►
From YouTube: GitLab Monitor:APM Feature and Roadmap Discussion
A
Monetary
log
aggregation
it's
in
a
monitor
stage
of
get
lab
and
we're
doing
a
really.
We
have
pretty
extensive
error
tracking
in
Gila
now,
based
on
century.
We
also
got
log
tailing
and
we
got
tracing
with
Jade
the
SIL
bit
lock
tailing,
at
least
looking
at
logs
at
machines
and
probably
want
to
go
to
log
aggregation
and
he's
been
working
on
that
and
he's
filling
me
in
on.
What's
what's
going
on,
yeah
cool.
B
B
B
What
we
do
today
is
log
tailing,
and
so,
if
you
attach
a
kubernetes
cluster
to
your
get
lab
project,
you
can
go
look
at
the
pod
logs
and
again
just
slowly
just
tails,
the
logs
we've
added
some
ability
to
filter
to
specific
pods
and
containers,
but
you
basically
I
get
to
live
log
view
and
we
are
adding
log
aggregation,
but
we've
required
some
log
aggregation
datastore
to
do
so.
So
we
added
the
ability
recently
to
add
elastic
the
elastic
stack
as
a
gift.
I've
managed
app,
it's
presently
behind
a
feature
flag.
A
I
wish
I
thought
we
hadn't
selected
log
aggregation
method.
Yet
so
I
was
suggesting
hey.
Maybe
we
should
look
into
Loki
because
then
it's
lighter
weight
and
we
can
run
more
out
of
the
box,
but
if
we
already
select
it
elastic,
then
I'm
cool
with
that
we
can
just
move
on
I
posted
a
couple
of
questions
in
my
last
response.
Would
you
like
me
to
articulate
those
or
what's
the
best
way
to
procedure
yeah.
B
A
Get
lap
comes
with
self
monitoring,
it's
done
in
it
special
project
and
work.
It
lab
sees
itself
as
an
application
gives
you
all
the
metrics
you're
saying
yeah
we're
dogfooding.
Some
of
that.
If
my
understanding
is
that
our
operations
team
is
still
using
Gravano
for
almost
all
their
metrics
needs,
so
but
you're
saying
we're
dog
fooding,
some
part.
What
is
your
take
on?
What
part
of
the
metrics
are.
B
Yeah
and
sorry
I,
there's
probably
two
terms
to
uses
of
the
word
dark
footing
there.
So
I
can
just
immigrate
them.
One
I'm
saying
we're
we're
dog
fooding
in
that
we're
giving
get
lab
administrators
of
not
give
up
calm
but
of
a
smaller,
give
up
instance,
exposure
to
the
same
tools
their
users
would
use
for
monitoring
their
applications
in
their
view
of
the
get
lab
instance
itself.
So
that's
not
actually
us
eat
dog
food
for
gala,
calm,
its
I.
A
B
I
will
say
we
do
dog
food
it
in
the
sense
that
the
support
team,
the
one
of
the
values
of
us,
doing,
give
up
self
monitor
and
the
way
that
we
are
doing
it
is
that
the
support
team
can
work
with
a
customer
and
say:
oh
I,
want
to
know
the
health
of
your
gait
lab
instance,
and
look
at
that
instance
project
and
see
the
metrics
coming
from
that.
Like
have
a
customer
screen
share
if
there's
some
sort
of
yeah.
A
Yeah,
okay,
so
let's
call
it
customer
feeding,
just
just
invigorate
yeah
I
think
with
with
that.
I
also
think
we're
not
there
yet
I
know
of
our
current
customer.
We
have
in
the
telecommunication
space
and
they
want
better
metrics
of
what
we're
doing
is
we're
sending
over
graph
Anna
sets.
So
the
customer
feeding
is
not
going
very
well
either
and
in
my
opinion
the
first
step
should
be.
We
should
be
using
it
ourselves
and
then
we
can
recommend
it
to
customers.
We
shouldn't
put,
we
should
doing
it.
A
The
other
way
is
making
our
customers
experience
every
single
problem
first
and
that's
leads
to
much
longer
iteration
cycles,
so
I
don't
think
the
customer
feeding
is
going
as
well
as
it
could
be.
I
think
we're
still
recommending
one
or
two
customers
and
I
think
the
solution
is
not
to
make
that
better.
I
think
the
solution
is
first
start
using
it
ourselves.
Yeah.
B
I
agree:
it's
actually.
If
you
the
documentation
for
get
web
self
monitor,
is
it
is
still
not
presently
available.
There
are
no
customers,
you
have
this
self
monitor
project
available,
so
they
are
defaulting
to
using
graph
Anna
if
they
wants
to
see
a
dashboard
of
their
gitlab
metrics
and
we
have
been
shipping
grow
fauna
as
kind
of
a
short
term
option
there,
because
we
needed
to
have
some
method
to
display
the
Prometheus
metrics
and
we
didn't
have
our
dashboarding
capabilities
or
this
project
available.
B
But
I
expect
over
time
that
we
will
continue
to
eat
away
at
having
more
built
in
default,
dashboards
and
not
pointing
to
or
fauna,
and
to
that
point
the
give
at.com
team
is,
does
not
have
dashboards
set
up
in
github
itself
for
monitoring
our
github
deployment,
and
that
should
be.
Our
starting
point
is
what
dashboard
are
using
ingre
fana,
let's
put
it
in
our
get
lab
project
and
then
then
start
shipping,
those
same
default
dashboards
to
our
users.
Yes,
it's.
A
There's
two
situations:
it's
like
you
make
custom
dashboards,
forget
lab
calm,
and
then
you
look
at
what
you
should
ship
and
get
laughs.
I
propose
to
do
the
other
way
like
just
make
everything
that
we
need
for
get
lab,
become
the
default
for
self-monitoring
and
then
half
of
them
don't
work
for
customers
and
we
remove
dough.
But
in
the
beginning
we
can
just
ship
everything
and
get
loud
because
I'm
much
more
afraid
of
not
shipping
enough
like
a
dashboard
that
doesn't
work.
It's
not
nice,
but
it's
understandable
not
having
the
dash
for
it.
A
In
the
first
place.
That's
a
big
problem,
so
I'd
love
for
us
to
error
on
that
side
and
just
make
make
sure
that
the
the
wait
get
lab
comes
out
of
the
box
works
for
our
accounting.
That's
the
that's,
probably
the
toughest
thing
in
the
world,
so
we
should,
although
we
should
iterate,
but
let's
do
that
and
then
have
a
bunch
of
passwords
that
don't
don't
make
sense
to
our
customers,
but
we
can
explain
that
I
think
a
lot
better
than
not.
A
B
I
will
I
do
think
that
that
is
the
way
the
team
is
approaching.
It
douve
has
almost
word-for-word
said
those
same
words:
it's
okay
to
have
dashboards
ship
that
don't
necessarily
work.
It's
more
important
to
have
more
information
possible
dashboards
that
do
work
than
the
other
way
around.
So
totally
agree
what.
A
B
I
think
it's
almost
entirely
yeah
of
gitlab
comm.
It's
zero.
The
starting
projects
have
been
a
couple
of
key
metrics
that
have
had
alerts
set
up
on
them
and
a
couple
of
non
gate
lab
application
services,
like
version
and
pyjamas
and
others
are
moving
to
auto
dev
ops
and
using
kind
of
like
the
full
monitoring
suite
yeah.
A
B
B
We
have
this
ability
where,
in
the
get
that
application
you
assign
an
external
dashboard
Lincoln,
so
you
can
have
my
starting
point
is
a
gate
lab
forward
with
what
I
know
to
be
some
key
metrics
and
then
I
can
like
jump
to
an
external
link
to
grow
fana.
If
I
really
need
that
backup
available-
and
so
maybe
it's
just
getting
started
there
and
finding
more
and
more
cases
where
it
would
have
been
great
to
have
this
dashboard
and
we
can
create
it
or
this
metric
view
and
we
can
create
and
get
labs,
makes.
B
A
B
B
Yeah
that's
correct,
and
that
was
the
the
health
team
was
finalizing
error
tracking
they
had
some
stuff
in
the
queue
before
I.
Think
six
weeks
ago
we
had
to
talk
about
prioritizing
that
flow
and
they've
been
adjusting,
but
we
had
a
plan
to
get
air
tracking
to
viable.
That
was,
the
team
was
finishing
off
okay,
but
yeah
Sarah
has
is
planning
on
adjusting
and
as
validation
and
other
items
that
are
all
about
the
flow.
As
a
result,
okay,
great
yeah,.
A
B
Hard
and
I
think
on
the
dashboards
part
the
the
dashboards
are.
You
know,
they're,
based
on
the
each
arts
library.
There
are
the
kinds
of
things
that
we
talk
about,
dogfooding
and
I
talked
to
product
managers
about
dogfooding
is
not
just
a
mandate
for
us
to
prioritize
the
product
team's
effort,
but
also
as
an
opportunity
to
leverage
existing
teams
that
might
be
putting
effort
in
other
places.
So
maybe
that's
a
good
reminder
to
me
to
go
back
and
engage
with
the
sre
and.
B
B
B
A
B
One
thing
that
I
would
actually
appreciate
your
thoughts
on
this
was
a
discussion
that
douve
and
I
had
and
I
was
a
little
bit
unsure
about
the
answer.
You
mentioned
the
log
tailing
experience
that
we
have
today.
There's
been
some
question
about
whether
or
not
we
should
deprecated
that
current
experience,
which,
in
favor
of
the
log
aggregation
kind
of
like
log,
explore
and
part
of
that,
has
been
well.
A
Yeah
we've
decided
to
do
but
yeah
I
think
if
you're
gonna
look
half
a
year
from
now.
What
percentage
has
the
alleged
state
search
managed
step?
It's
gonna
be
fewer
than
50%,
maybe
even
fewer
than
10%,
like
what
percentage
is
running,
get
a
bond
kubernetes,
your
vocal,
but
it's
probably
10
or
20%,
and
then
and
then
you
still
have
to
hit
that
elastic
search
button.
So
and
then
you
have
an
elastic
search
installation
which
is
known
in
the
industry
to
be
hard
to
run.
A
So
that's
gonna
be
a
very
tiny
percentage,
so
I
I
think
it
makes
so
little
sense.
Also,
people
are
not
used
to
get
a
lot
of
offering
logging
so
just
getting
that
in
front
of
them
and
seeing
that
log
tailing
and
then
like
this
is
great
but
I
need
more
in
to
say,
oh
well,
you
need
more.
This
is
the
path
forward,
so.
B
That's
great,
actually,
it
never
makes
me
think
that
drawing
that
connection,
oh
you've
got
log
tailing.
Here's
where
you're
getting
logged
more
or
find
more.
It
requires
us
all
lastest.
Oh
the
elastic
stack
on
this
cluster,
but
hey
at
least
it's
us
get
it
giving
them
some
progress.
So,
okay,
okay,.
B
When
I
pointed
that
issue
to
you,
I
read
back
through
the
history
there's
some
dialogue
between
myself
and
Daniel
grew.
Sir.
The
product
manager
about
seems
to
be
more
about
the
I
think
there
is
a
use
case
for
people
and
adding
a
cluster
that
is
not
actually
about
deploying
their
applications
to
it
and
so
I
think
Daniels
argument
was.
We
might
be
installing
all
the
things
when
we
expect
you
to
be
deploying
your
application
to
this
cluster
versus
oh
I.
A
B
A
B
B
A
B
Appian
team
has
been
working
on
low
logging
stuff
in
there.
The
plan
is
to
been
working
on
removing
blogging
from
minimal
to
viable
and
that's
involved,
both
the
installing
of
elastic
stack
and
the
log
Explorer
view,
and
there's
been
a
couple
of
other
things
in
that
epics.
That
I
can
link
here
and
there
actually
have
been
a
couple
of
adding
new
chart
types
over
the
last
couple
of
releases
by
the
APM
team.