►
From YouTube: GitLab 12.6 Kickoff - Enablement:Geo
Description
Fabian Zimmer (Product Manager, Geo) highlights new features and direction for GitLab Geo in 12.6.
A
Oh,
hello,
everyone:
this
is
Fabian
the
product
manager
for
a
good
lab
geo
and
today
I'm
going
to
walk
you
through
the
road
map
of
Geo
and
what
we're
up
to
in
the
next
release.
So
yeah.
Let
me
start
with
one
thing
that
I
talked
about
last
time
here
on
a
design
repositories,
so
we
had
intentions
to
ship
this
in
12.5.
A
Unfortunately,
this
is
going
to
be
a
little
bit
delayed.
Currently
we
are
working
on
exposing
all
of
the
design
repository
information
in
the
admin
area
and
in
order
to
polish
that
we
need
a
little
bit
more
time,
so
this
is
going
to
slip
from
twelve
point
five
to
twelve
point
six.
So
that's
the
first
thing
to
know
so
we
hope
to
deliver
this
by
the
next
release
of
the
22nd
of
December.
A
I
spoke
about
it
in
the
past,
so
I'm
not
going
to
go
too
much
into
detail,
but
this
is
ideally
going
to
end
and
come
to
a
conclusion
in
12.6.
Okay.
So
let's
talk
about
some
of
the
things
we
are.
We
are
doing
right
now.
So
we
closed
actually
out
a
few
other
epochs.
We
focused
a
lot
on
improving
our
upgrades.
We
added
H
a
instructions
we
made.
A
The
foreign
data
wrapper
tables
a
little
bit
more
robust
and
I'm,
not
showing
those
epics
here
anymore,
but
for
12.5
we
closed
out
a
few
things,
so
that
has
an
impact
on
what
we're
doing
right
now
and
I.
Think
the
first
thing
to
to
highlight
is
something
that
we
are
like
starting
to
work
on
at
the
moment,
which
is
building
a
scalable,
self-service,
G,
replication
verification
framework,
and
so
I
want
to
spend
a
little
bit
of
time
on
this.
A
Just
to
give
you
some
context
in
the
last
releases,
usually
there
was
a
large
epic
that
spoke
about
replicating
unreplicated
data
types,
so
geo
does
not
currently
cover
all
of
it
and
also
verifying
it,
but
we
found
that
that
was
actually
like
preceding
a
little
bit
more
slowly
than
we
wanted,
and
there
are
a
couple
of
reasons
for
it.
The
one
of
the
reasons
is
that
gate
lab
is
growing
right
and
we're
adding
many
new
data
types
and
geo
and
the
team,
even
though
everybody
works
very
hard,
can't
really
keep
up
as
much
anymore.
A
A
So
this
is
going
to
be
something
that
will
finish
in
12.6
and
this
is
a
lot
of
groundwork,
but
we're
also
going
to
work
on
a
POC
here,
so
a
proof
of
concept,
and
that
should
be
really
time
boxed
and
quick,
because
we
have
a
lot
of
assumptions
at
the
moment,
also
on
a
technical
level,
and
we
we
want
to
avoid
essentially
going
down
a
path
for
many
months,
even
though
we
have
some
technical
challenges.
So
this
should
be
really
quick,
ideally
like
a
week
or
two
just
to
validate
that.
A
We
can
do
something
like
that
that
our
ideas
are
sound
and
then
we'll
move
on
and
break
all
of
that
into
much
smaller
iterations
and
like
push
this
forward
in
the
incremental
way.
But
that's
a
that's
a
big
effort
in
the
team
and
if
you
just
look
at
the
issue,
there
are
some
like
proposals
on
how
to
actually
do
this
or
the
team
is
really
engaging
with
this
at
the
moment
and
trying
to
figure
out
how
to
how
to
do
that.
A
Yeah,
so
this
is
probably
going
to
continue
for
a
while,
and
this
is
sort
of
the
first
step
here.
We
started
thinking
about
it
already
this
release,
but
for
12.6
the
POC
is
really
what
what
will
come
out
of
it
and
other
than
that
we
are
switching
now
and
we
are
also
like
in
you
know,
sort
of
as
a
second
track,
and
this
is
a
tiny
bit
cluttered,
because
I
can't
collect
subtopics,
but
the
second
track
is
essentially
a
renewed
focus
on
disaster
recovery.
A
So
the
first
thing
that
we're
doing
at
the
moment
is
we
are
rolling
our
geo
to
staging
the
staging
environment
of
Quetelet
calm,
because
we
believe
that
dogfooding
and
using
our
own
products
is
very
valuable
and
we
want
to
do
that
by
the
end
of
the
year,
so
release
this
with
the
22nd.
That's
a
sort
of
a
prerequisite
for
actually
rolling
it
out
onto
production,
because
it
allows
us
to
understand
exactly
how
all
of
that
works.
It
also
I
think
it's
a
great
insight
into
the
automation.
A
We
can
build
around
that
and,
as
you
can
see,
there's
there's
a
few
tasks
here
that
are
going
to
follow
in
12.6,
so,
for
example,
an
enabling
replication
of
other
data
types
we're
currently
working
on
enabling
Postgres
replication
executing
on
test
plans
and
so
on.
But
this
is
really
a
big
priority
for
the
team,
because
it
is
relevant
on
many
different
levels
and
we're
focusing
on
this
as
sort
of
a
part
of
our
disaster.
Recovery
focus
so
hopefully
by
12.6
the
end
of
this.
A
We,
we
are
essentially
saying
okay
with
all
the
things
that
we
need
to
improve
and
disaster
recovery
space.
A
good
starting
point
is
actually
to
improve
the
support
for
a
planned
failover,
because
every
customer,
including
us
who
will
roll
out
a
disaster
recovery
solution,
will
be
interested
in
like
doing
a
really
structured
plan,
shake
overs
and
as
part
of
this,
of
this
effort,
there
are
a
few
feature,
features
that
we
are
investigating
right
now,
but
there's
also
quite
a
bit
of
discovery
like
work
going
on
so
I'll
zoom
in
here.
A
A
little
bit,
for
example,
it
lab
is
making
big
steps
to
actually
provide
reference
architectures
for
specific
deployment,
so
for
AJ
environments
and
we're
going
to
go
through
and
a
discovery
effort
at
the
moment
to
understand.
You
know
where
are
the
the
points
in
our
failover
process
that
are
cumbersome
or
error-prone
right
and
really
like,
focusing
on
understanding
the
the
current
limitations
so
we'll
run
at
least
two
demos
of
our
planned
failover
process
in
12.6.
This
is
a
little
bit
similar
to
what
we
did
with
upgrade
demos
which
we
do
every
month.
A
We
just
want
to
make
sure
that
you
know
we
we
get
into
the
habit
of
doing
this
and
also
then
on
a
separate
tech,
we'll
build
sort
of
the
the
alternation
surrounding
this.
But
before
we
automate,
we
want
to
sort
of
really
be
sure
that
the
process
is
right.
So
that's
that's
something
we're
going
to
do,
and
then
there
are
two
interesting
feature
proposes
that
I
would
like
to
highlight.
The
one
is
a
maintenance
mode.
This
is
actually
bigger
than
the
Geo
in
in
general.
A
There
are
at
least
a
few
use
cases
where
this
can
be
very
helpful.
So
if
you
are
a
systems
administrator
and
you
need
to
perform
some
kind
of
operation
at
which
you
would
like
users
not
to
interact
with
the
gate
lab
installation,
we
would
like
to
add
support
for
essentially
putting
an
instance
into
a
read-only
mode
where
no
changes
can
be
made.
A
This
is
relevant
for
plan
fail,
overs,
because
you
can
imagine
that
if
you
have
a
primary
in
the
secondary,
if
you
need
to
always
sing
over
Changez,
the
primary
is
often
a
tiny
bit
ahead
of
the
secondary,
and
if
you
put
the
system
into
maintenance
mode,
then
actually
you
can
guarantee
at
some
point
that
the
two
systems
are
really
in
sync:
there's:
no,
no
data
being
lost.
We
have
a
little
bit
more
control.
A
So
this
is
why
this
is
on
our
plate,
but
there
are
plenty
of
other
reasons
why
systems
administrators
are
interested
in
in
the
maintenance
mode.
So
we
are
taking
a
stab
at
that
and
also
in
a
similar
fashion,
to
what
we
are
doing
with
the
replication
model.
We
just
need
to
validate
some
assumptions
that
we
have
on
a
technical
level,
so
we
are
also
going
to
do
a
small
QC
here
and
then
from
there
I
actually
move
forward
again
with
an
MVC
for
for
a
maintenance
mode.
So
that's
one
of
the
things.
A
A
So
imagine
a
situation
where
you
have
a
primary
and
a
secondary
server
and
the
secondary
server
is
putting
quite
a
bit
of
load
on
to
the
primary,
and
you
know
that
can
be
very
undesirable
and
especially
now
that
we
are
moving
towards
putting
these
systems
on
comm.
We
are
thinking
about
ways
of
of
handling
that,
and
we
spoke
with
some
of
our
sres
and
the
team
and
I
think
it
may
be
desirable
to
actually
improve
hours
of
pause
and
resume
logic
in
geo.
A
So
we
to
be
able
to
pause
replication
on
a
secondary
so
that
you
know
the
secondary.
Essentially,
it's
idle
not
putting
any
load
on
the
on
the
primary,
and
that
again
can
also
be
quite
helpful
in
other
situations
where,
for
example,
if
you
want
to
upgrade
the
primary
right,
you
can
essentially
isolate
the
secondary
a
little
bit
from
from
the
primary
and
yeah.
So
we're
hoping
to
do
this.
This
is
technically
actually
quite
interesting
because
it
deals
a
lot
with
phosphorus
and
the
specifics
on
how
that
database
technology
works.
A
I
think
many
of
the
systems
administrators
that
use
G
as
well-
and
this
is
improving
our
user
experience
on
the
administrative
level
and
I'm
very
happy
that
since
this
month,
actually
singeing
joined
us
as
a
product
designer,
and
so
we
are
going
to
now
actually
take
a
look
at
the
or
starting
to
take
a
look
at
the
administrator
panel.
What
the
experience
is
sort
of
baseline
requirements
are
for
systems
administrators
interacting
with
this
and
then
sort
of
having
a
push
towards
making
that
easier
to
use
for
for
you.
A
So
at
the
moment,
we're
still
in
early
phases
but
I
think
at
the
in
December.
This
is
going
to
pick
up
a
little
bit
and
it's
going
to
continue
throughout
the
beginning
of
2020,
and
then,
lastly-
and
this
is
really
important
and
I
think
this
is
potentially
going
to
accelerate
a
little
bit
but
we'll
see
currently
geo
is
quite
hard
to
install
it's.
You
need
a
lot
of
manual
steps
and
it's
not
the
most
pleasant
experience,
especially
with
very
large
installations.
A
So
at
the
moment
another
team
inside
get
lab
distribution
is
already
interested
in
helping
us
automate
some
of
those
steps,
and
this
is
something
that
that
may
kick
off
very
soon.
So
I'll
keep
you
posted,
but
you
know
with
all
the
di
work
on
the
go
at
the
moment.
We
believe
that
sort
of
by
the
beginning
of
2020
we
will
ramp
up
that
effort
as
well,
maybe
a
little
bit
earlier.
A
So
that's
definitely
on
the
menu,
and
that
may
actually
have
a
very
positive
excited
like
it
may
have
just
been
ofit's
in
many
other
areas
as
well,
because
it
will
mean
increasing
automation
and
making
testing
easier.
So
we
do.
We
do
care
about
that,
and
that
is
geo
in
twelve
six.
So,
to
summarise,
in
terms
of
like
features,
you
can
look
forward
to
design
repositories,
we're
going
to
look
at
the
maintenance
mode,
we're
going
to
do
discovery
on
you
know.
A
The
plan
failed
over
proceeded
procedures,
we're
working
on
like
pausing
and
resuming
post
post
replication,
and
then
there
is
some
really
exciting.
We're
kicking
off
here
in
terms
of
building
a
scalable
and
self-service
to
your
application
framework,
which
I
believe
will,
if
you
know,
if
we
have
good
signals
in
the
beginning,
will
really
enable
us
to
move
even
faster
in
2020
and
that's
it
I
hope
you
enjoyed
that
call.