►
From YouTube: Distribution Demo - Dec 31, 2020
Description
The distribution team demos using a patroni standby cluster for geo replication.
A
Hello
and
welcome
to
the
distribution
team
meeting
december
31st
2020
today,
I
am
going
to
be
running
through
maybe
for
the
first
time
on
on
my
machine
here,
a
setup
of
gitlab
and
gitlab
geo,
but
with
the
get
lab
geo
using
petroni
standby
cluster.
A
So
this
is
a
change
that
has
been
added
to
the
omnibus
for
the
previous
release
and
something
the
geo
team
has
been
working
through
in
the
patrony
support,
go
ahead
and
share
my
screen.
A
So
for
this,
I'm
currently
working
on
some
reviews
related
to
changes
around
the
standby
cluster,
and
so
when
I'm
working
on
reviews,
I.
A
Kind
of
very
much
focus
on
booting
up
the
minimal
or
very
close
to
the
minimal
cluster
needed
to
test
to
test
it
in
some
cases,
especially
early
on
the
review
to
be
able
to
iterate
fast
back
and
forth
on
the
review
in
these
test
environments.
So
I'm
actually
using
docker
and
docker
compose
to
set
up
basically
a
full
environment
that
will
exercise
the
standby
cluster,
but
not
not
necessarily
be
an
environment
that
anyone
would
use
in
production.
A
So
what
that
looks
like
is
our
three
console
nodes
of
our
primary
application:
two
database
nodes
for
that
primary
application,
one
pg
bouncer
for
the
primary
application
cluster
and
one
app
node,
because
I'm
not
really
I'm
I'm
focusing
on
the
database
replication.
A
That's
why,
even
though
the
rest
is
redundant
another
the
pg
bouncer
that
I
only
have
one
app
and
then
on
the
geo
side.
It's
basically
a
mirror
of
that
with
three
console
g,
three
consoles
for
geo,
two
geodbs,
a
gopg
bouncer,
a
geo
app
and
I've
also
split
out
the
geo's
tracking
db
into
a
separate
container
as
well.
A
So
these
are
all
running
the
omnibus
image
all
running
the
nightly
and
I've
just
started
prior
to
this
meeting
just
started,
bringing
up
some
of
the
primary
cluster
so
so
far
I've
just
got
the
console
instances
up
and
just
the
db1.
So
I'm
going
to
continue
with
that.
A
These
are,
these
are
taken
from
the
geo
installation
docks.
Actually,
we
can
jump
over.
A
A
So
on
the
geo
database
replication,
we
actually
have
a
section
now
on
petroni
support,
and
so
this
is
for
the
database
setup.
My
my
primary
is
basically
following
settings
here.
Main
difference
between
this
and
the
normal
setup
is,
of
course,
rep
manager
is
set
to
false.
We've
got
patrony
set
up
with
pg
rewind,
and
we've
declared
a
physical
replication
slot
for
petroni
to
set
up
in
postgres
and
then
on
the
standby
cluster
side.
A
We
once
again
have
rep
manager,
disabled
and
then
we
have.
We
have
petroni,
that's
interesting.
A
A
As
I
mentioned
before,
for
especially
for
these
h8
setups,
I
like
using
docker
compose
just
because
of
how
quickly
I'm
able
to
bring
up
the
instances
versus
setting
up
vms.
A
Which
is
a
lot
nicer
for
review
iteration
when
working
on
reviewing
some
changes
that
are
impacting
the
aha.
A
A
It
looks
like
we
did,
get
the
pg
bouncer
set
up
and
it
got
the
found
the
correct
master.
A
So
in
this,
in
this
scenario
I
haven't
moved,
I
haven't
moved
redis
into.
I
haven't
moved
redis
to
have
the
sentinel
set
up
or
anything
like
that.
So
everything
else
other
than
the
databases
are
basically
running
in
the
app.
A
Nodes
once
again,
not
some,
not
necessarily
something
you
would
do
in
production,
but
something
that
lets
us
exercise.
The
database
replication
portions
quickly.
A
So
one
nice
thing
is
the
the
switch
to
patroni
and
the
use
of
these
standby
clusters
has
made
it
easier
to
set
up
this
sort
of
scenario.
This
was
pretty
difficult,
or
at
least
included
a
lot
of
a
lot
more
configuration
in
order
to
set
up
the
same
sort
of
thing
previously,
with
git
lab
to
set
up
a
secondary
cluster
in
rep
manager.
That
was
replicating
from
the
first
here
in
patroni.
A
A
Basically,
didn't
have
to
write
a
whole
lot
else
under
the
hood.
This
config
right
here
is
being
passed
right
into
patrony
and
that
that
is
settings
up
setting
it
up
for
us.
We
saw
some
limitations
here,
we're
still
working
on
getting
service
discovery
and
console
setup
in
a
way
that
we
can
make
use
of.
A
A
Okay,
now,
let's
start
bringing
up
the
geo
cluster,
so
I
haven't
fully
gotten.
I
haven't
previously
set
up
this
config
that
I'm
working
on
here.
I
have
gotten
the
databases
up,
but
I
haven't
gotten.
A
This
will
be
the
first
time
I've
done
the
pg
bouncer
and
the
geo
app
on
this
standby
cluster,
so
we'll
bring
up
the
geo.
A
C
A
A
A
A
A
A
A
A
A
A
B
A
B
A
C
And
you're
changing
that
board,
primarily
because
you're
running
all
on
the
one
instance,
one
one
host.
A
No,
it's
just
the
rp
pg
bouncer
instances
are
by
default.
They
run
they're
running
this
port
versus
postgres,
which
is
running
on
five.
A
So
basically,
it
just
means
that
this
previous
config
I
had
was
copied
from
a
non-pg
bouncer
geo
setup.
But
when
we're
going
through
pg
bouncer,
we
have
to
change
the
you
have
to
tell
rails
that
it's
a
different
port
than
the
post
right.
The
default
postgres
port,
because
we're.
B
B
A
A
Okay,
that
does,
in
fact,
look
like
at
least
from
my
installation
setup,
but
it's
happy,
let's
see
what
else
we
might
need
to
do
here.
So
one
of
the
things
you
have
to
do
normally
is
run.
This
set
primary
node,
but
this
is
in
the
in
the
docs.
This
is
steps
without
patroning.
B
B
A
A
A
Interesting,
so
it
is
operating
enough
to
be
actually
up
here.
Is
it
replicating
it
hasn't
replicated
yet,
but
I
just
turned-
or
at
least
it's
not
look,
it's
replicating
the
database,
but
not
hasn't
yet
finished.
Replicating
the
repositories.
B
A
B
A
A
B
B
A
B
C
A
A
But
yes,
replication
is
now
set
up
both
on
my
previous
instance
and
on
my
previous
created
project
on
and
on
the
new
one
I
created
with
after
we
had
set
up
the
geo
apps,
okay
that
actually
all
went
better
than
I
was
expecting.
The
main
thing,
that's
still
throwing
errors
is
that
the
geo
log
cursor
is
enabled
on
well
there's
a
couple
things.
The
geo
log
cursor
is
enabled
on
my
database
nodes.
B
A
Secondary
okay,
so
on
their
secondary,
they
aren't,
including
in
the
instructions
they
are,
including
the
geo
secondary
rule,
which
I
am,
which
is
probably
the
cause
of
my
issues.
A
The
geosecular
okay,
so
that's
that's
where
my
config
is
wrong.
Is
that
if
I
am
going
to
use
this,
it's
going
to
enable
that
geolog,
cursor
and
I'd
have
to
manually
set
it
to
disabled
for
that?
Okay,
so
that
makes
sense.
But
then
the
other
thing
they
had
was
petroleum
enabled
false.
A
So
that's
that's
the
last
thing
we'll
try
at
the
moment,
but
other
than
that.
I
think
we're
good,
I'm
just
going
to
jump
into
the
database
node
and
restart
it
and
see
what
happens
or
reconfigure.
It.
C
A
C
C
A
C
A
B
A
A
A
C
A
I'm
gonna
bring
it
back
and
we'll
call
that
good
for
today's
demo.
A
Yeah
in
terms
of
reviews,
I'm
working
on
this
because
we
are
we're
trying
to
update
patrony
to
the
2.01
release,
we're
on
1.64
or
something
like
that.
So
I
just
wanted
to
make
sure
that
I
had
environments
that
could
actually
confirm
and
test
that
upgrading
worked
and
needed.
This
standby
cluster
set
up
to
be
able
to
test
the
upgrade.
A
Yeah,
so
that
should
be
it,
so
we
have
petrol
up
and
running
being
that
this
one
was
down
for
a
while.
We
can
see
this
is
geodb1
it
we
didn't
watch,
but
it
probably
should
have
failed
over
to
db2.
B
C
A
Registering
service
as
yeah
standby
leader
cool
well,
that's
good
looks
like
everything's
working.
I
went
a
little
bit
over
time
on
the
demo.
I
think,
but
yeah
we'll
call
it
there
for
today.