►
From YouTube: Developer Experience Office Hours: Datadog
Description
OpenShift already provides you with some cluster metrics, but not much at an application level. In this session, Joel Lord from OpenShift will be joined by Ryan MacLean to see how to install Datadog in an OpenShift cluster and see how it can be used to find issues within your applications.
A
Good
morning,
good
afternoon,
good
evening,
hello,
everyone
welcome
to
another
developer
experience
office
hours,
I'm
very,
very
happy
to
be
joined
by
Joe,
Lord
and
Ryan
Jarvan,
and
today
we
also
have
a
special
guest
Ryan
McLean,
the
one
and
only
Ryan
McLean
from
data
dawg.
He
is
here
today
to
talk
to
us
about
all
things
day
to
dog
and
open
shift.
It's
what
I'm
hearing
where
aids?
Let's,
let's
just
do
a
quick
round
of
introductions,
so
everybody
here
knows
each
other
and
Joel.
If
you
want
to
start
off,
that'd
be
great
yeah.
B
I'm
gonna
get
started,
hi
everyone,
I'm,
Joel,
Lord,
I
work
as
a
developer
advocate
on
the
OpenShift
platform
here
at
Red,
Hat
I
come
from
a
developer
background.
I
started
using
some
of
those
technologies
like
OpenShift,
kubernetes
and
even
containers
not
very
long
ago,
so
I'm
still
kind
of
new
to
all
this
world,
then
I
come
from
a
very
developer
background
really
thing
that
I
was
trying
to
do
recently
was
to
get
data
dog
and
to
see
how
it
can
help
me
to
kind
of
see.
B
What's
going
on
inside
my
applications
and
I
was
trying
to
set
that
up
on
OpenShift
and
I
had
a
few
issues.
So
that's
when
I
actually
figured
out.
Why
not
reach
to
the
experts
and
I
reached
out
to
Ryan
and
I
said
hey?
Why?
Why
don't
we
actually
do
a
twitch
stream,
so
I
can
actually
learn
as
the
same
time
as
everybody
else
so
I'm
glad
to
have
you
Ryan?
Why
don't
you
introduce
yourself
yeah.
C
No
problem
so
I'm
Ryan
McLean
that
data
dog,
my
official
title
is
a
tech
evangelist.
We
were
laughing
about
that.
A
little
bit
earlier
might
change
in
the
near
future.
You
never
know,
but
I've
been
been
in
the
industry
for
about
20
years,
I
started
as
like
at
bench
techs
all
the
way
down
from
you
know,
working
on
laptops
and
stuff,
like
that,
all
the
way
up
to
helping
people
out
with
your
infrastructure,
both
large
and
small
these
days.
C
What
I
try
to
do
is
help
people
with
getting
things
set
up
in
the
cloud
and
starting
to
monitor
their
apps
as
well
as
getting
that
that
big
picture
kind
of
thing
clang,
so
so,
I
think
when
you
and
I
talked
we
talked
a
little
bit
about
so
the
difference
between
like
infrastructure,
monitoring
and
application
monitoring
and
how
the
two
might
work
together
and
that's
that's
kind
of
where
a
lot
of
people
are
these
days
as
well.
So
that's
this
one
trend,
oh
yeah,.
B
Exactly
so,
I
was
my
main
question
was
why
why
should
I
actually
use
data
dawg
when,
like
I
I,
do
have
all
of
those
metrics,
and
it
tells
me
stuff
about
this
usage
and
and
memory
usage,
which
you
know,
I,
don't
necessarily
really
care
a
lot
about
as
a
developer,
but
I
was
like
well.
Why
would
I
even
bring
data
doc
to
get
more
metrics
right,
yeah.
C
So
you
can
see
things
like
stuff
like
function,
calls
or
method
calls,
or
even
calling
third-party
api's
and
get
metrics
on
those,
but
not
just
for
like
those
calls
in
aggregate
or
maybe
on
average,
but
also
how
they
might
depend
on
each
other.
So
if
you've
got
micro
services
like
if
one
is
calling
another,
what
that
looks
like
or
if
you're,
seeing
like
a
lower
rate
of
calls
for
your
team
service,
but
it
looks
like
everything's
flying
on
your
end,
sort
of
being
able
to
see
what's
upstream,
that
can
be
really
valuable.
C
I
think
that's
where,
like
that,
the
APM
crowd
kind
of
comes
in
right
and
that's
that's
sort
of
our
place.
So
the
infrastructure
side
I
mean
to
be
honest,
like
you'll,
get
a
lot
out
of
a
default
Cooper
days,
and
so
you'll
get
a
lot
out
of
a
default
cloud.
Install
like
a
VMware,
just
install
a
red
heavy
em
install
all
those
will
give
you
like
they,
your
CPU,
your
memory
or
disk
IO
stuff,
depending
on
being
able
to
use
that
as
context
for
your
applications
performance.
So
what
the
business
really
cares
about
right?
C
Can
people
actually
check
out,
as
opposed
to
like
it's
a
server
app
I
think
is
maybe
more
important,
so
sort
of
flipping
that
problem
on
its
head
and
we
get
it
from
like
the
the
business
angle
and
then
drilling
down
into
your
CPU
memory
and
metrics.
Like
that,
so
it's
just
a
different
way
of
looking
at
the
same
stuff.
It
give
me
a
bit
more
context.
At
the
same
time,
yeah.
B
Absolutely
when
you
showed
me
the
other
day,
this
APM
stuff,
I
was
really
impressed
and
I
was
very
excited
and
then
I
tried
to
install
it
and
and
unfortunately,
I
wasn't
able
to
get
it
to
work.
To
be
honest,
I
didn't
try
very
hard.
It
seemed
like
something
that
was
very
straightforward,
but
I
just
had
five
minutes
to
test
it
out
and
and
I
had
never
actually
got
it
got
it
to
work.
But
knowing
that
you
were
gonna,
be
here,
I
figured
you
know.
Well,
we
can
just
do
it
together.
C
Yeah,
yes,
so
now
I'm
here,
let's,
let's
just
talk
about
like
if
it
doesn't
work
so
so
one
is
on
I.
Think
one
of
the
things
you're
trying
to
do
is
install
something
that's
pretty
new
to
us
as
well.
So
there's
there's
a
few
methods
of
installation
inc
in
kubernetes
if
you're
not
used
to
it.
Basically
there
there
was
a
whole
push
behind
a
lot
of
people,
got
home,
working
and
like
to
sort
of
use
as
a
package
manager
for
kubernetes
we're
moving
away
from
that
to
operators
and
I.
C
Think
redhead
is
as
well,
but
on
the
operator
side,
I
got
to
admit
where
we're
still
pretty
new
to
this
area
in
terms
of
having
it
being
customer
facing,
even
though
we
use
it
internally
and
because
of
that,
its
beta,
so
I
think
some
of
the
problems
you
were
having
is
because
its
beta,
we
can
try
it
out
again
today
and
figure
out.
What's
going
on
there.
C
The
long
and
the
short
of
it
is
the
the
install
into
openshift
is
very
similar
to
a
vanilla,
grenades
install,
but
there's
there
are
some
things
like
roles
and
permissions
and
privileges
I
think
are
specific
to
open
shift
to
allow
us
to
get
some
of
that.
Like
networking,
for
example,
I
think
we
need
a
little
bit
more
and
I
have
a
feeling
that
when
I
looked
at
the
operator,
I
didn't
see
a
lot
of
that
in
there.
It
could
be
that
just
needs
to
be
massaged
a
little
bit.
C
I
also
know,
and
I
was
hoping
that
there
would
be
a
newer
release
of
the
operator
coming
out.
It
hasn't
launched
yet,
but
it
should
be
out
in
a
little
bit
we're
trying
to
do
an
internal
conference
here.
So
I
know,
we've
pulled
a
lot
of
resources
in
that
area,
but
the
operator
definitely
is
our
focus
and
India
is
where
we
want
to
be
there's
anything
wrong
with
helm
or
default
manifest
it's
just
what
we're
using.
It
seems
to
be
a
big
push
in
the
industry
as
well,
so
we're
gonna
follow
suit.
Yeah.
B
D
B
C
C
Infrastructure
is
set
up
in
the
cloud
we
can
just
sort
of
log
into
and
show
people
you
know
just
like
a
normal
application
and
and
what
that
would
look
like
so
something
with
quite
a
few
micro
services
and
quite
a
bit
of
interaction
between
these
that
it
allows
us
to
just
talk
about
how
you
can
drill
down.
Look
at
look
at
different
micro
services
talking
to
each
other
kind
of
thing,
I'm
just
going
into
it
here
and
I'll
see
if
I
can
drag
this
window
over.
Let
me
know
if
the
screen
res
is
too
high.
C
A
C
D
C
Can
sort
of
read
on
my
end?
Can
you
please
read
that
there
that's
better,
better
good?
This
is
some
inside
baseball
here
it
on
the
tree,
sensei,
okay,
so
I
just
logged
in
you
can
see
my
email
there
and
you
can
see
that
I've
got
some
stuff
running
I'm,
just
gonna
quickly,
minimize
this
sidebar
because
we're
not
getting
any
of
it.
Basically,
when
you
log
in
you
can
see
like
all
your
services
and
we're
talking
about
a
p.m.
C
So
we'll
just
talk
about
like
some
of
these
services
here
and
specifically
what
we
were
talking
about,
which
ole
is
just
tracing
I,
guess
in
in
particular
so
ways
of
looking
at,
and
this
is
like
a
like
a
15
minute,
but
we
can
look
at
live
trace
here,
so
what's
actually
going
on
in
in
this
okay,
so
it's
life
actually
looking
at.
What's
going
on
in
this
store
and
I'll
just
give
some
quick
background
here.
So
this
story
is
imagine
you've
decided
to
start
selling
t-shirts
with
dogs
on
them.
C
Maybe
some
hats
with
dogs
on
them,
some
sort
of
dog
branded
merchandise.
For
a
fictional
story.
It's
got
an
ABS
ad
service.
It's
got
a
web
store.
It's
got
some.
You
can
see
Redis
going
through
here.
It's
good,
quite
a
few
components.
It's
got
some
80
authentication
as
well.
If
you
want
to
just
have
have
that
stuff
set
up,
so
it
allows
you
to
look
at
like
a
bunch
of
services
working
together.
You
can
see
that
there's,
there's
actually
traffic
going
to
the
site
right.
C
It's
flying
by
there's
errors
that
kind
of
stuff
I'm
just
going
to
quickly
pause
this.
So
we
can
sort
of
drill
down
and
a
couple
of
these
errors.
So
you
can
see
like-
and
this
is
typical
you'll
you'll
get
like
your
your
200
or
your
2x.
X
is
3
X
X,
maybe
not
so
much
your
4
X,
X
or
5x
X,
though
they'll
come
through
and
generally
what
you'll
want
to
do
is
not
necessarily
look
at
them
specifically,
but
maybe
in
the
aggregate
or
on
the
whole,
so
kind
of
looking
at
the
whole
thing.
C
So
traditionally,
what
you
do
is
you
kind
of
go
in
and
then
just
just
gonna
left
hand
side
here,
I
got
rid
of
those
okay,
so
all
those
200
certs
are
gone
and
now
I'm,
looking
at
the
four
hundreds
and
the
five
hundred
here
so
specifically
the
five
hundreds.
What
do
we
got
and
sort
of
I
was
telling
you
all
is
that
we
can
actually
click
into
these
and
then
we
can
get
it
almost
like
a
flame
graph.
So
if
you've
ever
used
something
like
dtrace
I
in
the
command
line
to
look
at
your
app.
C
This
is
what
they
look
at.
If
you
haven't
that's
fine
as
well,
but
basically
in
this
case
you
can
see
at
the
top
here.
We've
come
in
and
we've
hit
the
the
laravel
service
for
pricing
and
in
this
case
it's
using
a
clustering,
goop
cloud,
but
basically
imagine
it's
it's
any
infrastructure.
You
can
see
that
it's
look
like
like
a
third
of
a
second
to
load.
It
up
the
kind
of
things
which
it
takes
well,
I'm
just
clicking.
C
Like
a
nice
juicy
one
here,
okay,
so
basically
what
happens
with
the
flame
graph?
Is
you
see
the
top
coin
come
in
and
then
under
that
you'll
see
anything
else
that
it
called,
and
some
of
these
can
be
can
be
pretty
involved.
Some
of
them
can
be
a
little
bit
less
interesting
and
basically
you
can
see
them
on
the
left
hand
side
here
when
there's
multiple
services,
so
just
going
through
to
see
if
I
can
get
one.
That's
got
like
quite
a
few
calls
in
it,
so
this
one
might
be
a
little
bit
better
again.
C
It
hits
level
it
goes
into
level
calling
itself
basically
for
an
action
when
I
click
on
these
I
can
actually
go
right
into
what's
happening
here
and
I
can
see
these
areas.
So
there's
a
guzzle,
a
cgb
sort
of
exception
and
saying
that
there's
a
bad
gateway,
a
kind
of
thing,
but
I
can
actually
go
all
the
way
down
into
each
one.
So
this
is
expressed,
for
example,
I
can
get
down
all
the
way
into
the
Express
middleware
and
then
all
the
way
to
the
HP
out
get
in
this
case.
C
It
failed
at
the
top,
so
a
lot
of
the
failures
aren't
gonna
percolate
down
into
the
middle.
We
can
go
all
the
way
down
here
and
see
that
you
know
Express
is
trying
to
call
another
service
and
it's
actually
getting
a
find
missing.
Now
you
can
see
that
it's
then
calling
the
middleware
that
actually
goes
down
to
Redis
and
says
like
hey.
C
This
is
not
how
you
use
a
get
call
to
call
rid
us
kind
of
thing
so
along
the
shirt
here
is
a
you
can
actually
drill
down
into
each
call
for
each
service
and
within
them.
If
it's,
if
it's
a
problem
in
the
actual
code,
you
can
actually
look
at
the
the
functions
or
the
methods
themselves.
Now.
Some
of
this
is
done
automatically
and
that's
great.
C
What
I
would
recommend,
though,
is
once
you
start
using
a
service
like
this
you'll
probably
want
to
start
making
your
own
metrics
or
your
own
business,
metrics
or
things
that
are
of
more
value
to
you.
Then
then,
possibly
like
the
box
stuff.
That
being
said,
the
others
box
stuff
will
get
you
started
and
get
you
started
quickly,
which
is
what
we're
hoping
for
today
is
to
do
what
we
call
monkey
patching,
but
a
way
to
sort
of
almost
side
load
your
stuff
in
there.
C
So
I
think
we
were
talking
about
node,
there's
ways
of
just
essentially
importing
at
the
start
of
your
file.
You
might
need
to
you,
know,
rebuild
your
container
and
relaunch
it
or
or
to
launch
to
it
like
a
test
cluster
or
staging
cluster
prior
to
just
you
know,
Yolo
eating
it
to
product
kind
of
thing,
making
sure
that
you
got
it
working
first.
I
put.
C
That
being
said,
it's
it's
generally
pretty
straightforward
in
terms
of
kubernetes
I'm,
pretty
familiar
with
that
part
and
Joel
I
might
lean
on
you
a
little
bit
on
that
on
the
note
slide
in
terms
of
communities,
it's
the
same
of
installing
any
other
like
any
other
helm
charge
or
any
other
manifest
applying
it
to
the
clusters
very
similar
methodology.
So
that's
basically
it
that's
that's
kind
of
like
the
big
tour
you
can
go
from
here
into
anything
else.
You
can
click
on
this
I'm
Pro.
C
If
you
want
to,
you,
can
actually
go
right
into
the
hosts
that
serve
this
in
this
case
is
probably
a
container
excellences
gke.
You
can
get
it
all.
Those
metrics
that
we're
talking
about
before
the
infrastructure,
metrics
will
come
off.
No
problem
and
I
can
go
in
and
look
at
the
network
traffic
dis
latency
you
can
from
here.
You
can
start
going
into
the
cluster
and
stuff
like
that
as
well,
and
as
you
go
through
here,
you
can.
If
you
need
to
correlate
anything
in
here,
you
can
do
that
as
well.
That's
basically
it
so!
C
You
can
pivot
back
and
forth
between
the
the
front
end
different
microservices,
different
API
is
different,
API
backends
and
then
the
Peter
clusters
as
well
so
allows
you
to
hit
quite
a
few
different
angles.
If
you
want
to
drill
all
the
way
down
into
logs
as
well,
you
can
do
it
that
way.
So,
if
you
want
down.
B
C
C
Abilities
will
sort
of
taking
all
of
these
and
bringing
them
together.
No
people
will
say
it
means
different
things
for
different
vendors
right,
you're
logging
vendor.
It
will
tell
you
it's
it's
about
logs
your
event,
vendor
away
it's
about
events
and
for
us
it's
more
about
grabbing
the
signals.
So
what
signals
can
we
give
you
that
allows
you
to
give
you
a
big
picture?
I.
F
D
F
C
It
can
be
so
so
yes
I
know
so
so.
Basically,
a
lot
of
this
you
can
see
it's
X
request.
Id
is
is
actually
how
we
tag
everything
together.
So
basically,
as
things
are
coming
in
to
your
services,
we're
putting
things
matter
to
say,
as
it
goes
downstream,
how
do
we
know
what's
what
kind
of
thing
the
tags
themselves
like
if
we
go
into
infrastructure
here?
C
Some
of
these
tags,
like
in
in
openshift,
the
tag
might
be
like
a
project,
I
think
we'd,
probably
a
common
tag,
maybe
the
node
that
you're
running
or
possibly,
if
it's,
if
it's
somatic,
a
specific
name
space
or
a
user's
name
space.
That
can
be
a
tag
as
well
in
this
case,
I
think
most
of
the
tags
that
you're
seeing
here
are
things
like
the
environment.
So
an
environment
can
mean
different
things
to
different
places,
but
a
lot
of
times
it
would
be
like
your
test.
C
Staging
your
product
might
be
a
tag
so
that
it
takes
a
key
value
just
to
just
to
clear
that
up
so
generally,
a
tag
will
be
like
what
type
of
tag
and
what
is
the
value
of
the
day.
So,
in
this
case,
it's
a
little
bit
small
here.
Also
you
can,
if
I
can
move
in,
but
basically
this
clustered
location
here
are
the
clustered
name,
for
example
our
tags
with
a
cluster
ID.
We
can
use
those
tags
so
then
pivot
into
the
cluster
or
pivot
into
the
environment.
That's
a
good
question,
like
the
tags.
D
F
F
C
So
a
lot
of
this
is
done
default
in
the
cloud
provider
or
I.
Guess
you
could
say
default
in
the
infrastructure
provider,
so
in
this
case
we're
in
gke.
So
there's
a
lot
of
gke
tags
in
here,
but
if
we're
in
AWS
or
in
Azure
or
what
have
you
we've
grabbed
those
tags
from
that
environment
as
well,
but
the
the
corollary
here
is
to
get
that
data.
We
need
to
actually
have
access
to
read
from
that
environment.
So
a
lot
of
times
you'll
have
data
dog
almost
almost
scraping.
C
F
That
sounds
like
a
huge
advancement
to
automatically
get
this
level
of
markup
appended
on
requests
as
they
fire
through
a
distributed
system
and
then
allow
you
to,
like
you
said,
do
that
pivot
and
and
then
drill
down
with
a
lot
more
certainty
to
as
to
what
you're
looking
at
and
what
the
related
edge
cases
may
be.
If
they're
highlighted
via
these,
these
tags
right,
yeah.
D
E
C
That's
your
favorite
shape,
so,
to
be
honest,
I've
been
using
data
dog
for
almost
10
years
now
it
feels
like
that's
sort
of
impossible
but
yeah,
basically,
as
it
came
out,
I
kind
of
get
interested
into
it.
So
this
is
what
I
was
used
to
when
I
when
I
started
using
it
as
well
as
a
hex
to
you
right.
So
so,
where's
where's,
all
my
in
fright
and
you
can
tell
we've-
got
a
fairly
large
demo
environment
here,
but
basically
the
the
hexes
are
kind
of
a
way
of
it.
C
The
reason
we
use
hex
is
just
just
so
you
all
know
it's
just
because
you
can
fit
more
on
a
graph
and
you
use
hexes
as
opposed
to
like
squares
or
circles.
You
can
just
get
more
on
here.
So
that's
that's
why
we
use
them,
but
it
the
hexes,
do
it.
They
basically
allow
you
to
look
at
your
infrastructure
as
a
whole
on
one
screen
pretty
easily
and
then
sort
of
just
pick
out
the
orange
ones
right.
Oh
the
red
ones
or
whatever
your
visualization
is
to
see
what's
having
problems
but
they're
still
there.
C
So
there's
a
lot
of
tabs
here
on
the
left
like
when
I
started,
using
gaeta
dog
this
left
hand,
side
may
have
had
a
couple
icons
and
and
we're
up
to
a
whole
bunch
of
them
now
with
like
sub
subheadings
here.
So
the
hex
stuff
generally
exists
under
this
infrastructure.
So
it's
the
host
map
is
what
we
call
it
so
we're
all
the
host
is,
but
we've
now
got
like
containers
processes,
servers,
server,
less
network
network
map,
all
that
stuff,
there's
quite
a
bit
going
on.
So
one
of
the
more
common
ones.
C
These
days
that
we
like
to
talk
about
is
sort
of
like
the
network
map.
So
how
do
things
relate
to
each
other
now?
This
is
a
bit
of
spaghetti
here
that
I'm
showing,
but
if
we're
going
to
look
at
those
services
like
how
do
they
relate
to
each
other,
this
is.
This
is
maybe
like
the
new
way
that
the
new
hex
map,
if
you
will
like
the
the
new
APM
kind
of
way
and
and
again
like
there's
a
there's,
a
few
vendors
out
there.
C
That
will
do
this
as
well
like
I'm,
not
I
can't
do
the
hard
sell
here
either
but
ways
of
relating
your
services
together
graphically-
and
you
know
the
red
green
right.
So
if
something's
red,
let's
stroll
on
into
it,
let's
see
what's
going
on
anything
but
I'm
glad
you
asked
cuz,
yeah
I
might
have
just
glossed
over
it
and
said:
let's
talk
about
APM
today,
but
this
does
allow
you
to
certainly
drill
into
your
in
for
us.
Well,
if
you
want
to
look
at
it
from
that
aspect,
I.
A
There's
a
lot
of
back-and-forth
about
your
bitter
things,
but
you
know
no
specific
questions
regarding
a
topic
at
hand.
I
mean
folks
tracing
is
huge.
Right,
like
is
the
the
capabilities.
I
think
can
be
a
little
daunting,
potentially
maybe
for
some
people
right
like
there's
a
lot
of
stuff
that
you
might
have
to
filter
through
to
find
what
you
actually
need,
given
that
like
moment
context
whatever
it
is
you're
looking
for.
So
how
would
you
help
you
know
kind
of
people
just
getting
started
with
Tracy
and
try
to
figure
out
like
what
is
I
mean?
C
So
it's
almost
to
say
like
if,
if
what
matters
to
you
is
daily
active
users
or
monthly,
active
users
or
or
signups
or
check
outs,
what
in
your
code,
do
you
need
the
instrument
to
give
you
that
visibility
and
then
how
do
you
trace
from
there?
So
how
do
you?
How
do
you
put
like
a
tracer
bullet
into
that
specific
area
that
code?
So
if
it's
daily
active
users
may
be
signups
is
something
you
want,
or
maybe
it's
like
front
page
hits
or.
C
And
that
could
be
hard,
but
a
lot
of
times
those
conversations
involve
sitting
down
with
your
stakeholder
and
saying
you
know
if
we
had
to
put
a
KPI
on
our
site.
What
would
that
keep
yet
like
if
we
had
to
have
one
KPI?
What
would
that
look
like
a
lot
of
times?
They'll
give
you
you
know
four
or
five
and
they
can
say
okay
if
we
were
to
wait
these,
which
should
we
wait
higher
or
if
we
were
to
score
these
like
it.
C
If
you
know
a
plus
B,
plus
C,
plus
B
plus
C
equals
X
once
we
need
to
wait
them
for
to
get
it
and
I'm
kind
of
I'm
being
generic
here.
But
if
we
were
to
talk
about
like
this
site,
for
example,
the
store
site,
one
of
them
would
be
like
how
many
checkouts
do
we
have
in
the
day
like
how
many
sales
right,
that's
pretty
common
for
a
sales
org,
how
many
and
if
you're,
gonna
look
at
it
as
a
funnel
right
like
how
do
people
get
into
sales?
C
I'm
kind
of
often
left
field
here,
a
little
bit,
but
sometimes
when
you've
got
that
done,
it
allows
you
to
do
a
little
bit
of
almost
like
testing
like
like.
Let's
try
this
new
story
and
see
if
it
works
better
or
let's
try
this
new
landing
page
and
see
if
it
works
better,
so
allows
you
to
experiment
as
well
when
you
get
there,
but
basically
the
long.
The
short
of
it
is
to
figure
out
what's
important
to
your
business,
a
lot
of
times
you
have
to
talk
to
the
business.
Basically
right.
A
You're
all
after
the
same
thing,
but
if
you're
not
all
talking
about
the
same
thing
and
you're,
not
looking
for
the
same
things
and
you're,
not
finding
the
right
signals
to
indicate
success
or
failure
in
some
instances,
you're
gonna
be
kind
of
lost
in
the
sea
of
you
know.
What
is
it
we're
trying
to
do
here
so
having
that
conversation
and
getting
that
balanced
as
far
as
what
you're
looking
for
specifically
is
vitally
important.
C
You
got
it
and
you
mentioned
yes.
Are
we
there
abyss
well
I'll
touch
on
that,
but,
like
then,
there's
also
like
the
downtime,
pretty
the
expected
downtime,
where
the
times
of
things
fail.
Maybe
your
planned
maintenance,
but
there's
some
of
that
too,
and
and
when
it
comes
to
the
business
like
it's
the
same
thing,
you
gotta
communicate
like
what
your
your
error
budget
is
or
or
like
what
the
reputational
damage
of
five
minutes.
Downtime
during
Black
Friday
might
be
versus.
Yes,.
C
In
in
March
some
random
day
kind
of
thing
right,
it's
it's
totally
different,
but
at
all
it
all
kinds,
it's
sort
of
packed
together.
You
say
so
I'm,
honestly
glad
coming
from
like
a
few
different
angles
in
the
business
that
we're
starting
to
realize
that
you
have
to
have
shared
goals
and
you
have
to
be
able
to
understand
the
indicators
you're
using
or
share
in
the
whole
company
and
you're
all
working
towards
that
same
thing.
C
So,
if
it's
more
sales,
then
everything
should
be
enabling
that
if
it's
more
viewers
on
the
twitch
stream,
then
everything
we
do
should
enable
that
as
well
I'm
sort
of
like
working
towards
that
one
or
those
two
were
those
three
KPI
is
to
make
sure
you're
there
and
I
know.
Kpi
is
kind
of
a
word
that
people
don't
like,
but
basically
those
indicators
right.
Those
things.
A
A
Kpi
see
what
you
want
about,
you
know
KPIs,
and
it's
it's
more
so,
just
like
anything
right,
DevOps
sres,
the
way
you
implement
it
right.
So
if
you
have
a
negative
experience
with
KPIs,
it's
it's
sad,
but
it
might
have
been
the
way
that
this
was
done.
So
the
the
idea
behind
having
those
indicators
are
critical
right,
like
monitoring,
is
like
that
base
level
of
this
Ori
pyramid.
A
You
know
having
tracing
as
part
of
that
scene
where
certain
people
are
falling
off.
The
cart
potentially
are
falling.
You
know
falling
out
of
the
buying
process
or
falling
off
the
website
or
whatever
it
is
that
metric.
It
is
you're
looking
for
right,
like
they
fall
off
the
stream.
They
turned
out
by
mail,
boring
me
whatever,
but
yeah
the
the
the
those
those
metrics
for
success
are
critical
and
and
having
them.
You
know
indicated
and
traced
and
observed,
or
vitally
important,
yeah.
C
And
I
see
Ryan's,
adding
some
some
stuff
here
on
the
acronyms.
That's
great
cuz
I
often
do
forget
to
define
like
what
Nestle
is.
What
missile
is
is
I
tend
to
its
acronym
super
bright,
but
he
he's
got
a
good
point
here.
So
it's
it's
important
to
say,
like
a
solo,
is
service
level
objectives,
what
you're
trying
to
achieve
and
they're
informed
by
the
SLI.
So
when
we're
talking
about
those
indicators
as
opposed
to
the
KPI,
is
the
SL
is
of
the
service
level
indicators
thanks,
Ryan
I
totally
apologize
for
not
deploying
them.
A
C
D
C
Know
I
worked
at
a
mobile
phone
company
today.
I
was
here
not
today
prior
to
today,
so
it'd
be
like
is,
is
a
daily
active
user
on
your
phone,
using
like
a
tick-tock
or
an
Instagram,
or
are
they
interacting
with
your
mail
app
like
what
does
that
daily
I
couldn't
I
mean
it's?
It's
really
different,
depending
on
the
company
that
you're
at
so
it's
it's
worth.
Having
that
conversation,
it's.
A
A
They
should
you
know,
keep
you
I
should
align
to
that
across
the
organization.
I
mean
there's
some
tools
that
can
help
you
with
this,
but
there's
a
lot
better
outputs
from
people
that
work
as
an
organization
that
then
people
that
work
in
silos
we've
seen
this
over
and
over
we've
seen
entire
industries
get
upheaval
to
buy.
You
know
DevOps
and
people
just
working
better
with
data
right
yeah,
so
you
know
I
mean
going
back
to
you
know.
A
A
A
that's
a
public
health
concern
so
having
tracing
in
here
and
having
you
know
that,
like
hey,
they
hit
this
checkbox
and
then
they
hit
submit,
and
the
next
page
comes
up
it's
taking
too
long
because
of
this
weird
Java
library
or
whatever.
Right
for
this
one
page
like
you,
won't
know
that
unless
you're
actually
observing
and
monitoring
these
things
and
doing
it
in
something
like
data
dog
makes
it
a
little
bit
easier
and.
C
I
think
you
you
sort
of
you're
hitting
in
a
home
here,
but
that's
the
other
thing
is
that,
at
least
in
my
career,
a
lot
of
what
I
looked
at
was
like
the
mean
or
the
average,
and
if
it's
just
one
of
those
surveys
that
somebody's
not
doing
for
their
Tuco
mid-nineteen
test,
just
like
that,
you'll
ignore
it.
But
these
days
a
lot
of
what
we're
doing
is
looking
at
like
the
99th
percentile.
The
ninety-nine
point,
nine,
the
ninety-nine
point,
nine
nine.
So
let's
look
at
those
bad
experiences.
C
C
Like
everybody
else
is
fine,
but
the
VP
you
got
in
the
elevator
and
tried
to
load
the
site
and
it
didn't
load,
and
then
they
come
back
and
tell
you.
Ok,
your
site
didn't
load
and
then
you're
looking
through
your
logs
and
everything
seemed
to
load,
and
there
isn't
is
like
you
didn't
you
didn't
get
that
hit
like
you,
weren't
instrumenting,
the
app
on
the
phone,
your
instrumenting
on
the
back
end
and
the
hit
didn't
come
to
you.
So
you
didn't
see
it.
C
So
some
of
this
is
like
some
of
its
end
to
end
some
of
its
not
looking
at
the
average
looking
at
your
actual
nines.
So
your
percentile
sorry
back
actually
being
able
to
look
at
your
1%
percentile,
your
slowest
traffic,
basically
or
your
your
most
error-prone
traffic
and
concentrating
on
fixing
that,
as
opposed
to
concentrating
on
making
your
average
lower,
were
hired
right.
A
Right
right,
yeah
and
like
tracing
like
this,
can
also
tell
you
right
where
you
might
have
Geographic
issues
too
right,
like
let's
say
you
have
a
pocket
of
customers
in
a
region
where
you
don't
necessarily
have
a
great
you
know
fit.
You
know
edge
presence
whatever
you
want
to
call
it.
Yeah
like
this
can
indicate
that
to
you
to
an
extent
right
like
if
you
see
a
lot
of
traffic
coming
in
through,
you
know
like
Tokyo,
and
you
don't
have
a
REIT.
You
know
you
don't
have
a
point
of
presence
in
Japan
anywhere.
A
That
is
like
instant
red
herring
right,
like
not
red
herring,
but
like
something
that
you
need
to
definitely
the
opposite
of
red
herring
right
like
that's
something
that
I
can
address
pretty
easily
in
most
cases
right
like
spin
up
an
instance
or
you
know
a
region
in
Japan
and
off
you
go
right,
like
you'll,
you'll
get
better
user
experience
from
that.
You
know
one
or
five
percent
with
you
know
the
the
effort
it
takes
to
have
what
you
have
now
right
so
and.
C
It's
it's
funny.
You
bring
that
because,
like
I
literally
had
the
issue
so
I've
launched
a
couple
services
in
China
before
and
China
has
got
a
host
of
concerns
that
are
a
little
bit
different
when
it
comes
to
cloud
providers.
One
of
the
main
is
geography
right,
so
so
getting
data
into
a
region.
That's
not
in
the
US
when
you're
a
u.s.
centric
company
is
difficult,
so
Japan's
a
great
example.
C
I
can
talk
about
one
from
my
past,
so
we
in
China
that
the
main
thing
that
we
had
to
do
is
actually
just
set
up
a
CDN
right,
like
it's
inside
edge
compute,
but
a
way
to
get
something
set
up
locally,
and
that's
that's
how
we
found
it
was
through
tracing.
So
we
could.
You
can
see
that
all
these
users
are
having
a
good
time
and
these
10
users,
or
these
10
million
users
are
not
having
a
good
time
what's
going
on
with
these
10
million.
C
A
D
A
It
was
someone's
really
low.
It
was
to
be
loaded
in
one
second,
less
than
one
second
page
load
times
and
like
transaction
times
with
less
than
500
milliseconds,
or
something
like
that.
It
was
something
ridiculously
low
right
and
it's
like
if
you're
getting
to
that
level,
you
have
instrumented
Freight
for
yes,.
A
C
E
C
The
nice
thing
about
tracing
or
observability,
or
even
just
having
these
indicators
in
general
is,
is
knowing
that
you
can
you've
almost
got
a
safety
net,
so
you
know
what
the
baseline
is.
You
can
then
say
like
hey:
let's
start
experimenting:
let's
start
lazy
loading
I
start
putting
our
fonts
in
a
CDN
start,
making
our
images
a
little
bit
smaller
than
they
are.
C
They
don't
need
to
be
4
megabytes,
that
kind
of
stuff
where
you
can
start
running
those
experiments
and
like
blue
green
deployments
and
kubernetes
right
so
openshift,
or
what
have
you
your
canary
deployments
as
well?
So
you
can
start
doing
things
in
DNS
and
pointing
things
to
just
subsets
of
users
and
running
those
experiments
and
comparing
the
two
numbers
and
I
think
that's
for
me.
That's
really
where
the
magic
is
and
I'm
not
a
front-end
person,
but
I've
seen
front-end
people
do
amazing
things
with
blue
area.
D
C
A
Would
be
right,
like
the
wee-wee,
we
put
a
lot
of
effort
into
UX,
the
Red
Hat.
We
see
more
effort
coming
every
day
and
I
know
why?
Now
it's
because
of
stuff
like
this
right,
you
mentioned
like
for
Meg
pictures
like
being
loaded
and
instead
of
you
know
like
optimized
for
the
web
type
stuff.
A
When
the
event
happened,
they
were
sitting
there
with
their.
You
know
digital
SLR,
a
satphone
and
an
FTP
connection
and
just
uploading
picture
after
picture
after
picture
after
picture
into
our
CMS,
and
you
know
the
journalists
and
editors
of
the
newspaper
loved
it,
so
they
just
put
them
all
on
a
web
page.
A
Files
rendered
down
so
that
they
actually
fit
on
a
web
page
but
they're
being
served.
Millions
of
time
like
this
went
like
viral.
The
post
actually
went
viral
because
it
was
like
one
of
the
very
few
people
with
you
know
live
footage
coming
out
of
the
environment
at
that
time.
So
right
like
finding
that
out
very
quickly
hitting
our
you
know,
you
forgot
to
run
the
batch
optimization
process
on
these.
A
It's
like
it's
saved
tens
of
thousands
of
dollars
in
a
day
right,
but
that
was
just
with
basic
billing
alerts.
Right,
like
we
had
crushed
the
billing
threshold.
If
we
would
have
had
something
first
like
this,
we
would
have
seen
this
huge
increase
and
seen
this
huge
anomaly
happen
and
been
like
whoa.
What's
going
on
and
could
have
started
looking
way
early
in
the
process,
potentially
saving
so
having
this
stuff
sooner
rather
than
later,
is
vitally
important
and
I
can
personally
attest
to
that
through
my
s3
bucket
was
lately
so
yeah
yeah.
C
I
think
there
was
something
this
chat
about
billing.
So
when
I
read,
we
can
set
up
billing
alerts
so
we're
not
like
across
management
company
or
anything
like
that,
but
we
do
have
access
to
this
API,
so
you
can
certainly
set
up
here,
I
think
if
you're
already
in
a
co-op
or
buddy,
you
might
have
something
like
that,
like
your
AWS
I
think
has
you
has
like
a
billing
thing
built
in
so
does
Google.
C
So
does
this
your
a
lot
of
people
these
days
are
in
multi
cloud
and
I
realize
that
that's
not
it's,
not
everybody,
but
there
are
people
doing
it
and
setting
up
those
billing
alerts
across
multiples
can
be
a
pain.
So
that's
that's.
Certainly
there
I've
done
it
myself
to
my
trial.
Accounts
I've,
certainly
gone
way
over
budget
Oh,
kubernetes
and
since
I
left
running
on
a
long
weekend.
It's
generally
my
mo
you
set
it
up
on
Friday
and
you
come
back
to
say,
look
how
that
was
yeah.
D
F
Could
set
this
up
not
well
you'd
set
this
up
on
probably
on
my
development
cluster
but
on
multiple
clusters,
and
is
this
going
to
show
me
traffic
flowing
between
clusters
for
that
kind
of
larger
infrastructure
view
and
could
I
have
data
dog
on
on
gke,
plus
some
data
dog
on
openshift,
possibly
both
installed
via
the
operator?
Would
that
make
sense
or
or
is
there
different
installation
paths,
but
still
opportunity
to
view
cross
cluster
traffic?
F
C
All
says
I
should
say
so.
I'm
new
to
this
role,
like
I've,
only
been
a
tech
evangelist
for
six
months.
But
what
I'm
learning
to
say
is
it
depends
I
think
this
is
like
a
common
refrain
so
or
maybe
like
yes
and
so.
I
understand,
I've
not
done
the
multi-cloud
tracing
portion.
That
you're
talking
about
specifically,
but
just
to
talk
about
how
it
works
is
when
things
get
injustice
when
they
first
come
in.
C
C
So
it's
not
like
it
will
overwrite
or
anything
like
that
at
the
very
least,
you'll
be
able
to
see
the
traffic
calling
out
from
one
cluster
into
another
cluster
if
they're
in
the
same
cloud,
no
problem,
oh
so
ever
they're
in
the
same
app
or
what-have-you,
although
those
infrastructure
tags
will
allow
them
to
really
outside
of
them,
though.
So
if
you're
going-
let's
say
from
kubernetes
in
AWS,
to
like
a
Red,
Hat
VM
in
Azure,
for
example,
the
only
thing
that's
going
to
relate
those
two
will
be
that
header.
F
C
D
F
C
Got
it
yeah,
so
it's
it's
a
bit
of
both
right
so
for
us
to
handle
the
volume
that
we've
got
outgoing
we
buffer
a
little
bit
locally,
which
is
good.
So
if
something
goes
down,
you'll
be
able
to
replay
the
buffer,
but
generally,
what
we're
doing
was
we're
streaming
everything
out
as
fast
as
possible
to
today
to
dog.
C
But
if
your
use
is
let's
look
at
all
of
my
communities,
clusters
deployed
locally
or
multiple
clouds,
or
what
have
you?
You
can
certainly
put
those
all
in
one
page,
right,
I.
Think
I
can
probably
demo
that
here
basically
like
what
does
that
look
like
right.
So
how
do
I
see
all
of
the
communities
clusters
together
and
we
can?
We
can
certainly
allow
you
to
do
that
from
all
of
them.
That's.
F
D
C
So
maybe
I
can
steal
if
we've
got
some
time
just
to
steal
an
example
from
my
past
because
of
I've
been
at
like
a
senior
engineer
at
a
pretty
big
company,
so
I
think
when
I'm
working
with
like
dev
team
leads
a
lot
of
what
they
want
to
know
is
how
is
my
app
doing
like
a
cross
region
and
how
is
my
app
doing
across
environment?
So
if
it's
scene
staging,
they
can
come
to
me
and
say:
hey
it's
fast
and
staging.
C
C
So
you
can
see
the
tags
roll
everything
around
here,
but
basically,
when
you
go
into
like
a
service,
you
can
then
allow
you
to
sort
of
drill
down
into
each
of
them
by
just
looking
at
like,
like
only
look
at
this
service
for
this
region
and
I
I
sort
of
showed
it
in
the
communities
one-
and
this
is
a
can
dashboard.
So
you're
not
going
to
see
it
here.
What
if
we
go
back
to
the
kubernetes
one
just
like
for
one
that
you've
made
it
on
your
own
and
I?
C
Don't
I
don't
have
an
example
for
an
app
sorry,
but
basically
these
drop
downs
at
the
top
here
and
it's
a
bit
small,
so
I'll
try
to
zoom
in.
But
these
allow
us
to
say
like
let's
look
at
just
this
one
cluster
or
let's
look
at
just
this
one
deployment
or
this
one
namespace
and
you
can
imagine
for
an
app
like
it
would
be.
Let's
say
the
environment
is
something
like
the
let's
look
at
just
this
one
region
or
if
it's
at
your
environment,
to
something
like
staging.
C
Let's
look
at
only
the
staging
installs
or
let's
look
at
it,
test
staging
and
prod
for
this
one
region.
So
there's
different
ways
of
slicing
it,
but
I
totally
agree.
You
might
not
care
about
the
infra.
You
might
not
care
about
the
containers,
you
might
not
care
about
the
controllers
or
any
of
that
you
just
want
to
see
like
how's
my
app
running,
how
many
instances
my
app
do
I
have.
C
A
I'm
going
back
through
the
questions
here,
so
okay,
so
I'm
gonna,
say
the
these
name
wrong,
but
cease
and
tampura
sorry
does
data
dog
do
you'll,
contribute
open
source,
wise
to
zip,
can
or
open
tracing
or
any
other
projects
like
that?
Oh
this.
C
Zipkin
user
prior
to
coming
to
data
dog
and
there's
a
lot
of
work
in
that
community
for
sure
opens
up
can
I
think
is
specifically
what
we're
talking
about
at
that
point.
What
happened
is
many
people
want
to
do
their
their
own
version
of
this,
so
so
sort
of
two
camps
came
out
of
that
one
of
them
was
I,
think
it's
open,
metrics
and
then
the
other
was
open
census
and
we
basically
we
contributed
to
one
I
think
it
was
all
consensus,
but
we
contributed
sort
of
the
tracing
area
of
the
agent
for
them.
C
What
happened
was
there?
There
was
like
a
I
push
to
bring
the
community
all
together
under
the
branding
of
open
telemetry.
So,
basically
again
the
answer
is
like
yes,
definitely
tracing
is
important
and
what
we
do
in
the
tracing
world
certainly
does
have
an
open
source
component,
we're
for
open
source
champions,
the
digital
most
of
what
we
do
is
open
source.
You
can
see
the
Cofer
agent
on
github
in
cynical
for
like
a
lot
of
what
we
do
for
even
just
our
like
infrastructure
architecture,
that
kind
of
stuff.
C
In
terms
of
like
how
we
set
things
up
automatically
also
conserve
got
a
pretty
big
github.
Repo
definitely
do
poke
around.
But
what
I'm
trying
to
say
here
is
that
in
open
telemetry
we
are
part
of
that
community.
We
do
contribute
back
and
we're
very
interested
in
making
sure
that
you
know
it
keeps
going
forward.
C
So
what
we
want
to
see
is
people
actually
involved
in
the
tracing
world,
not
just
specific
to
us,
but
actually,
you
know
reaping
the
benefits
of
the
community
here,
because
I
think
Zipkin
was
a
big
change
in
the
way
that
we
look
at
apps
and
it
wasn't
I
wasn't
pretty.
It
was
a
little
bit
hard
to
set
up.
There
was
a
lot
of
moving
parts
to
get
it
working,
but
once
it
was
there
I
think
that's
where
the
magic
happened.
C
Once
you
see
these
flame
graphs
of
traces
right
across
all
your
services,
it
it's
a
pretty
it's
a
pretty
powerful
visualization
of.
What's
actually
going
on
your
service,
so
yeah
sweet.
We
definitely
do
contribute
back
to
open
telemetry,
as
well
as
the
slew
of
other
projects
that
one
specifically
is
for
tracing
if
you've
not
seen
it
definitely
look
it
up.
I
think
it's
just
open,
telemetry,
slash,
open
telemetry
is
is
the
main
project
and
in
the
community
you'll
find
people
calling
it
hotels.
C
Right
so
the
way
we
work
in
in
general
is
stats,
D
and
there's
ways
of
multiplexing
stats,
D.
So
let's
say
prior
prior
to
getting
us,
you
could
take
the
same.
It's
for
the
most
part.
It's
it's
like
vanilla,
stats,
D
that
we
said
we
call
it
dog
stats,
D!
C
You
need
it
like
10
20
years,
and
we
give
you
a
certain
amount
in
the
cloud
and
you
want
to
make
sure
you've
got
it.
You
can
certainly
do
that.
So
there
are
a
few
multiplexers
out
there
for
stats.
Dean
I
can't
really
recommend
any
of
them
because
I
haven't
used
them,
but
I've
talked
to
people
at
other
companies
that
do
you
do
it.
They
do
work
and
there's
there's
different
ways
of
setting
up
that.
One
cautionary
note
I
have
here
is
that
there's
a
lot
of
metrics
going
out.
C
It's
not
the
same,
can't
do
it
but
be
prepared
for
like
a
lot.
Basically,
there's
there's
a
lot
of
signal
that
comes
out
of
a
typical
coronated
cluster
and
yes,
you're
being
kind,
listen
I've
done
it
before
I
set
up
300,
node
elasticsearch
cluster
and
seen
it
fail
overnight.
While
I
was
like
in
the
midst
of
needing
a
root
canal.
So,
like
I've
got
the
horror
stories
there
and
I
know
it
can
be
bad,
but
it's
fun
and
it's
it's
definitely.
C
Good
engineering
like
you've
got
the
cycles
if
you've
got
the
cycles
internally
to
look
at
this
or
if
you
wanted
to
figure
out.
What's
going
on
under
the
hood,
definitely
look
at
stats,
T
and
what's
happening
there,
because
that's
really
the
meat
and
potatoes
I
think
that
the
same
person
who
asked
the
question
said
like
we
predate
some
of
this
tracing
stuff
and
basically
the
answer
there
is
that
we
evolved
with
the
community
as
it
came
through,
like
that's
T's,
completely
open
source.
What
we
do
is
also
there
as
well.
C
You
can
see
the
documentation
on
set
like
dog
stats
d
compared
to
stats
T,
but
it's
it's
important
for
us
to
actually
see
what's
going
on
in
the
industry
and
serve
our
customers
at
the
same
time,
while
not
falling
too
far
behind,
because
it
it
changes
really
quickly.
Right,
like
it
feels
like
every
week.
There's
a
new
way
to
monitor
things
or
a
new
thing
to
monitor.
D
A
So
you
could
that,
and
you
could
totally
like
step
D
out
to
spark
if
you
wanted,
but
lots
of
data
be
aware.
Yeah
spark
expects
a
whole
fun
new
ballgame,
yeah,
yeah,
that's
and
I
would
like.
I
would
like
to
have
somebody
come
on
talk
about
spark
and
like
what
you
do
with
it
with
openshift
right
like
in
the
future,
but
maybe
for
this
channel,
or
this
show
you
know,
yeah
I've
been
there
done
that
yeah.
C
A
D
A
C
Yeah,
it's
percs
vikes,
pretty
interesting
I!
Guess
that
the
main
thing
we're
trying
to
say
here
is
that,
like
that
as
a
metrics
come
in
and
as
you
want
to
grab
the
signal
from
the
noise
from
those
metrics,
it
often
means
the
infrastructure,
and
it
often
means
like
being
prepared
for
that.
It
could
be
that
you
just
want
to
first
upset
as
well.
I,
certainly
seen
that
like
not
grab
everything,
but
just
for
this
one
out,
yeah
grab.
C
C
A
Is
yeah
data
dog?
Is
that
ultimate
build
versus
by
question
and
that
I
think
a
lot
of
people
have
to
answer
for
themselves
and
oftentimes.
They
realized
that,
like
there
aren't
like
monitoring,
gurus
anymore,
like
there
used
to
be
right
like
there's
no
one
out
there.
That's
like
a
Nagios
expert
right,
you
know,
is
that's
a
blast,
not
Android.
D
A
If,
if
you're
yeah,
that
is
a
blotch
one
path,
but
if
you're,
if
you're
out
there
as
like
a
monitoring
person
right,
like
that's
a
rare
skill
set,
does
that
make
sense
right,
yeah,
please
make
sure
you're
being
paid
accordingly.
C
D
A
Skill
set
is
an
expertise
that
isn't
like
a
huge
focus
right
right
now
in
the
industry,
for
whatever
reason,
but
we
we
always
say
that
monitoring
is
like
one
of
the
most
important
things.
Yet
we
don't
emphasize
like
the
tooling
to
do
it
kind
of
deal
so
yeah
that
build
versus
my
conversation
is
a
huge
one
but
I
think
I,
don't
know
anyone,
that's
unhappy
with
data
dog
and
when
they
decide
to
use
a
dog
right
like
they.
C
Yes,
I
will
say,
like
I
started
out
with
Nike
and
cacti
with
a
dog,
okay
and
it's
sort
of
silly,
but
what
I
was
doing
was
keeping
my
negiah
stuff
locally
and
using
data
dog
as
sort
of
like
that
cloud.
Backup,
if
you
were
just
again,
it's
not
not
maybe
the
best
use
case,
but
it
works
for
me
excited
swamp,
Ram
I
was
basically
HIPPA
like
I
was
in
a
healthcare
right
area
and
I
couldn't
necessarily
send
everything
to
the
cloud.
C
But
what
I
wanted
was
like
my
events,
my
grass
we
do
integrate
with
Nagios
and
cacti.
There
were
older
integrations,
but
they
certainly
do
so
if
you've
got
those
on
crown-like.
Don't
love
me.
Let
me
laugh
at
that
experience.
Cuz
I've
certainly
been
there.
I
know
that
pain,
writing
pcre
checks
every
day,
all
day
yeah.
But
if
you
do
need
like
a
like
a
safety
net
kind
of
thing,.
F
C
Similar,
it's
almost
like
a
like
a
a
thorough
Venn
diagram
like
there's,
there's
what
we
can
do
and
then
there's
what
Prometheus
can
do
that
being
said,
that
that
small
circles,
pretty
big,
you
can
do
a
lot
with
Prometheus.
C
Do
it
generally
not
something
that
like
it's,
it's
meant
to
have
like
one
Prometheus
and
then,
if
you
want
to
scale,
then
you
figure
out
how
the
multiplex,
at
that
kind
of
thing
what
we
provide
there
is
similar
to
like
the
Negi
O's.
We
can
aggregate
from
multiple
previous
installs
or
if
it's
just
the
prometheus,
install
that
you've
already
got.
We
can
certainly
grab
data
from
there.
So
if
you've
already
got
Prometheus
like
there's,
no
no
real
step
to
like
our
integration
in
communities
is,
is
set
to
work
with
Prometheus.
C
Well
grab
those
from
you
there's
a
couple:
config
changes
you
need
to
do,
but
they're
they're,
basically
there
as
like
a
and
you
say
it's
like
a
default
config,
so
the
endpoints
already
define
is
just
common.
Today,
you'll
uncomment
it
and
you'll
start
grabbing
from
Prometheus
for
you
Prometheus
itself
like,
but
it
it's
sort
of
caught
on
quite
a
bit
right.
It
is
really
popular
in
the
open-source
community.
So
the
onus
is
on
us
to
integrate
with
that
right
as
opposed
to
just
ignoring
it.
Not
it
like.
Definitely,
if
you're
using
prometheus
like.
C
Why
would
we
not
help
you
get
those
metrics
into
the
clouds
we?
We
certainly
do
integrate
to
it,
but
yeah
like
a
I've,
dealt
with
this
to
do
clusters
and
a
lot
of
the
info
that
you
get
from
any
CEO
like
sorry,
initio
cougar
and
is
to
enable
kubernetes
cluster
lots
of
words
in
there
yeah,
but
they
already
send
as
the
metrics
to
that
sync
already.
So
why
not
just
grab
from
there
right?
Why?
Why
set
up
a
whole
new
path,
kind
of
thing,
yeah.
D
C
F
C
Got
it
yeah
so
so,
for
me,
at
least,
prometheus
is
kind
of
like
a
like
a
metrics
database
like
it's
a
really
good
way
of
storing
metrics
or
in
some
cases
like
the
meadow
did
about
traces,
but
maybe
not
necessarily
like
a
way
to
relate
all
of
those
very
easily.
The
UI
I
can
that
there
are
front
ends
for
a
theist
for
a
reason
right.
C
So
some
of
the
value
that
we
provide
there
is
just
a
way
to
correlate
between
all
those
metrics
or
aggregate,
or
look
at
your
percentiles
for
those
metrics,
which
Prometheus
also
has
just
in
a
way
of
looking
at
like
multiple
clusters
or
multiple
installs
or
multiple
different
services
in
just
a
different
way.
A.
C
A
Not
having
a
fun
day
today,
with
moderation,
I'll
just
leave
it
like
leave
it
at
that.
The
the
first
question
you
asked
about
the
pod
scraping
Carlos.
Can
you
ask
that
please,
you
know
I'm,
not
quite
sure
what
you
mean
there,
but
for
service
ryan
mclane.
Would
what
would
we
do
for
service
metrics
there?
You
can't
really
scrape
you
know
you
kind
of
have
to
omit
those.
What
do
you
recommend?
Yes,.
C
In
this
case,
I
must
admit:
I've
I've
not
done
server
listing
kubernetes
I
know
that's,
maybe
a
full
pot
I've
done
service.
I've
done
grenades,
I've
not
done
one
inside
the
other.
That
means
that
I
can
sort
talk
about
how
that
works.
So
in
general,
when
you're
loading
your
libraries
for
whatever
application
you're
running
so
let's
say
it's
Ruby
or
PHP
or
Python,
or
what
have
you
that's
when
you'd
actually
load
in
sort
of
like
the
data
log
library
that
would
emit
out
to
either
like
a
public
endpoint
or
if
its
internal?
C
We
can
talk
about
like
that
dog
stats
de
sort
of
workflow
and
you
would
actually
send
it
into
there.
I've
not
done
server
listing
meeting
metrics
to
Prometheus
again
because
I've
just
not
launched
service
in
kubernetes.
That
being
said,
if
it
already
does
that
it
would
be
the
same
way
now
in
terms
of
tagging,
the
the
way
it's
sent
out
for
us
like
if
you've
tagged
it
in
Prometheus,
we'll
grab
that
tag.
Basically,
if
you
have
not
them,
we
won't
have
the
tag
to
grab.
C
What
we
will
do,
though,
is
things
like
the
cluster
of
the
region.
The
availability
zone,
all
that
stuff
where
Prometheus
is
launched,
will
get,
but
if
there's
anything
specific
to
your
function
like
the
function
name,
no
problem,
that's
metadata
that
we'll
get
if
you
need
to
have
like
a
function
name
and
a
version
pair
together
like
if
it's
function,
joel's
awesome,
function,
v1
kind
of
thing.
C
If
we
want
to
like
tag
the
v1
and
do
that
string
concatenation
with
the
tags,
that's
something
that
you
probably
want
to
do,
as
you
admit,
add
to
Prometheus,
so
that
we
can
scrape
it
from
Prometheus.
For
you
there
you
go
it's
a
long
answer,
but
I
think
that's
how
it'll
work
yeah
now,
there's
a
what
are
the
what's
the
state
of
serverless
in
kubernetes,
cuz,
again,
I
probably
should
know
I
I
know
there
was
Oakland
whisk
at
one
point,
a.
A
C
A
Likewise
yeah
was
it
I
think
it
was
2017.
I
was
sitting
next
to
burst,
shutter
at
all
three
Lawson
and
he's
like
hey
yeah
man,
great
great,
to
have
you
here,
blah
blah
whole
was
before
I
worked
at
Red
Hat
he's
like
yeah
we're
looking
into
open
west
for
a
bunch
of
stuff
and
I
start
looking
at
it
like.
While
he
was
talking,
I
was
like
you're.
D
A
A
C
For
dropping
the
links
from
chat-
oh,
this
is
great,
so
I
I
had
a
team
like
I
helped
deploy
as
your
functions
has
a
library
that
you
can
deploy
into
a
Cass
but
I
honestly
sort
of
turned
a
blind
eye
to
it.
It
ran
and
it
seemed
to
work,
but
I
didn't
I,
didn't
really
know
what's
going
on.
It
was
just
like
that,
like
a
schedule,
tasks
kind
of
thing,
if
you've
ever
done,
those
in
kubernetes
like
once
a
week
run
this
task
or
any
kind
of
thing.
C
C
Yeah,
so
on
one
twitter
is
probably
the
easiest
way
either
that
or
I'm
in
a
bunch
of
the
slacks,
but
I'm
generally
either
myself
on
twitter.
I
I
miss
getting
my
my
vanity
username
swamp
ryan,
underscore
mclean
as
opposed
to
ryan,
mclean
and
poor
thing
yeah.
I
know
you
can't
win
them
all
right,
but
that's
something
into
it.
Otherwise,
I'm
just
Ryan
macLane,
all
in
word,
I'm
I'm
out
there.
Let's
see
if
I
can
stop
this
zoom
here
but
yeah.
Basically
you
can.
You
can
find
me
anywhere.
Ask
me
any
questions.
C
We,
we
didn't
actually
get
to
Joel's
question,
but
that
that
the
main
thing
I
was
gonna
say
it's
just
like
if
you're
in
node
any
want
a
to
dog,
we've
got
talks
for
their
great
docs,
its
one-liner,
its
its
import
data.
Dog,
basically,
is
what
does
all
the
magic
and
everything
is
done
aside
from
the
environmental
variable,
so
apologize
for
not
getting
that
we
got
into
a
bit
of
an
SRE
rant,
which
is
a
favorite
topic
of
money.
Yeah.
A
Sorry
I
might
have
been
my
fault
too.
Certainly
a
guilty
party
here
all
right.
Well,
thank
you,
Ryan
and
Ryan
angel'
for
joining
us
today.
I
really
appreciate
you
coming
on
Ryan,
McLean
and
Ryan
J.
Obviously,
angel
can't
forget
role
he's
what
made
this
all
possible.
So
thank
you
very
much.
Everybody
join
us
here
at
2:00
p.m.
A
Eastern,
1800
UTC,
we're
gonna
have
one
of
my
favorite
interns
here
at
Red
Hat
this
year,
Cedric
he's
gonna
be
doing
like
his
first
six
weeks
with
open
shifts
so
Jason
Toby's
is
gonna,
come
on
on
with
Cedric
and
kind
of
walking
through
what
they've
been
up
to
during
his
internship.
So
it
should
be
pretty
fun
and
I'm
looking
forward
to
it
for
sure
so
yeah
until
then
we'll
catch
you
next
time
thanks
everybody.