►
From YouTube: Plan self-managed decomposition epic
A
A
So
I
think
in
general,
the
plan
is
good.
So
in
that,
if
we
work
backwards,
the
whole
idea
is
that
we
want
to
allow
the
schemas
to
diverge
by
18.0
right
so,
which
is
a
long
way
away,
and
it
gives
a
lot
a
lot
of
time
for
users
to
upgrade
and
then
and
then
further
backwards
by
17.0,
we'll
we'll
make
a
migration
mandatory
to
a
decompose
database.
So
that's
fine
I
think
is
so.
A
A
If,
for
some
reason
in
a
minority
of
cases,
you
are
using
too
much
postgres
connections,
it's
it's
quite
easy
to
just
say:
okay,
how
on
the
temporary
switch
back
to
the
single
connection
for
now
right?
This
is
to
support
and
no
data,
I'll
also
work,
which
is
the
main
problem
with
the
decomposed
one.
A
And
also,
at
the
same
time,
we
also
kind
of
blocked
slightly
by
we're
waiting
on
distribution,
some
distribution
tasks
to
create
the
CI
database
by
default
in
Omnibus
and
also
charts,
I
suppose
as
well.
So
I,
don't
know
whether
that
will
make
16.0
or
not
as
well
so
four
chances.
But,
as
we
know,
distribution
is
has
a
lot
of
other
competing
priorities.
So
that's
that's.
A
A
It's
not
so
this
is.
This
is
kind
of
a
long
longer
term.
It
shouldn't
affect
the
new
installations
or
whatever,
but
in
terms
of
migration,
which
is
required
by
17.0
I.
Think
the
the
consensus
is
that
we
should
focus
on
documenting
first
what
we
can,
because
a
lot
of
things
probably
can't
be
automated,
because
we
we
just
the
main
thing
is
that
Adobe
there
will
definitely
be
cases
where
we
don't
control
their
postgres,
their
PG
bouncer,
as
in
we
didn't
install
it
for
them.
A
A
B
A
So
those
are
good
things
to
fix,
because
those
are
issues
that
customers
will
face
as
well
when
they
run
or
decomposed
and
then
and
then
the
bigger
things
will
be
then
like
supporting
the
migration
somehow
a
little
bit
of
a
very
large
thing,
and
that's
definitely
more
than
enough
time
to
do
that,
and
also
supporting
the
delivery
team
to
to
migrate
the
Dev
instance
and
Ops
instance
to
decompose
as
well
yeah.
So
there's
plenty
of
time
for
that.
For
those
things
for
sure
up
to
17.0.
A
Let
me
see
my
screen:
yes,
yeah,
okay
cool,
so
I
I
was
proposing
that
we
dropped
this
like
we
dropped
this
kind
of
special
casing
for
new
installations,
and
and
basically
we
replace
it
with
require
two
Connections
in
state
for
16.0,
and
then
we
we
keep
this
decompose
requirement
in
17.0.
It
does
there's
no
work
right.
Let's
just
saying
this
is
saying
that
adding
any
documentation
or
this
in
sometime
in
16.0
or
16.1,
say:
okay,
you
must
move
to
decompose
by
17.0
right.
A
Most
of
the
work
is
troubleshooting
and
writing
docs
and
migrate
and
session
in
migration.
Tooling,
no
migration
dogs
so
like
that
shouldn't
take
up
like
a
lot
of
effort
and
there's
a
lot
of
time
as
well.
This
one
might
might
be
might
take
one
or
two
Milestones,
because
we
don't
have
this
option
thing
potentially,
but
other
than
that.
It's
sort
of
a
doing
this
sort
of
like
switching
it
to
default.
That
and
then
writing
some
dogs
and
for
for
troubleshooting,
mostly
I've,
already
got
a
feature
Docs.
A
B
A
It's
just
that
we've
never
done
it
before
that's
the
thing:
okay,
yeah
and
also
I.
Don't
think
that
we
should
be
rushing
to
to
kind
of
achieve
this
when
a
lot
of
the
prerequisites
are
not
there
like
we
don't
we
don't
have
the
auto
creation,
we
don't
necessarily
have
test
and
migration
properly.
So
a
lot
of
things
are
not
there
and
then
like.
If
something
goes
wrong,
migrating
back,
there's
no
there's
no
migrating
back.
A
A
But
things
like
this
effectively
is
done
right
like
effectively.
Today,
someone
can
opt
in
to
decompose
if
they
want
to
follow-up
problems,
full
of
gaps
but
still
yeah
and
it's
experimental,
but
gearlab.com
is
running
it.
So
if
you
know
what
you're
doing
potentially
you
could
but
not
not
recommended,
of
course
yeah
and
then
yeah.
This
is
the
one
that
I
just
reopened
and
then
new
installations
are
decomposed
by.
A
B
A
A
The
tooling
I
have
no
idea
what
tooling
even
we're
gonna
do
so
I
would
say:
I
will
push
this
out
actually
yeah
yeah
to
like
next
one
yeah
yeah,
that's
right
so
I
didn't
add.
Other
things,
I
think
are
correct,
require
deprecate
six
non-county
compulsive
system,
but
I
know
that's
correct,
require
back
law.
This
is
simply
17.0,
but
it
has
17.0
Isn't
created
yet
so
yeah.
Okay,
that's.
B
It's
more
the
it's
more,
the
next
three
so
for
for
that
planning
issue
that
quality
planning
issue
I've
got
it's
just
mapping
those
back
to
the
indicative
kind
of
Milestone
aspirations
that
we
have
and
I
just
want
to
be
able
to
go.
Okay.
Well,
we've
now
got
I
think
because
if
you
go
to
the
top
of
that
planning
thing,
I
think
we
don't
have
anything
yeah
we
haven't
got
anything
for
15,
10
and
15
11.
B
B
B
Yeah
I
mean
again
if
it.
If,
if
there
is
no
15
10
15,
then
it's
replaced
by
something
else
and
I'll
just
replace
it
here,
but
yeah,
essentially,
that's
that's
just
call
them
the
the
the
further
two
Milestones.
Is
there
anything
that
we
can
be
and
of
course
it
would
if
they
are
going
to
be
replaced
with
something
then
it'd
be
easier
to
say.
Well,
we
had
everything
at
15
10
here
and
we
can
replace
the
milestone
for
it.
B
A
A
B
A
B
B
B
Not
for
my
side,
no,
it
was
just
how
it
for
me
it
was
the
more
of
the
mapping
to
that
to
the
planner.
Now
that
we
can
do
that.
From
my
point
of
view,
that's
great
I
wanted
to
make
sure.
From
your
point
of
view,
everything
was
realistic
and
you
were
happy
with
the
timelines.
So
that's
great
too
so
yeah
I'm,
I'm,
all
good
okay,.