►
From YouTube: GitLab Kickoff 13.3 - Enablement:Geo
Description
Nick Nguyen (Engineering Manager, Geo) discusses planned features and improvements for GitLab Geo in 13.3. Planning issue: https://gitlab.com/gitlab-org/geo-team/discussions/-/issues/4972
A
A
All
right,
one
main
theme
for
13-3
is
a
focus
on
moving
our
disaster
recovery
capabilities
towards
viable
maturity.
We
have
made
significant
progress
over
the
last
couple
of
months
and
have
completed
over
60
percent
of
the
issues
in
our
viable
maturity.
Epic,
this
work
aims
to
increase
the
amount
of
data
types
we
replicate
to
a
Geo
secondary,
improve
support
for
a
plan
failover,
as
well
as
improved
geo
scalability
by
simplifying
backfill
operations,
making
them
more
performant.
A
Some
new
data
types
that
we
plan
to
support
in
13-3,
our
external
merge
request,
diffs
terraform
state
files
and
vulnerability
export
files.
We
had
planned
these
for
13-2
and
made
good
progress
in
that
milestone,
but
along
the
way
we
discovered
some
considerations
that
have
now
pushed
the
outlook
back
to
13-3.
A
Moving
on
to
the
theme
of
geo
scalability,
we've
made
a
lot
of
iterative
improvements
to
the
backfill
performance
for
different
replicable.
Having
geo
enabled
on
get
labs.
Staging
environment
allowed
us
to
better
troubleshoot
scalability
issues
and
validate
improvements
there
in
13-3,
we
plan
to
continue
testing
the
improvements
that
are
still
behind
a
future
flag
and
we'll
remove
them
if
no
problems
arise.
Our
customers
with
large
installations
should
see
a
noticeable
improvement
in
the
time
it
takes
to
sample
large
amounts
of
data,
we'll
also
be
able
to
remove
code
and
documentation
related
to
find
data.
Wrappers.
A
A
Currently
secondaries
only
support
a
single
database
node,
and
this
is
problematic
for
customers
using
geo
for
disaster
recovery,
because
a
failover
to
a
secondary
can't
use
a
high
availability
configuration
immediately.
Now
that
Patroni
is
going
to
be
bundled
with
omnibus.
We
are
going
to
investigate
using
Patroni
to
manage
a
standby
cluster
on
the
secondary.
A
Another
way
that
we
want
to
improve
the
failover
experience
for
our
customers
is
to
improve
our
failover
documentation.
Currently,
the
documentation
is
a
bit
scattered
and
we'd
like
to
update
it
to
provide
a
more
runbook
style,
linear
process
for
failing
over
on
well
supported
geo
configurations
such
as
a
two
site
single
note,
geo
with
an
omnibus
managed
database
on
the
front
end.
We
have
some
exciting
developments.
A
All
right,
well
that
does
it
for
this
walkthrough
of
the
Geo
team's
plan
for
13-3
I'll
end
with
our
usual
disclaimer
that
we
plan
ambitiously
and
work
in
a
continuous
iterative,
Kanban
style.
So
these
plans
are
always
subject
to
change
and
we
will
be
sure
to
provide
regular
updates
in
our
epics
and
issues
all
right.
Thanks
for
watching.