►
From YouTube: Magento Architectural Discussion -- September, 4, 2019
Description
- Consumer processes improvements
- PSR-12
- auto_increment in MySQL 8
Meeting minutes: https://github.com/magento/architecture/issues/258
B
B
The
process
in
the
background
and
the
process
will
check
if
there
are
messages,
and
then
it
goes
to
sleep
also
after
every
after
it
was
the
first
method
and
goes
to
sleep
and
then
comes
back.
So
the
problem
which
raised
was
there
are
some
kills,
which
which
will
be
running
the
processes
for
a
much
longer,
and
it
will
just
be
exhausting
resources
without
being
efficient,
because
each
process
takes
up
some.
B
So
that
was
one
suggestion
and
the
other
part
was
that
they
were
also
concerned
that
you
know
and
when
we
update
any
new
future
and
if
you
want
to
run
a
new
code
for
the
cat
I
have
to
pick
up
also
say
we
have
the
poison
film
process,
so
we
it
does
check
any
new
message
comes.
It
does
first
checked
that
is
it
a
new
code?
Do
I
need
to
get
a
new
code,
or
if
it
does
tavon
the
poison
pill
check,
then
it
will
restart
itself,
it
will
kill
itself
and
we
crawl.
B
The
main
runner
will
means
initiate
the
process,
but
that
the
it
will
not
going
to
happen
until
there
is
some
message
in
the
queue.
So
it
will
keep
running
again
until
they
find
some
message
in
the
queue.
Then
it
will
do
the
poison
pill
check
and
then
kill,
kill
itself.
So
the
suggestion
was
maybe
to
put
a
message
dummy
message
into
the
kill
so
that
it
does
check
the
poison
pill
or
I.
Think
the
combination
of
the
maximum
idle
time
and
and
this
feature
that
we.
B
Maximum
idle
time,
this
is
the
time
that
how
much
time
that
it
can
stay
idle,
so
that
is
configured
and
then
this
flag
only
spawn
when
messages
available.
So
this
is
the
proposal
there
on
the
curve,
the
initial
runner,
which
initiates
the
q
processes
that
can
all
be
little
bit
intelligent
and
see
if
there
is
a
message
in
the
queue,
then
only
spawn
the
job
queue
process.
Otherwise,
just
don't
even
bother
because
them
you
will
run
it
and
then
it
will
continue
running
for
a
certain
period
of
time.
Then
it
will
kill
itself.
B
So
why
bother
creating
another
process
if
there
is
no
myself
so
I
think
the
introduction
of
this
only
spawn
when
message
available
with
maximum
idle
time
will
solve
both
of
the
issues
long-running
process,
and
if
there
is
a
it,
doesn't
have
to
wait
for
a
new
message
to
to
shut
down
itself.
So
both
of
these
will
be.
B
Yeah,
it
will
be,
it
will
get
the
flexibility
to
the
side
administrator
to
configure
it
the
way
they
want
to
set
up
the
site.
So
that
was
also
the
intention
to
give
more
flexibility.
If
they
don't
want
to
read
at
all,
maybe
they
can
set
it
to
zero
and
as
soon
as
the
prod
the
to
process
will
find
that
there
is
no
message
for
their
message
to
process.
It
will
want
to
shut
down
itself
immediately,
but
the
ability
to
configure
it
will
be
on
the
administrator
of
the
side.
B
The
only
thing
there
based
on
my
I
don't
know
that
cold
very
well,
but
I
found
through
the
code.
It
looks
like
the
the
cron
job
runner,
which
initiates
the
job.
He
doesn't
have
any
intelligence
right
now
in
terms
of
checking
the
queue
and
is
morally
like
kind
of
like
to
call
that
fossil
that
dead
triggers,
so
that
process
will
need
to
have
some
kind
of
a
refactoring
done.
Some
intelligence
will
be
there,
checking
the
keyword.
There
is
a
message:
I
am
working
with
Alex
in
Germany
to
find
out
that
it
is
possible.
B
A
A
A
A
Expected
so
in
some
circumstances
it
just
doesn't
return
any
information,
so
it
returns
information
only
if
there
is
an
auto
increment
field
on
the
table
and
if
somebody
has
already
access
to
this
table
and
I
think
if
there
is
data
in
this
table.
In
all
other
cases,
it
just
returns
now,
which
means
that
you
have
just
created
a
table
and
nobody
was
reading
from
it.
You
will
get
no.
A
A
D
A
A
A
A
One
more
option
that
we
are
considering
is
disable
table
status
cash.
It's
possible,
it's
what
they
actually
recommend.
What
people
are
doing
there
is
still
one
problem
is
that
it
doesn't
give
us
necessary
result
in
some
cases
it
doesn't
result.
If
it's
a
newly
created
table
is
not
very,
it
resolves
the
issue
in
some
other
cases,
but
still
not
all.
A
D
D
E
Think
if
we
there
is
another
side
of
of
this
issue,
whether
or
not
we
want
to
do
fix
it
to
the
tree,
to
give
everyone
performance
benefits
and
if
we
can
fix
it
into
the
three
and
two,
the
three
will
become
faster
than
it
might
make
sense
to
keep
existing
problem
in
the
codebase
and
reliable
implementation,
but
make
everything
else
much
more
performant
and
included
included
four.
We
would
a
factor
all
of
the
usages
and
eliminate
dependency
on
this
of
the
increment
field.
What
did
what
do
you
think
about
that?
E
D
You
know
this
is
table
statistics
and
we
don't
want
to
run
state
table
statistic
gathering
during
our
regular
workflows.
So
when
we
need
this
method,
we
we
be
calling
this
method
during
import
to
best
in
UAD,
which
is
logically
wrong.
Okay,
let's
assume
this,
we
want
to
keep
it
in
case
we're
on.
We
will
put
analyze
table
into
entosis
logic.
We
will
round
table
statistic
rich
by
gathering
for
each
for
each
product
row
that
we
are
going
to
insert
I.
D
E
D
Poem,
that's
okay!
For
some
cases,
when
we
see
that
we
have
a
feature
gap
or
something
world
works
completely
incorrect,
we
have
to
do
breaking
changes
but
and
actually
right
now
we
don't
have
an
actual
interface.
We
have
an
array,
and
this
array
has
some
key.
We
you
just
remove
this
key
from
our
a
so
the
film
would
break
all
clients
all
suspended.
It
will
affect
only
only
only
clients
who
were
trying
to
utilize
this
method,
but
yet
this.
E
D
D
E
E
D
D
D
D
Because
this
is
two
different
databases
and
we
can
put
this
in
years
this
information
into
the
communication.
So
on
our
side
we
can
do
our
best.
We
can
fix
such
usages
to
remove
dependency
on
particular
bicycle
version
for
a
community
guys.
I
believe
this
is
good
enough
to
say.
Okay,
my
keep
my
sequel
aid
behaves
in
other
ways
and
my
sequel
5.
So
the
ever
that's
we
don't
suggest
you
to
use
of
increment
field
to
graze
in
any
UF
increments.
But
if
you
still
want
to
do
this,
be
aware,
that's
my
sequel.
C
A
D
C
C
D
F
D
D
Do
it
for
2
to
3
is
enough
to
fix
existing
places
in
code
and
I,
like
idea
to
maybe
introduce
the
interface
real
interface
for
2
to
4
and
replace
this
method
with
clean
interface,
which
you,
which
will
return,
only
values.
This
we
really
need,
and
we
want
to
expose
as
a
part
of
ajanta
database
table
metadata.
D
A
D
D
I'll
be
in
case
we
will
IIIi,
don't
think
this
is
good
idea
to
force
any
mystical
configurations
on
magenta
level
magenta.
This
is
just
an
application.
My
simple
configuration-
this
is
responsibility
of
administrator,
in
particular
merchant
we're
trying
to
force
some
values.
You
know
II,
it
can
be
unpredictable
in
de
it
can
cause
conflicts
with
merchant
installations
in
merchant
configuration
and
we
cannot
cover
all
possible
cases.
D
D
D
A
E
If
I
would
say,
if
we
can
keep
the
same
performance
for
the
scenarios
where
the
simplemente
is
used,
when
be
clear,
cash
and
other
scenarios
improve,
then
I
would
personally
vote
for
delivery
in
what
around
into
the
three,
because
I
was
naughty
us.
Improve
scenarios
where
we
rely
on
incremental
gives
the
same
performance
in
this
clear
in
cash.
It's
it
has
value
and
it's
backward
compatible.
So
I
would
propose
to
do
this
for
two,
the
three
I'm
not
sure
about.
E
A
A
C
C
A
C
F
A
A
A
D
D
D
C
Depends
it
depends
on
type
of
engines.
We
use
I
had
a
project
like
many
years
ago,
like
six
or
seven
years,
and
we
helped
they're
nice
people,
my
sound
table,
and
there
was
a
periodical
crush
of
indexes
and
the
just
analyze
table
helps
it
so
that
I
had
the
analyze
table
around
my
chrome.
But
this
is
definitely
not
that
the
solution
we
want
to
do
here
but
like
it
may
be.