it's
one,
it's
one,
if
it's
none,
it's
none.
A
A
A
So
thanks
for
that
I
have
heared
three
issues
and
one
of
them
is
show
health
of
environment
inside
of
the
merge
quest
and
the
way
that
I
was
thinking
about
this
is
based
on
one
of
these.
Let
me
see
the
health
check
is
very
important
to
keeping
these
metrics
and
I
was
thinking,
as
we
have
on
the
environment,
dashboard
or
the
environments
list
page
it
changes
into
the
environments
dashboard
as
soon
as
a
Cuban
IVs
cluster
is
attached
to
it.
We
show
additional
information.
A
I
was
wondering
if
that
health
information
he
would
not
better
be
suited
to
some
extent.
If
we,
if
the
user
is
merging
a
merchant
class,
so
the
bird
shall
purchase
what
happened
and
then
we
have
the
pipeline,
and
then
we
have
an
additional.
In
fact,
that
kind
of
you
know
showcases,
do
you
know
the
deployment
and
that
deployment
is
then
made
visible.
Visual
inside
of
that
merge
request,
which
the
release
manager
or
the
maintainer
of
the
project
has
been
set
to
emerge.
I,
don't.
A
A
There's,
of
course,
the
idea
to
be
able
to
figure
that
out
on
in
the
research
might
be
still
unknown
for
us,
like
which
ones
we
need
for
which,
but
perhaps
it
might
like
it
might
be
that
we
want
them
to
be
configurable.
So
with
that
in
mind
that
we
can
give
the
users
an
option
for
them
to
be
configurable
within
their
project,
say
like
hey.
For
these,
we
want
to
order
rollback
for
the
others.
We
don't
I'll,
leave
them
as
three
separate
ideas.
C
E
E
All
right,
so
that
is
the
first
one
which
I
wrote:
liver-lips
it's
2
for
a
sink
reporting
on
world
work
process.
So
I
think
one
of
the
critical
point
here
during
the
whole
process
is
reporting
right.
This
is
super
important.
This
robach
process
is,
like
you
know,
important
and
I
would
say
even
like
stressful
for
the
user,
so
I
think
our
focus
should
be
made
on
reporting
and
I
think
the
WebSockets
could
be
leveraged
to
do
so
so
WebSockets.
E
For
for
those
who
don't
know
what
you
know
that
were
there,
for
there
are
it's
just
whatever
enables,
like
I,
think
a
synchronous
display
of
information
right
so
just
use
use
WebSockets
and
in
our
application.
So
let's
principle
when
you
wanna
merge,
request
page
and
you
just
look
at
the
pipeline.
The
pipeline
disks
right,
you're,
gonna
see
them
like
being
updated
without
refreshing,
the
page
right,
so
that's
behind
the
scenes.
E
A
E
A
A
A
E
Yes,
and
also
when
or
it
was
going
over,
the
inside
the
new
green
tickets,
there
was
this,
like
you
know,
focus
on
just
reporting,
and
so
that's.
Why
that
what
maybe
made
me
think
of
what
circuits,
just
like
the
importance
of
reporting
but
as
far
as
like
green
cars
goes
I
think
one
of
the
first
ones
was
related
to
that.
E
So
that's
that
second
one.
Second,
one
I,
just
it's
in
it's
in
that
square,
so
at
manual
slash,
auto
switch,
but
settings
I.
Think
one
of
the
green
card
refers
to
like
the
user
wanting
to
have
manual
road
box
or
auto
world
box,
so
we
usually
have
like
switches
in
the
project
settings,
but
my
curse
like
the
the
default
way
to
this
type
of
situation.
The
problem
personally
I
find
that
settings
are
be,
are
becoming
like
very
busy
completed
with
like
lots
of
switches,
and
there
is
like
not
much
UX
I.
E
There
should
be
some
sort
of
like
UI
every
design
and
some
UX
work
needs
to
be
done.
This
is
like
off
topic,
as
I
said
in
my.
In
my
note,
the
thinks,
if
we're
going
down
that
roads
and
it's
like
the
whole
part
of
settings
into
the
project
settings
general
pipeline
I-
think
that's
the
section
that
has
like
the
most
switches.
E
A
I
mean
like
from
settings
point
of
view,
I
think
it's
always
that
we
always
need
to
consider
like
do.
We
want
to
hide
Empyrean
settings,
or
can
we
make
an
informed
decision
for
the
users?
Can
we
potentially
you
know
service
it,
a
more
contextual
area
or
preferably
can
we
skip
it
skip
a
study
out
at
all,
but
you
know
like
in
some
way
the
idea,
I
think
related
to
the
one
that
I
described.
C
B
B
A
B
Your
if
your
are
going
to
be
sending
yeah
exactly
if
you're
going
to
be
sending
like
an
alarm
to
say,
you've
got
my
email
right
like
something
went
over
a
threshold.
You
send
out
an
alarm
through
email
or
slack
or
some
other
mechanism
reusing
the
same
kind
of
trigger
mechanism
to
then
also
point
to
a
rollback
procedure,
or
you
know,
or
or
the
deployment
procedure
right
stall
here,
if
there's
a
way
that
we
could
not
like
make
that
thing
extensible
and
not
have
to
reinvent
it
every
time
for
every
specific
case.
D
D
Second
one
is
notifications
for
a
rollback
started
and
ended
via
email.
It's
like
and
that's
pretty
similar
to
what
che
said
and
also
ITN.
So,
like
the
main
events
overall
back,
they
should
initiate
some
kind
of
notification,
whether
it's
using
WebSockets
or
something
else.
That's
fine
ad
setting
for
manual
or
automatic,
rollback
and
I.
Think
the
me
sure
that's
something
that
you
had
mentioned
as
well.
A
D
A
D
D
D
Maybe
here
triggering
so
again,
the
idea
is
through
the
release,
orchestration.
You
can
configure
a
sanity
test
that
will
run
the
closed
deployment
and
then
they
will
do
something
if
they
shout
and
then
I
have
users
set
rollback,
stop
deployment
based
on
a
selection
of
metrics,
so
like
check
or
uncheck
specific
thresholds.
A
A
Sounds
good
with
that
I
think
we
can
congratulate
ourselves
already
with
this
kind
of
like
this
feels
so
new,
very
uncomfortable,
but
we
get
there
in
the
end
and
I
think
it's
kind
of
interesting
too,
if
you
like
the
ones
that
highlighted
here.
I,
don't
know
if
this
shows
up
for
everyone
else,
but
threshold
management
has
ideas
from
almost
all
of
us
in
there,
which
is
interesting
right,
like
I.
D
A
Like
I'm
sure
that
that
is
a
good
idea
to
add
them
all,
on
the
other
hand,
like
I,
would
like
to
see
for
this
specific
instance
like
perhaps
we
get
to
an
idea
which
is
more
infused
like
it's
like
infused
with
all
of
our
ideas
into
kind
of
like
one
which
kind
of
like
no
like
seems
like
the
best
idea
out
of
all
of
them,
or
is
the
one
that
kind
of
combines
elements
for
multiple
ideas?
It's
one,
it
kind
of
makes
it
most
likely
a
good
idea
to
to
to
get
there.
A
How
that
is,
gonna
be
formed
into
like
a
MVC
like
scope,
an
issue
that
is
a
separate
discussion.
I
would
say,
I
would
say
for
that:
let's
open
up
an
issue
and
have
a
discussion
there.
That's
perfectly
that
reasoning,
but
I
think
that
the
idea,
like
what
kind
of
problem
we're
solving
here
exactly
is,
is
the
most
important
thing
for
us
to
do
so.
A
If
I
look
at
this
I
see,
you
know,
we
want
users
to
set
a
rollback,
slash
stop
for
deployment
based
on
a
selection
of
metrics,
so
we
want
to
check
and
check
some
kind
of
things
there.
It's
kind
of
it
kind
of
related
or
relies
on
this,
or
relates
to
that
that
we
have
some
kind
of
like
if
I
get
it
correctly.
We
want
the
users
to
be
able
to
configure
when
it's
going
to
be
able
to
either
roll
back
or
stop
deployment.
I
was
making
it
either
auto,
rollback
or
manual
rollback
is
there?
A
D
So
and
there's
another
little
start
that
I
want
to
put
to
say
that
after
so
there's
a
difference
between
auto
and
manual,
in
the
sense
that
the
manual
action
all
we
do
is
just
you
know,
surface
the
alerts
and
let
the
user
make
the
final
call
whether
they
should
do
it
and
so
I,
don't
know
you
can
think
about
it.
Making
flashing
lights,
saying
you
have
a
problem
with
your
pipeline.
Do
something,
of
course:
no!
D
That's
not
it
in
that
sense,
just
so,
you
know
get
their
awareness
and,
at
the
end
of
the
day,
they're
there
they're
the
ones
that
have
to
press
the
button,
and
this
has
a
lot
of
importance,
not
only
because
they
said
it.
It
also
has
to
do
these
permissions,
who
is
allowed
to
press
the
button
and
also
for
compliance
and
auditing.
So
this
has
a
lot
of
you
know
things
that
we
need
to
think
about
and
when
you
give
on
a
medical
back
you're,
basically,
you
know
I'm.
D
E
Yeah,
no
sorry
right,
yeah,
yeah
I
was
just
gonna,
say
something,
but
my
connection
dropped
at
the
same
time,
so
that
was
not
great
yeah
just
to
jump
on
what
or
it
said
that
that
ability
to
give
the
control
to
manually
roll
back.
This
is
something
that
can
be
leveraged
today,
I'm
just
thinking
about
it.
Now
in
Piper
pipeline
configuration
in
the
grid
live
CI
file.
There
is,
there
are
some
attributes
that
are
when
manual
I
mean
there
is
like
manual
manual.
True,
there
is
something
that
exists
already.
That
sets
a
job.
E
D
E
Yeah
yeah
I'm
more
familiar
with
the
mo.
My
point
was
just
there
may
be
some
overlap
between
CI
configuration
and
settings
right,
and
so
we
need
to.
We
need
to
find
like
work.
Where
is
place
to
to
to
define
all
of
this,
which
is
like
interracial
management
section
right
so
there
we
need
to
consider
also
this
a
pipeline
and
the
tools
that
it
provides.
D
Yes,
absolutely
so
so
that's
the
manual
part
and
automatic
prototype
is
basically
putting
on
a
mission
on
it
and
kind
of
gives
you
as
a
whoever's
in
charge
of
the
deployment
piece
of
mind
that
you
know
something
goes
really
wrong,
something
it's
gonna
be
taken
care
of,
and
you
don't
have
to
be
really
mindful
about
going
and
pressing
a
button
and
getting
all
the
alerts
on
time.
You
know
so
we're
all
for
automation.
If
you
can.
D
Depends
we're
asking
some
of
the
users
wanted
manual
and
some
of
them
wanted
automatic
the
again.
The
reason
why
people
don't
want
automatic
is
for
two
reasons:
one
they
may
not
have
an
automatic
pipeline,
so
automatic
robot
doesn't
give
them
anything.
They
just
don't
have
CIC
Dean,
please
or
number
two.
Is
they
don't
trust
it
in.
C
D
D
A
D
Yes,
I
wanted
to
mention
that
you,
you
wrote
on
the
orange
sticky
notes
predefined
table
of
metrics.
It's
important
to
note
that
users
can
also
create
custom
metrics.
It
doesn't
have
to
be
our
out-of-the-box
metrics,
because
we're
connected
to
Prometheus
you
can
configure
whatever
you
want.
So
we
may
not
know
the
metric
that
they
want
to
say,
roll
back
from
yeah.
A
Would
you
say
in
that
case
it
makes
kind
of
a
case
for
them.
It's
like,
like
have
a
predefined
table
of
metrics
I,
understand
that
predefined
canal,
a
bias
towards
existing
ones,
but
in
the
creation
of
a
custom
metric?
Would
it
then
be
registered
to
this
kind
of
table
where
you
immediately
have
access
to
it?
I
say
like
hey.
D
D
A
E
E
It's
just
that
this
exercise
made
me
think
of
a
previous
functionality
and
I
remember
putting
a
switch
in
the
settings
and
I
think
that
was
the
right
thing
to
do
so
it
could
be
the
right
thing
to
do
here
as
well,
but
maybe
the
right
thing
it's
all
CI
configured.
You
know
how
to
have
it
in
the
CI
configuration
so
I.
Don't
know
there
is
a
maybe
here
I
should
have
added,
maybe
before
add
manual,
auto
switch
button
settings.
E
E
D
I
agree
with,
and
it
doesn't
have
to
be
a
setting,
but
it
needs
to
be
super
clear
to
the
user.
What
they're
expecting
do
I
want
my
pipeline
to
automatically
roll
back
yes
or
no,
and
do
I
want
to
give
an
options:
specific
users
whether
they
can
manually,
you
know,
click
the
button.
So
I
know
it
can
be
a
tableau
in
the
environment
page
as
well.
It
says
manual
automatic
choose
one,
but
it
definitely
needs
to
be
defined
for
her
project
for
environment
and
I.
D
I
was
very
predictable,
although
people
have
different
projects
with
different,
you
know
granularity
and
they
may
care
about
some
areas
in
there
in
their
code,
which
may
be
a
separate
project
and
then
the
main
business,
the
core
business
they
want.
You
know
roll
back
song,
but
some
you
know
experiment.
They
don't
really
care
right.
Would.
A
D
Vinegar
has
thresholds,
for
example,
let's
say
we're
talking
about
CPU
usage
and
I
want
to
automatically
roll
back
in
case.
My
seeking
new
usage
is
over
90%,
but
I
know
100%,
but
in
cases
80%
I
just
want
to
know
about
it.
I
don't
want
to
roll
back.
I
want
to
know
about
it,
maybe
on
the
side
manually
to
go
back,
but
I
don't
want
the
machine
to
do
that
on.
D
A
All
right
thanks
so
much
I
think
we're
gonna,
we're
gonna,
leave
it
at
that
for
now.
Here,
I
think
we
got
to
some
extent
and
we
explored
one
of
the
kind
of
themes
that
came
out
of
at
least
three
of
the
four
people
here,
and
you
know
there's.
This
was
like
one
of
the
first
takes.
We
got
at
this
I
I
want
to
thank
all
of
you.