►
From YouTube: 2023-07-06 Scalability Demo
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
All
right,
this
is
the
scalability
demo
and
Sylvester.
You've
got
the
first
and
the
currently
only
item.
B
Right,
okay,
let
me
share
my
screen.
B
I
think
most
of
us
know
that
chat
needs
to
go
to
reddit's
clusters.
Soon
we
have,
we
have
saturation
forecaster
and
it's
not
getting
better,
but
the
traffic
system
so
I
think
there
was
this
question
on
whether
what
should
we
have
any
short
term
visual
process
for
based
on
this
thread,
which
is
linked
here.
We
we
were
hopeful,
so
we
we
because
last
week
the
it
went
down,
you
can
see
it's
every
every
spike
is
a
week,
so
I
think
last
week
it
went
down
and
we
wanted
to
observe
the
Civil
classes.
B
Everybody
did
not
so
I,
don't
think
we
can
waiting,
waiting
and
hoping
it's
not
gonna
work,
so
we're
gonna
start
doing
something
which
is
also,
which
is
very
likely
going
to
be
some
sort
of
a
short-term
like
shutting
out
of
another
workload
and
I
was
walking
around
with
the
main
epic,
this
one
through
five
Financial,
which
is
not
formed
yet,
but
I
have
an
issue
under
it
which
tracks
all
the
all
the
components
of
real
set,
which
needs
to
be
my
which
needs
to
be
fixed.
So
that
is
when
it's
faster
compatible.
B
So
one
big
one
that
I
found
out
today,
while
fixing
NMR
so
I've,
opened
this
Mr.
But
I
did
one
simple
change
it
and
which
is
this
one
I
added
shared
State
into
the
radius.yaml,
so
it
would
run
all
it
would
run
shots
the
against
a
radius
cluster,
which
means
everything
that
is
not
compatible
with
brick.
So
the
pipeline
will
just
be
ready
and
I
can
click
through
and
see
what's
broken
and
the
goal
is
to
fix
those,
so
I
fixed
bunch
of
them-
and
there
are
some
that
are
not
not.
B
B
Yeah
also
trying
to
add
we're
trying
to
introduce
hashtags
so
that
it's
red
is
cluster
compatible.
So
these
efforts
are
going
on
in
the
background,
but
one
of
the
major
blockers
I
found
today
was
that
pops
up
has
a
gold
dependency
which
is
inverse
Dependable
cause
the
the
goal,
library
that
it
uses.
B
The
gold
Library
users
does
not
support
redis
cluster,
which
is
travel
zone.
But
the
thing
is
it
for
pop
sub
modules.
You
don't
need
to
have
a
radish
cluster,
because.
B
You
don't
technically
need
one
because
very
slutter
clients
publish
to
everyone
so
as
long
as
you're
as
long
as
the
subscriber
is
connected
to
any
of
the
nodes,
the
subscriber
will
get
the
message,
but
Workhorse
also
does
a
get
request,
so
they
get.
Requests
will
need
to
be
ready
to
clustered
like
that.
The
client
needs
to
be
valid
clustered
valuable
because
it
will
likely
get
a
moved,
redirection
and
the
client
needs
to
know
how
to
shift
to
the
next
the
the
right
node.
B
As
part
of
this
epic
net,
it
would
increase
the
complexity,
I
guess,
because
it's
basically
changing
stopping
out
a
gem
and
not
a
jam.
Sorry
A,
Go,
Go,
Library
dependency,
yeah
so
and
I
was
looking
back
at
this
issue.
Pops
up
is.
B
B
Does
they
got
quite
a
large
chunk,
but
as
to
like
the
proportion
of
the
CPU,
we
probably
need
a
flame
graph
or
a
deeper
dive
so
to
better
estimate
if
it's
worth
slitting
out
pops
up,
but
now
I'm
just
discovered
an
additional
reason
why
it
might
be
good
if
we
move
Parts
about,
in
addition
to
saving
some
like
buying
something.
We
also
do
not
need
to
upgrade
the
the
goal.
That's
not
upgrade
but
support
the
goal,
dependency
on
low
cost,
so
that
keeps
things
a
little
simpler,
yeah.
B
Yes,
sorry
about
I
was
adjusting
the
volume
when
you
started
so
could
you
repeat
so
yeah.
C
Sorry
so
the
casting
that
I
discovered
like
last
week
or
something
like
that
I
believe
that
also
pops
up
workload
isn't
it
so
maybe
it
doesn't
all
go
into
the
same.
B
B
Right,
the
one
is
for
kubernetes
in
general
hold
on
hold
on.
Yes,
that
one
doesn't
run
on
Rails
right
if
I'm
not
wrong.
B
B
B
A
A
I
know:
we've
done
some
optimization
to
kind
of
buy
us
some
time
over
the
last
year
or
so,
but
we're
again
bumping
into
the
the
limitations.
My
question
would
be.
B
There
is
like
shuttered,
publish
evidence
yeah,
so
this
one
is
small.
This
is
Charlotte
pops
up.
It's
meant
for
radish
cluster,
so
I'm
not
sure
how
this
migration
process
would
be
like
wouldn't
be
a
two-step
thing
or
like
cut
over
thing,
but
yeah
I
have
to
read
out
more
about
it,
but
I.
Think
yeah
it'll
be
good.
If
you
can
move
into
organized
cluster,
otherwise
it
will
just
become
a
time
bomb
like
we
will
have
to
do
a
bit,
maybe
a
year
later
or
two.
B
All
right,
yeah
in
any
case
I
I've
painted
Stephanie
and
Ariana,
is
called
so
I'll,
probably
wait
for
more
more
news
for
them,
but
I
think
this
is
the
general
direction.
So
the
next
step
is
just
to
figure
out.
Is
it
worth
like?
Is
there
a
lot
sizeable
enough
and
and
what
people
mentioned
the
strategies
to
move
it?
Are
we
going
to?
Is
it
just
a
simple
move
or
we're
going
to
move
it?
B
Oh
tomorrow,
with
time
zone,
wait
I'm
trying
to
call
in
time
zones,
but.
B
Hours
time
we'll
be
doing
a
production
migration,
let
us
start
of
it,
which
is
the
Dual
writing,
but
then
that
is
still
over
the
weekend,
because
all
the
keys
are
TTL.
We
will
observe
the
Friday
if
anything
goes
back,
we'll
roll
back,
but
if
it's
smooth
then
we'll
let
it
next
week.
So
if
all
goes
smoothly,
we'll
be
running
soon,.
B
C
B
We
have
those
like
we
have
like
I,
think,
probably
10
million
or
20
million
Keys,
which
are
pipeline
status,
Keys,
which
we
can't
TTR,
because
there's
just
too
many
of
them
and
yeah
I
think
yeah.
It
would
be
nice
if
you
can
get
it
out
so
I'll
discussing
is
definitely
depending
on
the
state
of
the
redis
migration
Mr
like
how.
How
ready
is
it
if
it
goes
out
and
it
can
be
run
before
we
do
before
we
start
running
the
migrations
group?
B
We
might
delay
the
read
card
over
just
so
that
we
have
TTR
the
keys
and
so
that
we
can
reduce
the
time
which
the
validation
skill
runs.
Because,
right
now
we
don't
have
a.
We
don't
have
a
very
good
idea
on
how
long
it
takes
to
run,
because
when
we're
running
on
the
staging,
we
could
wrap
up
the
rate
to
as
fast
as
possible,
because
the
staging
not
production,
which
has
to
find
a
comfortable
rate
that
doesn't
push
CPU
too
much
even
through
the
peak.
So
that
might
extend
the
length
of
the
validation
so
yeah.
B
But
the
idea
is
that,
when
the
violation
script
starts
running
every
step
like
from
then
on,
like
let's
say
it
starts
at
Point
T
from
time
T
onwards,
a
new
keys
should
be
dual
written,
so
we
don't
have
to
worry
about
those
we're
just
migrating
the
keys
that
haven't
been
touched
since
since
the
start
of
the
Dual
right
yeah,
because
we
have
here
that
ttrs
are
one
week
or
like
two
weeks
in
cash,
which
is
pretty
long
yeah.
B
B
Yeah
we
can
try
to
ignore.
So
that's
the
thing
Reddit
scan
doesn't
let
you
do
an
inverse
solver
inverse
period
you
have
to.
You.
Have.
C
B
B
C
A
B
Think
the
structure
the
structure
might
need
work,
but
overall
the
logic
is
all
right.
So
it's
more
of
structuring
and
yeah
things
like
that.
So
I
think
my
thing
met
from
database,
not
not
metformin
yeah,
so
we
look
at
it
yeah.
Let's
stop.