►
Description
Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io
Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects
From Issue to PR Merged: A Fluentd “Tail” - Jordan Hamel, Amgen
Do you often find yourself opening an issue or looking for an alternative open-source project with support for your use case? Not sure where to start in contributing a fix for an issue? Getting involved in the Fluentd ecosystem and submitting a PR helped make it possible for Amgen to effortlessly collect CloudTrail logs from hundreds of AWS accounts owned by separate teams. We'll take a look at the details of how to collect and annotate logs stored in any format or account in AWS with Fluentd where hundreds or any number of accounts are in use. We'll also follow the details of contributing this now merged PR to the Fluentd S3 plugin that made it possible. Whether you're a new or long-time user of Fluentd, come and be inspired to consider contributing back to observability related open-source projects like Fluentd and the benefits it can bring to your organization and the community.
https://sched.co/UaYe
A
A
How
many
of
you
are
active
users
of
flu
in
D,
so
majority,
and
how
many
of
you
have
ever
tried
to
contribute
to
flu
and
E
or
the
echo
system?
None
okay,
great.
So
my
goal
at
the
end
of
this
is
to
have
at
least
one
of
you
raise
your
hand
and
say:
oh
I'm,
opening
a
PR
now
I
think
we'll
get
there.
So
let's
talk
a
little
bit
about
Amgen,
so
Amgen,
it's
a
values
based
company,
deeply
rooted
in
science
and
innovation,
to
transform
new
ideas
and
discoveries
into
medicines
for
patients
with
serious
illnesses.
A
Our
mission
is
to
serve
patients.
So
I
highly
encourage
you
it's
you
know
it's
a
very
regulated
industry,
encourage
you
to
go
to
Amgen
comm,
to
see
more
public
information
about
Amgen
Amgen
is
a
global
company
they're
located
in
over
a
hundred
countries,
and
they
definitely
live
by
their
mission
of
helping
the
patients
and
I
think
you'll
get
from
today.
Why
open
source
helps
patients
as
well
and
and
how
the
dots
connecting
between
what
we've
done
with
fluency
and
how
it
continues
to
evolve
as
a
as
a
product
with
being
in
use
at
Amgen?
A
So
let's
get
started.
So
if
you
ever
read
a
lot
of
mythology
and
novels
you'll
hear
a
lot
about
the
the
hero's
journey,
and
sometimes
when
we're
working
in
software
development,
you
feel
like
a
hero
when
you've
failed
as
hard
as
you
could
possibly
fail
and
you're
able
to
then
recover
from
it.
So
it
was
December
November
6th,
2018
I
was
in
a
situation
where
I
had
a
log.
I'd
already
been
working
on
this
for
like
two
days
straight
and
I
had
solved
many
individual
problems
and
it
actually
had
had
the
pull
request
merged.
A
A
The
specified
Q
does
not
exist
for
this
wisdom
version
a
lot
of
stuff
that
at
that
moment
in
in
the
in
the
darkest
hour
of
like
debugging
for
myself,
that
was
very
challenging
because
at
this
point,
I've
already
like
fixed
like
at
least
six
or
seven
different
things
over
the
last
like
48
hours
prior
to
this
moment
and
I
was
at
that
moment
where
it's
kind
of
like
do.
I
want
to
continue
working
on
this.
It's
just
open
source
I
can
find
something
else.
A
That's
very
similar,
but
I
where
we'll
get
to
in
this
story
today
is
how
we
overcame
this
particular
problem,
but
also
the
the
journey
of
getting
through
it
and
I.
Think
it's
important
one
of
the
things
I
hope
you
can
get
out
of
today's
session.
Is
that
there's
a
there's,
an
importance
in
modeling
what
it
takes
to
actually
go
from
opening
your
own
issue
to
then
fixing
your
own
issue
and
all
the
communication
that
goes
on
in
between
that
I.
A
Don't
really
see,
I,
don't
really
see
that
model
very
much
in
a
talk
format,
so
we'll
see
how
it
goes
today
in
may.
In
my
naivety,
I
asked
the
question:
well,
what
wins
diversion
works?
Then
it
says
it
doesn't
work
for
this
was
diversion,
but,
as
you'll
find
out,
that's
not
really
pertinent
to
the
solution.
A
A
Why
contribute
to
fluent
D
and
we'll
go
into
some
really
specific
examples
of
why
I
believe
that
fluent
D
outside
of
kubernetes
I
would
almost
argue
that
fluent
D
is
one
of
the
most
cloud
native
projects.
There
is
and
I
hope
that
you
will
leave
today,
understanding
that
it's
a
lot
more
than
just
a
lager
and
then
we'll
look
a
little
bit
about
biotech
going
cloud
native
and-
and
that
is
a
kind
of
an
oxymoron
for
people
that
have
know
much
about
the
biotech
and
pharma
industry
that
are
very
regulated
and
very
stable.
A
Where
not
a
lot
of
change
is
introduced.
You
probably
haven't
been
to
many
talks
today
that
we're
featuring
biotech
companies-
what's
my
guess
right
so
I
think
that
you'll
be
hopefully
pleasantly
surprised.
So,
let's
get
into
the
issue,
every
good
issue
starts
off
with
a
problem.
What
am
I
trying
to
do
so?
A
What
we
needed
to
do
inside
the
Lewin
d--
was
use
a
plugin
called
the
s3
input
plugin,
it's
one
of
the
original
plugins
that
was
built
for
flow
Indy
a
lot
of
years
ago,
and
what
Amgen
needed
to
do
was
take
all
of
the
AWS
accounts
that
Amgen
uses,
and
we
are
already
writing
off.
The
cloud
trail
logs,
which
are
the
API
events
into
a
single
s3
bucket
standard
AWS
practice
is
to
find
a
way
to
make
all
of
your
accounts,
use
a
similar
trail
and
write
to
a
single
bucket.
A
That
might
be
your
security
account
or
some
other
kind
of
logging
account.
So
in
the
image
you
can
see,
this
four
counts,
but
there's
actually
many
many
many
dozens
more
accounts
than
that,
and
it
so
is
coming
into
a
very
large
bucket
all
tagged
in
the
specific
account
a
that
will
call
it
from
there.
What
we
needed
to
do
was
take
the
s3
bucket,
enable
the
put
events
so
that
therefore
every
event
for
every
log
that
was
delivered
from
Cloud
trail
to
a
single
bucket
across
hundreds
of
accounts.
A
A
No,
it's
not
it's
an
object
store,
so
you
have
to
know
where
your
objects
are
being
stored
on
poun
delivery.
So
it
writes
that
s3
key
into
an
sqs
queue
for
all
the
accounts,
because
we're
using
a
shared
bucket.
So
whether
you
have
one
or
you
have
a
thousand
accounts,
it
would
work
the
same.
So
in
our
particular
situation,
the
account
be
that's
separate
from
where
the
bucket
was
in
a
secure
environment
was
where
our
foodie
was
running,
where
elasticsearch
was
running
and
also
where
s
qsq
is
located
from
here.
A
Basically,
we
wanted
foon
D
to
assume
a
role
into
account.
Access
the
bucket
read
the
sq
sq
in
account
B
to
read
all
the
object.
Keys,
get
the
GZ
file.
Do
the
decompression.
All
this
is
in
the
documentation
for
the
configuration
of
the
s3
input
plugin.
Then
we
wanted
to
be
able
to
write
that
to
our
Kafka
queue
for
various
reasons,
because
we
may
want
to
use
this
information
for
security
we
may
want
to
use
it
for
operations.
A
Cloud
trail
is
is
kind
of
like
AWS
is
API
logging
right,
it's
it's
events
that
are
specific
to
the
cloud
platform
that
you're
using
and
then
we
use
fluent
D
to
consume
that
topic
and
then
deliver
it
out
to
elasticsearch,
which
could
then
be
consumed
by
our
users
using
something
like
core
fauna
or
Cabana,
and
they
may
query
something
in
elasticsearch.
Hopefully,
they'll
come
to
an
end
result
like
this,
so
we
can
see
something
about
the
dates
of
the
the
record.
We
can
see
things
like
s3
bucket.
Where
did
it
come
from?
A
A
D
agents
writing
to
Kafka,
but
only
maybe
10
actually
consuming
the
information
and
writing
it
out
to
elasticsearch
so
fluid
plug
in
s3
issued
number
247
and
the
pull
request.
248
tried
to
use
the
I,
am
user
key
and
a
local
account
queue
and
the
whole
process
worked.
So
everything
that
I
wanted
to
do
in
the
previous
screen
worked
when
I
was
using
an
isolated
account
and
in
AWS
using
an
IM
user
and
key.
I
tried
using
the
assumed
role
feature
and
it
failed
with
an
arrow.
A
That's
saying
that
no
region
was
set
so
I'll
take
a
look
at,
and
this
is
this
is
the
actual
stuff
here
I
wrote
my
comment
as
a
new
issue
notice.
No
one
was
assigned
the
issue.
No
one
commented
on
the
issue.
I
was
alone
in
the
vast
internet
that
we
all
love
very
common
case,
but
I
tried
to
produce
like
what
my
configuration
looked
like,
but
basically
I
had
something
completely
wrong.
A
Of
course,
it
was
on
my
side,
not
the
product
side,
so
I
realized
that
I
needed
to
attach
the
assumed
role
to
the
instance
where
ec2
was
running
flew
indeed,
so
after
I
fixed
that,
then
I
realized
that
there
was
a
different
issue,
that
I
was
still
getting
I
couldn't
use
the
s3
key
or
the
the
I
am
key
to
assume.
The
role
had
to
use
the
instance
to
assume
the
role
which
that
worked,
but
I'm
still
getting
the
region
not
set.
A
So,
a
little
bit
later,
I
defined
out
that
this
issue
was
solved
four
years
prior
on
the
s3
output,
plug-in
and
so
I
highly
recommend,
as
something
will
kind
of
summarize
later
highly
recommend,
always
search
in
projects
for
their
closed
issues
to
see.
If
there's
some
hidden
gems
that'll
help
you
on
new
issues,
it
can
really
be
powerful.
So
there
was
a
fix
there
from
2015
and
that
was
back
in
pull
requests
number
122
so
which
basically
allows
you
if
you're
looking
at
it,
that
out
the
out
put
plugin
for
ruby
here.
A
A
Everything's
closed
great,
a
pull
request
for
122
fixes
out,
but
I
needed
to
go
ahead
and
do
it
for
mine,
use
the
exact
same
code,
everything's
great,
except
for
the
DCO
bot,
if
you're
not
familiar
with
DCO
bot,
that's
basically
the
developer
certificate
of
origin,
which
ensures
that
any
project
you
commit
to
has
been
signed
off
in
some
way.
The
origin
of
this
is
actually
way
before
github,
but
it's
the
linux
foundation's
way
of
basically
stating
that
you,
you
did
the
work
that
you
said
you
did
and
you
have
to
sign
it.
A
A
So.
The
exact
code
that
worked
to
fix
to
set
the
s3
region
is
here
and
now
we
open
up
ish
the
pull
request,
number
249
and
issue
250,
so
I
thought
my
pull
request
was
good
to
go
and,
of
course,
friendly
Cosmo
0,
9
2
Oh
reviewed
the
pull
request
and
realized
that
I
forgot
my
to
space
indentation.
So
very
common
newbie
mistake
in
some
ways
you
could
say,
but
the
friendly
enough
to
leave
the
message.
A
I
fixed
it
added
another,
commit
to
the
pull
request
and
I
was
good
to
go
so
then,
when
I
went
to
actually
test
it
I
noticed
that,
with
the
the
a
Tobias
key
and
secret
key
when
it
posts
to
the
sq
sq,
the
message
looked
correct,
where
you'd
have
an
account
in
the
queue
named
because
I've
already
set
that.
But
when
I
was
using
assumed
world
credentials,
it
was
posting
in
to
nothing
and
there
were
nothing.
A
There
was
nothing
there,
and
so
it
wasn't
working
as
proper
as
it
should
properly
and
I
didn't
want
to
have
to
use
an
AWS
I
am
user
and
key
because
that's
actually
not
the
best
security
practice.
If
you
talk
to
your
AWS
account
people
they'll
tell
you
the
same.
So
I
opened
up
a
new
issue
with
some
more
details
on
the
problem,
which
takes
me
back
to
my
original
problem
here
about
the
sqs
name,
s
qsq
name
not
being
added
to
the
uri
properly.
A
This
I
can
see
it's
not
adding
the
cue
name
something's
wrong
with
it
is
why
it's
working
fine
when
I
use
it
in
my
own
account
it's
working
fine
when
I'm
using
it
with
and
I
am
user,
but
something
else
is
missing,
and
this
is
at
the
point
of
frustration
where
you're
just
like
alright
I
could
probably
do
this
with
logs.
You
could
probably
do
this
with,
like
you
know,
a
million
other
tools
why
why
continue
on?
This
is
already
on
the
6.
A
So
it's
like
two
days
and
my
full-time
job
is
not
to
work
on
this
specific
issue.
So
I
did
what
everyone
ought
to
do
and
go
ahead
and
Google
it.
The
specified
queue
does
not
exist
for
this
whizzle
version
unfortunately
had
50,000
results.
So
it's
not
exactly
an
uncommon
issue.
If
anyone
has
ever
faced
this
issue,
you'll
get
a
lot
of
interesting
advice,
but
the
spoiler
alert.
It's
actually
not.
The
wisdom
version
is
problem.
A
So
one
of
the
things
I
learned
about
debugging
in
the
open
source
world
is
that
you
really
know
a
lot
more
than
you
think.
You
know
if
you
just
go
back
and
look
at
the
source
code.
What
Lloyd
I
can
see
here
from
the
s3
input
plug
in
in
in
this
is
that
it's
using
the
AWS
SDK
for
s3,
the
SQS
and
SQS
q
polar
so
I
start
to
take
a
look
at
where
I
probably
should
have
started
at
the
very
beginning,
which
is
at
the
AWS
documentation
for
Ruby.
A
So
Atos
sqs
client
in
ruby
has
a
lot
of
details
about
the
parameters
you
can
set
things
like
region
and
credentials
and
something
interesting
here.
A
queue
owner,
AWS
account
ID.
So
it
says
to
access
a
queue
that
belongs
to
another.
Aws
account
use
the
queue
owner
database,
account
ID,
parameter
to
specify
the
account
ID
of
the
queues
owner.
So
what
did
I
do?
A
I
go
back
to
flu
and
DS
configuration
I,
add
the
parameter
and
of
course
it
doesn't
work
because
it's
not
in
the
code
right
so
well,
eventually
that
what
this
is
when
you
create
your
your
client
with
sqs
you're
pulling
with
all
the
options,
you
have
the
ability
to
not
only
assume
a
role,
that's
attached
to
your
flu.
Indeed,
in
running
instance,
you
can
then
assume
another
role
where
the
bucket
exists,
which
is
actually
owned
by
someone
else,
account
a
and
account
B.
A
So
if
you're
following
this,
it's
because
we
have
security
over
here
who
needs
control
the
bucket.
We
have
a
different
team
that
owns
the
sqs
queue
and
also
pays
for
that
sqs
queued
in
a
different
budget,
and
we
have
to
kind
of
work
them
out
together.
So
if
we
look
a
little
bit
closer,
the
basically
we
see
a
queue
name
which
we
have
already
in
the
s3
input
plug-in.
But
there's
this
new
parameter
Q
on
or
a
2's
account
ID.
So
it
gives
me
an
idea:
let's
go
ahead
and
take
a
look
at
what's
there.
A
This
is
what
is
there
inside
the
s3
plug-in
at
the
time
that
I
did
the
pull
request
and
it
basically
says:
ok,
here's
your
response,
create
the
sks
queue
as
client
get
get
URL
the
queue
name,
but
no
account
owner
for
that
queue.
So
we
take
a
look
over
here
and
in
the
configuration
parameters,
I,
basically
just
added
an
additional
string
queue
on
our
a2v
account
ID,
with
a
default
of
nil,
it's
optional
and
then
added
that
into
the
parameters
for
the
response
of
the
actual
code.
A
So
it
looks
like
it's
probably
going
to
work
so
I
make
a
comment
and
say
modified
to
handle
cross
account
ID
that
owns
the
s
USQ
when
the
queue
URL
works
per
this
documentation.
It's
in
the
pull
request
now
well,
who
pol
is
ready
to
go
right?
Not
quite
there's
a
person
named
Travis
CI
who's,
always
ready
to
deny
you
and
Travis.
Ci
checks
were
unsuccessful.
My
D
Co
was
working,
but
the
integration
test
didn't
work,
and
so,
when
you're
contributing
always
make
sure
you
check
for
the
project's
integration
tests
that
are
there.
A
I
think
you'll
save
a
lot
of
time
in
understanding
what
mock
tests
are
expected,
so
that,
if
you
make
contributions,
you're
really
like
you're,
already
plugged
into
the
exact
process
that
the
maintainer
czar
using
because
you're
not
going
to
you,
may
have
to
write
your
own
test,
but
in
a
case
like
this,
where
it's
just
a
small
patch
you're,
actually
just
modifying
their
mock
tests
directly.
So
in
fact
it
failed
miserably.
It
failed
all
potential
tests
that
could
have
failed,
including
the
allowed
failures.
A
So
at
this
point
it's
kind
of
like
one
of
those
things
where
I
guess
the
next
thing
I
can
do
is
just
look
at
the
results
of
the
Travis
CI
tests
right
inside
of
here.
Unfortunately,
there's
like
800
lines
to
go
through
and
at
the
very
bottom
I
see
something
about
the
failure
of
the
gzip
multiple
streams
for
the
input
test.
Failing
so
on
that
multiple
streams,
it's
basically
doing
a
mock
test
to
see.
Can
you
get
the
gzip
file
where
all
the
cloud
trail
is
being
stored
in
the
s3
bucket?
A
But
thankfully
the
message
is
helpful
and
says:
unexpected
method
invocation
the
exact
method
that
I
just
added
to
the
plug-in.
So
that
gives
me
an
idea
of
where
to
look.
So
essentially,
you
know
at
this
point
before
I
get
to
look
there
sometimes
you'll
be
feeling
this
like,
as
a
developer,
is
kind
of
like
a
shock
that
your
stuff
doesn't
work
on
the
first
try
and
even
the
second
try
or
third
try
or
even
like
you
know.
This
is
a
35
minute
talk.
So
we're
compressing
a
lot
of
time
here.
A
You
you,
after
the
aftershock
generally
comes
sadness
and
that's
kind
of
like
where
I
want
to
encourage
you
to
continue
on
in
the
projects
that
you
use
like
within
the
cloud
native
echo
system,
there's
actually
a
lot
of
great
communities.
We're
gonna
talk
a
little
bit
later
about
like
the
community
of
fluency
and
kind
of
my
interactions
with
it
and
how
positive
it's
been
in
order
to
try
to
get
help
and
try
to
like
figure
out
like
how
you're
gonna
get
beyond
the
solution.
A
So
basically
I
finally
found
the
fix
and
I
had
to
go
into
the
test.
Ruby
file
and
notice
that,
in
the
test
we
basically
stub,
our
own
AWS
account
ID
one
two,
three,
four:
five:
six,
seven,
eight
nine
one,
two
three
that
matches
the
size
of
an
a
DBS
account
ID
string
and
then
add
that
to
the
test,
specifically
so
that
it
can
pass
its
tests
and
it
was
finally
fixed.
So
always
remember
that
you
don't
work
for
Travis
CI
Travis
CI
works
for
you
never
feel
like
it's
like
a
blocker.
A
It
should
actually
be
something
that
protects.
You
protects
the
project
from
getting
bad
code
in
and
I
think
that
that's
like
a
really
valuable
thing
to
learn
if
you're
just
getting
started
in
contributing-
and
so
it's
time
now
to
kind
of
go
on
to
the
the
pull
request
summary
here.
When
you
really
look
at
this,
the
the
the
two
days
that
I
spent
working
on
this
particular
issue
that
I
basically
made
for
myself,
because
it
helped
our
use
case.
A
Essentially,
there's
there's
just
eight
lines
of
code,
so
it's
eight
lines
of
Ruby
code
which,
when
I
calculated
it
that's
like
one
line
of
code
for
every
year,
the
Masahiro
has
been
working
on
mundi.
So
that's
pretty
impressive
like
it,
it
doesn't
take
much
code
actually
to
make
a
project,
that's
as
mature
as
flu,
indeed
to
work
for
you
in
a
new
way.
That's
a
really
important
lesson
that
I
think
is
would
apply
to
other
projects
that
you
may
work
on.
So
eventually
we
get
to
the
eight
lines.
A
We
add
the
description
for
the
config
parameter.
We
modified
the
response
to
when
it
regenerates
the
sqs
client
to
include
the
extra
parameter
of
who
owns
the
SQ
SQ
and
then
at
our
region
test,
which
is
the
fix
from
the
prior
region
issue.
And
then
we
add
in
our
testing
to
allow
for
a
fake
dummy
account
ID
for
AWS
and
allow
the
the
response
to
work
properly
and
we
get
to
our
favorite
part
repeatedly
says
merged
thanks,
which
is
like
a
great
feeling.
A
A
It
actually
works,
and
that
realization,
like
might
be
really
obvious
to
you
if
you've
worked
in
open-source
for
a
long
time,
but
I
think
it's
not
that
obvious
for
companies
like
Amgen
and
other
large
enterprises
that
don't
have
as
much
presence
in
open-source
contribution.
So
I
think
this
there's
definitely
some
things.
We
should
go
over
about
like
why
this
fix
matters.
So,
first
of
all,
just
to
talk
about
from
ambiens
perspective,
Amgen
squad
trail
data
is
now
fully
like
accessible
from
all
of
these
accounts
and
the
nativists
deliveries
around
like
five
minutes.
A
So
it's
still
not
quite
like
as
real-time
as
if
you
had
the
logs
delivered
in.
You
know,
maybe
from
the
console
I
I
forget,
even
if
that
is
a
five-minute
delivery,
but
it's
basically
if
I'm
in
at
SLA.
The
pipeline
accom
accommodates
now
are
kind
of
odd
security
structure.
Where
accounts
over
here
are
kind
of
part
of
the
core
accounts,
and
then
we
have
like
operational
accounts
and
application
accounts.
A
Now,
for
anything,
we
want
to
do
with
s3
I
can
have
cross
account
access,
which
is
really
key
for
teams
that
work
in
as
the
company
the
size
of
Amgen
there's,
not
one
development
team
working
on
one
revenue
generating
product
right.
So,
if
you
think
about
the
complexity
of
companies
like
Amgen,
Amgen
has
well
over,
you
know
it's
well
over
1500,
like
mainline
in
production
applications
and
really
closer
to
about
4,000.
When
you
go
beyond
that
and
those
are
all
sitting
in
different
types
of
8os
accounts
or
on-premise,
so
this
actually
helps
out
a
lot.
A
And
then
we
talk
about
like
the
different
plugins
that
fluidy
has,
and
this
is
really
where
the
magics
at
of
why
the
fixed
matters
so
I
can
use
things
like
record
transformer.
The
rewrite
tag
filter
the
exec
out
plug-in
to
manipulate
any
record
of
cloud
trail
for
specific
use
cases
by
by
the
account
the
region,
the
content
already
other
retention
needs
because
now
I
have
all
the
details.
I
have
s3
key.
A
If
you
remember
the
previous
screen,
I
have
the
s3
key,
which
gives
me
the
entire
object,
so
I
know
the
account
the
region,
maybe
I,
have
special
rules
for
Europe
or
maybe
I
have
special
rules
for
Asia
that
I
want
to
track
certain
kinds
of
data
for
different
regions
and
I
want
to
track
our
a
to
base
usage
in
those
regions
in
a
different
way.
I'm
able
to
do
it
really
easily
with
the
plugins
that
we
have
so
in
taking
a
look
at
like
an
example
of
a
config
of
how
this
works.
A
If
you,
if
you're
looking
at
like
the
filter,
plug-in
you
kind
of
use
like
star
star
to
match
everything,
we
use
the
record
transformer
to
add
a
tag
of
cloud
trail
to
everything
and
that
works
fine.
That
tag
will
be
like
a
key
value
pair
inside
of
elasticsearch
if
that's
the
final
destination
and
then
I
want
to
match
everything.
A
That's
tagged
as
cloud
trail
and
use
a
rewrite
tag,
filter
to
take
the
the
value
of
the
s3
key
and
look
for
the
account
number
so
now
I
can
tag
everything
with
the
actual
account
ID,
which
is
really
helpful
for
search.
Query
optimization
so
now
in
the
example
result.
Instead
of
account
that
cloud
trail
it'll
be
like
zero
one,
zero
to
nine
dot
cloud
trail.
A
The
the
the
other
interesting
thing,
though,
is
that
you
know
flu
and
E
has
a
lot
of
built-in
plugins
and
one
of
the
built-in
plugins
that
I
think
is
the
most
underrated
plug-in.
Is
the
exec
out
plugin,
so
this
this
is
actually
we'll
get
to
that
right.
First
time,
init.
What
I'm
going
to
do
is
I'm
going
to
grep
the
record
for
an
event,
something
in
AWS
that
I'm
not
really
like
wanting
to
have
happen
like
somebody
creates
an
IM
user.
A
So
I'm
really
concerned
with
like
the
europe
region,
and
then
I
can
tag
that
as
an
IM
user
create
so
now
from
a
security
perspective.
I
have
all
the
events
on
a
searchable
field,
with
the
tag
of
IM
user
create
just
from
the
region
uswest
and
that's
now
just
another
filter
block
within
my
fluid
e
config.
So
I
could
send
that
data
anywhere
and
do
different
things
with
it.
So,
for
example,
like
maybe
I
actually
want
to
delete
that
user
right.
A
Somebody
went
through
all
the
effort
of
creating
that
user,
but
actually
want
to
delete
it
right
away
because
it's
a
security
risk.
So
what
you
can
do
is
run
your
own
code
like
you
can
have
Ruby
here,
executed,
command,
delete,
I,
am
user,
dot,
RB
and
then
it
will
actually
because
you've
already
filtered
down
on
the
tag.
I
am
user,
create
eight
that
actually
passes
the
entire
value
of
the
s3
key
and
the
message
so,
where
you
see
keys
under
the
type
JSON
it'll
pass
those
as
like
data
to
the
code.
A
A
So
let's
go
on
a
little
bit.
One
more
one
more
benefit
again
was
about
this
the
scale
this
will
scale
with
the
number
of
accounts
that
you
have
and
if
you're
worried
about,
like
your
SQ
SQ
getting
too
big
that
actually
scales
as
well.
This
is
an
example
of
like
when
we
shut
off
the
flu
in
d,
the
r
SQ
s
missed
messages,
visible,
went
up
to
like
sixty
million
and
that's
a
lot
of
messages
for
a
lot
of
different
ad
bus
accounts.
A
But
all
I
have
to
do
is
spin
up
more
containers,
so
it
scales.
So
you
can.
You
can
spin
up
like
25
instances
of
flu
in
d
on
a
single
docker
host
and
in
like
a
couple
of
hours,
we
can
see
this
kind
of
like
down
into
the
right
motion
here
of
like
every
message
that
was
visible
in
the
queue
is
now
down
to
0
and
all
the
messages
received.
It's
like
up
to
250,000
during
that
time
per
per
moment,
the
so
to
speak.
A
It's
not
at
a
granular
level
here,
but
basically
it
keeps
up
with
as
large
of
your
queues.
You
have
and
that's
up
to
you
and
how
you
want
to
manage
it.
So
some
issue
PR
and
pull
request
takeaways
that
I
think
are
important
to
kind
of
go
over
stay,
curious
and
persistent.
I
think
that's
a
really
important
thing
to
do.
A
Learn
to
debug
and
troubleshoot
publicly.
You
know
plan
out
what
you
want
to
achieve
before
you
start
check
the
contribution
guide.
If
it's
available,
you're
learning
to
troubleshoot
in
public
is
like
hard
for
people
that
are
kind
of
shy.
They
don't
want
the
reveal
their
identity.
They
may
not
be
able
to
work
with
other
people.
I
think
that's
a
really
important
thing
to
do
and
again
search
the
close
PRS
before
you
open
them
overcome
your
impostor
syndrome
by
just
getting
the
code
out
there.
A
Your
contribution
matters,
whether
it's
on
documentation
or
actual
code
and
and
the
flu
indie
community,
is
very
welcoming
I.
Think
it's
highly
recommended
to
work
in
inclusive
communities.
Just
remember
to
thank
people.
You
know
these
people
are
dedicating
their
time.
They
may
be
paid
for
it.
They
may
not
be,
but
it's
never
hurts
to
be
kind.
A
So
we've
got
a
lot
more
to
go
through
and
let's
continue
on
to
a
little
bit
about
why
I
can
contribute
to
fluency,
specifically
and
I.
Think
these
are
some
of
the
lessons
that
are
important
flow
D
advertises
itself
as
a
unified
logging
layer.
You
can
take
data
from
here
and
put
it
over
here.
You
can
transform
the
data.
You
can
mutate
the
data,
it's
simple!
It's
easy!
It's
open
source
proven
reliability
again.
A
This
project
came
out
in
2011,
I
was
still
working
for
an
e-commerce
company
called
newegg.com
during
that
time,
and
we
were
already
using
Apache
flume.
But
looking
at
the
project
like
it's,
it's
a
really
long
time
in
internet
years
of
a
project
that
is
now
part
of
the
cloud
native
computing
foundation
and
has
a
proven
community
to
do
it,
I
think
it's
yeah,
while
we're
yeah,
so
I
want
to
talk
a
little
bit
about
this
part.
A
So
when
you
think
about
the
plug-in
ecosystem,
I
I,
like
to
think
of
the
analogy
of
like
music
and
and
and
deep
cuts
so
like,
if
you're
a
fan
of
any
particular
music
musician
or
rock
band,
some
of
like
the
the
music
that
is
a
little
bit
harder
to
find
for
that
particular
artist
can
be
considered
like
a
deep
cut,
because
it's
it's
really
good.
But
it's
not
a
lot
of
people
know
about
it
and
I
think
that's
the
way
that
the
the
plug-in
ecosystem
of
Fundy
is.
A
It
is
right
now,
there's
a
plug-in
for
serial
port
which
actually
does
exactly
what
you
think
it
does
if
your
company's
not
using
the
Internet
of
Things-
and
you
just
have
a
bunch
of
things
and
you
can
plug
it
in
the
serial
port,
you
can
communicate
to
it
JQ
record
transformer:
rewrite
stats,
notifier,
unique
counter,
validate
tag
filter.
All
of
this
can
be
used
to
create
a
secure
pipeline
so
that
you
can
get
any
data
out
of
the
your
pipeline
before
you
go
and
execute
other
things.
The
the
message
queues.
A
They
support
Kafka
MQTT,
a
MQ,
Kinesis,
sqs,
Google
cloud,
pub/sub
databases,
click
house,
time
scale
in
Flex,
DB
scribe,
which
is
really
interesting.
That's
like
an
open
source
database
from
Facebook.
This
is
like
the
original
plugin
for
the
first
version
of
that
database
that
they
released,
which
they
re-released
earlier
this
month,
level
DB
Vertica,
which
is
kind
of
HDFS
based
and
then
they
even
supported,
like
a
data
plug-in
to
take
things
from
collect,
D
and
stored
into
influx
DB,
like
is
like
a
pass
me
off
kind
of
like
thing
messaging.
A
If
you
need
notifications
there,
it's
all
there
in
the
cloud
each
one
of
these
cloud
platforms.
Ollie.
U
n--
as
your
AWS
GCP
OpenStack,
that's
the
count
of
the
number
of
plugins
for
each
cloud
platform.
That's
a
lot
and
so
use
really
a
deep
rich
integration,
if
you're
in
a
hybrid
environment
or
if
you're
in
a
cloud
native
environment.
Only
and
that's
why
I
say
it's
kind
of
like
the
like
the
cloud
native
project
outside
of
kubernetes.
A
If
you
think
about
like
this
in
in
terms
of
like
where
this
sits
in
the
ecosystem
projects
like
Telegraph,
collect
D
log
stash,
they
all
have
a
lot
of
plugins
but
flow
India's
like
over
650
and
that
even
that
is
kind
of
being
humble,
because
if
you
look
on
Ruby
you'll
find
over
990
gems
that
are
labeled
flow,
Indy
plugin
and
that's
that's-
that's
saying
a
lot
because
it
has
a
package
manager
great
community
way
to
build
out.
Your
echo
system
is
by
using
a
package
manager.
A
So
this
is
one
of
the
like
the
original
logos
that
funda
uses
to
kind
of
describe
like
what
it
does
and
what
happens
when
you
have
it
afterwards
and
I
want
to
take
a
minute
to
do
a
quick
demo
for
us
today.
So
what
I've
get
basically
running
is:
let's
we're
gonna
tail
this
file
for
some
results
and
then
I'm
going
to
send
things
to
the
HTTP
input
that
are
kind
of
like
events
that
are
happening.
A
So
this
way
you
can
see
some
example
of
like
how
this
works
so
I'm
going
to
talk
about
an
event
like
cube
con.
If
I
send
this
to
HTTP
phone
t9,
88,
that's
actually
running
locally
and
the
event
is
cube,
con
itself,
I
can
send
it
and
then
I
can
take
a
look
and
see
that
okay,
it's
cube
con.
It's
called
a
native
con
2,
so
there's
some
event
there
and
I'll
talk
to
you
about
like
where
this
is
going.
This
is
not
going
directly
to
the
file.
A
This
is
actually
calling
out
to
another
API
you'll
see
the
details
in
a
minute
there's
a
few
other
events
that
may
be
going
on
here
this
week
that
you're
doing
like
cube
con
coffee.
Everybody
needs
coffee,
anyone
else,
coffee
fan.
If
your
coffee
time,
you
know
the
open,
Java
bar
for
everyone,
that's
a
minor
joke
on
open
JDK,
the
cube
con
after-party.
A
You
don't
want
to
miss
that
because
we're
in
San
Diego,
so
you
may
want
to
get
your
tacos
and
it's
free
and
if
you
start
thinking
about
like
where
else
this
could
go,
let's
think
about
like
what?
If
what,
if
you
have
like
something
real
that
you
want
to
actually
like
make
a
make
like
a
statement
on
here
if
I
want
to
post
like
an
actual
air
like
say
there
was
a
problem
with
cube
control
or
something
like
that,
I
can
actually
send
a
message
and
then
let's
do
it
cannot
post
there.
A
Let's
do
this,
so
let's
say
we
have
a
cube
control
error.
Now
I
can
actually
execute
and
send
a
message
back
saying.
The
cube,
cube,
control,
delete
F
recent
deployment
llamo
ran
successfully
in
one
second
300
milliseconds.
So
what's
that
actually
doing
your
piping
things
together?
So
that
actually,
is
you
take
that
error
event?
You
do
something
with
it,
because
what
are
you
going
to
do?
Send
it
to
elasticsearch
and
wait
for
somebody
to
query
that
and
then
take
some
action.
A
You
can
plug
all
this
in
together
and
if
you--if,
we
want
to
take
a
quick
look
at
like
what
this
is
doing.
This
is
actually
like.
If
I
look
at
my
fluent
D
config,
this
is
actually
matching
the
event
type
and
then
running
executions
on
a
simple
bash
script,
which
is
then
calling
another
API
to
return
the
results
back
to
a
file,
which
is
what
I'm
tailing
here.
A
So
if
you
start
like
like
really
putting
this
in
your
mind,
you
can
control
anything
in
your
cloud
native
ecosystem
of
fluent
D,
and
that's
like
imagine,
you
like
you,
like
Hubb,
has
their
actions.
Now.
If
you
have
a
log
file
on
your
gig
version
control,
you
can
create
your
own
actions
very
easily
using
this
kind
of
thing.
So
it's
really
powerful.
A
A
You
can
handle
all
these
things
that
you
want
to
do
and,
if
you
think
about
like
companies
like
Afghan
Amex,
silastic,
splints
data,
dog,
New,
Relic,
you're,
fond
of
Loki
even
honeycomb,
all
of
these
are
like
actually
have
plug-ins
for
flu
and
E,
and
it
you
know,
if
you
look
close
enough
on
the
flu
in
D
bird,
it's
actually
smiling
knowing
that
they're
all
under
its
wing,
so
I
think
it's.
It's
really
a
testament
to
the
like
the
reliability
of
it
and
why
I
say
it's
built
for
beyond
observability.
A
A
You
don't
really
need
to
be
a
scientist
to
know
that
drug
data
that
is
blindage
not
be
given
to
the
unblinded
people
at
the
wrong
time
and
flu
India
is
a
great
project
to
use
that
in
its
open
source,
there's
a
lot
of
value
in
the
enterprise
open
source
and
user
growth.
That's
growing
and
I
think
I
expect
to
see
this
growth
continue.
It's
a
talent
and
career
thing.
A
If
you
can
take
a
joint
effort
of
people
that
are
working
on
the
same
problem,
space
and
you're
really
going
to
thank
yourself
in
the
future
when
you
move
to
another
company-
and
you
can
still
continue
to
work
on
the
project
that
you
are
working
on
at
a
company
that
you
left.
That
can
be
really
interesting
and
I.
A
Like
you
know,
what
are
the
pros
and
cons
of
using
it
and
the
performance
metrics-
that's
gonna
be
really
hard
to
find
in
like
a
commercial
product
because
they
need
to
sell
you
on
the
idea
of
it,
whereas
felides
really
a
down-to-earth
community
and
as
a
career
path.
I'm
someone
who's
actually
gone
to
management
gone
back
to
as
an
individual
contributor,
and
now
I'm
back
is
management
again.
You
can
actually
contribute
to
open
source
projects
without
getting
in
the
way
of
your
direct
reports.
A
Does
anyone
hear
a
manager,
anyone
here
manager
of
managers
yeah,
I'm
not
quite.
This
is
really
a
good
way
to
stay
technical
by
contributing
to
projects
that
aren't
like
a
direct
dependency
inside
of
the
teams
that
you're
with,
if
you
think
about
like
the
number
of
software
employees
and
employees
that
these
companies,
this
is
just
like
a
really
generous
estimate
of
the
number
of
software
engineers
at
these
companies
and
if
you
look
at
like
fortune,
500
there's
like
28
million
employees
and
approximately
4
million
developers
at
those
companies,
the
trend
is
it's.
A
The
trains
already
left.
It's
only
a
matter
of
time
before
companies
like
Amgen
and
other
fortune,
500
companies
continue
to
get
integrated
with
CN
CF
projects
and
I.
Think
that
there's
a
real
like
high
value
in
speeding
up
the
delivery
of
a
product
as
a
project,
because
that's
really
the
real
reality
is
that
projects
and
products
the
lines
are
blurring
it
like
what
they
are
so
just
to
kind
of
end.
It
today,
I
wanted
to
talk
quickly
a
little
bit
about
like
why
I
am
John
and
why
this
is
really
important.
A
The
FDA
really
likes
traceability
and
we
reproduce
ability
that
the
cloud
gives
you
the
audit
trails,
the
everybody
in
the
whole
industry,
including
patients,
wants
therapies
at
a
lower
cost.
What
does
it
take?
It
takes
the
ability
to
get
through
the
discovery
to
the
production
in
a
much
faster
way
and
I
believe
the
cloud
native
projects
can
help.
You
do
that
as
a
statement
that
you'll
find
on
our
website
at
Amgen.
A
We
discovered,
develop
and
manufacture
medicines
to
enable
more
moments
in
the
lives
of
more
people,
moments
of
joy,
moments
of
friendship,
moments
of
life
for
moments
like
these,
we
innovate
and
we
work
bold
lessly,
boldly
tirelessly
and
passionately
to
serve
the
ones
that
matter.
The
patients
and
I
think
that
Amgen
is
committed
to
doing
that
in
their
technical
fields
and
also
in
their
in
the
the
drugs
that
they
produce.
So
with
that
I'd
like
to
thank
the
whole
community,
and
thank
you
for
coming
to
today's
please
email
kubernetes
at
amazon.com.
A
If
you're
interested
in
building
meaningful
products
at
Amgen,
I'll
be
here
for
the
rest
of
the
week,
as
well
as
another
colleague
of
mine,
we'd
love
to
talk
to
you
if
you're
interested
in
finding
out
a
little
bit
more
hope.
This
was
a
fun
kind
of
introduction
to
flu
Indy
for
some
of
you,
if
you're
not
actively
contributing
highly
encourage
you
to
do
it
and
thank
you.