►
From YouTube: 2019 11 18 Memory Team Meeting
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
B
Yes,
like
we
continue
testing
kuma
1940.
We
continue
also
sings
time
performance
annuity,
like
we
enabled
that
today
on
web
0-2,
but
it
seems
that
there
is
t
like
about
like
15
to
25,
like
it's
really
depends
on
like
what
kind
of
person
are
you
you
are
looking
for,
they
see
like
penalty
on
the
duration,
I'm
gonna
be
diving
more
that's
tomorrow,
I
mean
even
like
open,
Sam,
so
much
more
on
doing
like
working
hours
to
kind
of
divide
it
in
more
in
more
def.
A
B
C
B
A
D
B
Like
the
development
is
done,
like
you
consider
these
features,
they
finished,
it
was
not
yet
validated
on
the
break
on
static
and
production
and
technically
we
should
close
the
issue
once
we
get
to
this
validation
when
we
find
in
later
to
actually
it's
working
properly
as
designed
on
at
least
I.
Think
I
guess
this
is
fact
a
minimum
to
close
the
issue.
This
is
what
we
usually
do
as
well.
D
A
Rather
than
look
at
the
board,
because
it's
kind
of
noisy
I
put
this
query
together,
so
we
could
run
through
what's
on
12
six,
so
the
main
goal
carry
over
for
import
and
export
in
sidekick.
A
lot
of
the
work
is
already
assigned,
there's
a
few
that
are
not,
but
let's
run
through,
what's
in
twelve
six
and
make
sure
it
makes
sense
and
reprioritize
what
we
need
to
and
even
move
things
out.
So
first
one
Lexi
looks
like
you
just
open
this
one
up.
Yes,.
E
That's
in
your
issue,
while
doing
my
fix
on
particular
relations,
not
being
comported,
we
discussed
with
Josh
and
Nicole
is
that
we
could
probably
work
this
edge
case
to
reduce
the
complexity
a
little
bit.
So
that's
a
proposal,
but
I
didn't
start
working
on
it
yet,
but
we
have
an
idea
what
we
could
do.
Oh
it's
some
sort
of
follow
up.
It
doesn't
block
us
from
merging
into
Peaks,
which
is
currently
viewed
by
maintainer.
Okay.
B
B
E
E
E
This
one
is
Nicola
started
to
look
at
to
be
honest,
I
wasn't
able
to
properly
design
and
work
on
it
and
we
created
the
Nishant
like
unwrapped,
some
ideas
that
Camille
initially
proposed,
but
our
current
strategies
to
have
a
meeting
this
week
to
prepare
some
notes
and
her
meeting
discussing
how
we're
going
to
tackle
this
issue
so
somewhere
midweek
will
meet
with
Camille
and
Colette.
Whoever
is
interested
on
the
team.
Maybe
we
will
even
have
this
two
meetings
together,
I.
C
F
D
B
B
Iii
much
like
the
right
iteration
but
like
I,
have
some
follow-up
topics,
basically
like
these
three
topics
that
I
want
to
follow
up
as
a
second
iteration
to
kind
of
conclude,
these
change
search
and
because
it's
actually
like
not
mentioned
exactly
how
to
deal
with
the
temporary
storage
and
everything
else,
especially
that
we
like
he
really
want
to
move.
We
try
to
get
to
move
into
12
faster
up,
which
process
will
be
quite
difficult,
so
describing
exactly
what
we
should
avoid
doing
in
the
future.
B
This
is
my
god,
so
if
you
could
send
it
to
myself
to
me,
I
just
I
think
that
myself
to
me,
I
actually
have
a
second
at
least
I.
A
C
A
Sometime
at
the
end
of
this
meeting,
so
we
can
make
sure
that
we
talk
about
the
goals
for
import/export
and
sidekick
that
will
run
through
the
issues
as
quickly
as
we
can,
probably
until
about
eight
o'clock.
I
have
this
meeting
extended
for
another
20
minutes?
So
just
so
you
know
if
I
stop
talking
about
issues
and
switch
over
that's.
Why
looks
like.
B
B
Take
a
look
at
the
comment.
I
think
it
was
mentioned
somewhere
here.
I.
Don't
think
that
right
week,
like
it's
so
easy
to
do
it
right
now,
because
the
artisan
library
is
oh
yeah,
okay,
this
is
actually
the
comment.
One
comment
both
because,
like
we
are
still
using
our
own
Fork
of
the
of
the
library
it
was
like
done
very
long
time
ago
and
there
is
upstream
on
which
is
supporting
perform
on
T
now
or
if
case
is
like
10
times
slower.
B
So
it's
not
very
easy
to
do,
because
we
would
have
to
pretty
much
be
right,
all
quakes,
that
we
implemented
in
our
fork
to
the
upstream
librarian.
So
this
was
what
this
issue
I
created
to
serve
like
some
specific
problem,
but
like
we
overcome
this
problem
with
the
separate
endpoint
on
side
K
exporters.
So
it's
not
really
that
much
relevant
at
this
moment.
B
B
Put
looking
the
backlog
but
I
think
like
a
misanthrope
if
we
have
the
epoch
for
that
story
because,
like
changing
this,
this
library
becomes
like
bigger
and
bigger
task
and
I.
Don't
think
that
we
really
track
all
the
deficiencies
and
problems
that
we
have
with
the
current.
We
don't
really
have
lot
of
problems
in
the
current
library.
That's
the
problem,
so
there
is
no
big
incentive
to
try
to
rewrite
that
now,
but
it's
not
really
something
that
can
be
done
in
the
single
mice.
Long
answer:
Side
Story!
B
A
A
C
It's
funny
because
continued
just
created
a
story
that
was
very
similar
and
like
the
main
problem
statement,
so
this
kind
of
came
out
at
this
time.
I
came
up
at
the
time
where
we
were
looking
at
the
probably
booking.com
we're
running
into
with
the
process
kind
of
looking
up
and
the
memory
killer,
not
kicking
it,
and
so
the
TLDR
is
I.
C
C
A
B
C
D
Maybe
we
have
a
definition
of
what
AJ
means
then
we
can
all
agree
on
and
that
seemed
to
alleviate
their
concerns,
and
so
I
just
need
to
return
back
to
that
and
trying
to
drive
that
forward.
The
sales
team
announced
a
couple
places.
I
didn't
go
through
the
entire
list,
one
team
to
support
first,
but
they
didn't
seem
to
push
back
so
I.
Don't
think
they
I
really
care
talking
about
this,
but
the
support
team
did
so
and.
B
D
There
is
so
list,
as
there
is
I've
picked.
A
few
people
in
the
sales
channel.
I
haven't
gone
through
all
of
the
official
of
checklist
items
for
paying
sales
and
there's
a
stronger
min
between
me.
That's
not
going
to
impact
sales,
because,
quite
frankly,
as
support
noter,
does
that
how
many
bus
isn't
aware
of
licenses,
and
so
you
can
currently
configure
this
even
with
your
not
even
for
not
premium
sorry
documentation,
Xin
this
premium,
it
doesn't
starter
I,
think,
would
be
kind
of
fit
starter
or
premium,
but
that
starter.
D
B
Do
so
so
so,
maybe
like
so
maybe
like
a
good
place
for
us
for
focus
degrees.
It's
like
close
definition
of
I
tell
you,
because
it's
also
like
way
that
we
film
and
I'm
gonna,
be
working
on
that
as
well
finishing
some
related
part
of
that.
So,
if
we
get
like
this
creamy
described,
we
could
clearly
define
what
is
the
place
of
the
sidekick
duster
in
that
picture.
D
D
In
effect,
there
are
missing
simple
things
and
the
fact
that's
not
regarding
sales,
and
so
once
we
get
I
think
we
should
got
resolved
to
support
concerns
which
they're
ok,
with
with
my
proposed
definition
of
a
che,
and
so
then
it's
got
to
run
that
definition,
a
passed
sales
and
and
other
folks
to
get
that
sort
of
noted,
because
right
now,
as
far
as
now,
we
have
no
actual
definition
of
a
che
and
it's
just
sort
of
whatever
the
support
team
feels
like
implementing
or
can
feel
like.
They
can
defend.
I
mean.
D
C
One
benefit
outside
of
neh
a
discussion
that
I
see
is
that
we've
reduced
the
number
of
combinations
of
the
like
configurations.
You
can
gun,
run
good
weapon
and
specifically
sidekick.
So
we
would
boil
this
all
down
into
like
just
one
case,
which
just
makes
it
way
easier
to
maintain
it
easier
to
think
about
it,
because
everyone
will
be
talking
about
the
same
thing.
C
D
It's
good,
it's
a
good
point
to
ensure
they're,
aware
of
it
is
noted
in
in
the
issue,
but
I'm
not
sure
if
they
read
all
the
way
through
or
necessarily
understood
that
that
was
a
benefit
of
this.
That's
the
only
one
of
the
benefits
I
noted
in
the
issue
that
it
will
simplify
our
testing
matrix
and
the
overall
cognitive
load
of
understanding
what
the
heck
is
going
out
at
the
instance.
D
So
yeah,
no
I
I
think
it's.
How
about
I've
been
wrapped
up
in
another
item
over
in
the
search
team,
welcome
to
taste
trying
to
get
that
through
so
I
I
need
to
return
my
campus
and
keep
up
in
this
board.
So
it's
on
my
plate
and
I
think
we're
we're
through
the
the
challenging
parts
of
the
search
item.
Some
so.
D
B
B
C
A
Will
move
this
up
twelve
six
and
then
we
can
find
the
time
to
sync
up
on
some
of
the
sidekick
related
work
planning,
especially
for
twelve
seven,
so
we're
at
eight
o'clock.
So
if
anybody
needs
to
drop
off
yeah
keep
taking
the
issues
list
gun
through
thingie
chaff
I
want
to
jump
over
to
the
bigger
epochs
and
start
talking
about
that.
A
D
B
C
A
couple
of
things
several
times
and
what
they
produce
if
I
just
wasn't
useful,
especially
since
the
last
we
asked
for
like
a
process
listing,
but
he
pulled
one
from
my
time
where
there
was
no
problem.
It
was
just
a
generic
like
system
process.
This
thing,
which
was
not
helpful
so
at
this
point
I,
don't
think
we
can.
You
know,
figure
out.
What's
going
on
just
just
based
on
these,
like
back
and
forth
conversations,
we
had
worth
first
one
so
I
guess
the
two
options
were
we
need.
C
E
C
This
problem
is
happening
which
I
don't
even
know
if
that's
an
option,
because
for
them
that's
pretty
much
a
partial
outage,
because
their
jobs
get
stuck,
they
queue
up,
so
they
go
in
and
they
kill
psychic
workers
manually.
When
this
happens,
so
they
probably
also
have
an
interest
in
you
know
unblocking
this
that
these
pipes,
you
know
as
soon
as
possible
as
it
happen
and
not
get
on
a
call
with
us
and
for
like
hours
to
figure
this
out
so
I'm
I
know
I'm
a
better
last
year
like
so.
B
Like
from
the
past
experience
with
the
with
the
crucial
clients,
we
were
actually
doing
drive
the
fighting
sessions
when
it
was
in
it
because
it
was
like
the
fastest
to
really
Depok
these
kind
of
problems.
I
think
it's
like
taking
so
long
to
read
the
likely
miracle.
Now,
if
we
ask
for
that
explicitly,
we
gonna
find
a
time
to
kind
of
chunk
on
that
and
find
the
guys
because
right,
it's
not
something
that
is
happening
like
it's
happening
at
random,
but
also
quite
frequently
and
for
quite
some
time.
Okay.
B
C
B
Yes,
buy
a
spot
right
but
like
they
can
still
have
multiple
notes,
running
side,
key,
oh
yeah,
yeah
sure
so
so
then
like.
If,
if
one
note
would
be
dormant,
he
would
basically
people
to
debug
that
why
they
continue
like
processing.
B
C
By
the
way,
Camille,
how
do
you
feel
about
because
we
have
the
new
the
demon,
the
demonized
memory
color
the
day
they're
on
12.2,
so
I
I
would
assume
I'm
pretty
sure
that
I
use
you
can
tell
by
the
log
message.
This
is
what
they
use
the
psychic
middleware
based
memory,
color,
so
I'm
also
wondering
if
there's
like
one
more
thing
we
might
want
to
try
I,
don't
know
hope
like
feasible.
It
is
for
them
to
upgrade
okay
I
I.
Don't.
B
Think
that
this
trip
schwa
returns
us
at
the
behavior
so
I'm
expecting
that
they
gonna
see
the
same
problem,
because
if
you
look
the
same
signaling
the
same
shutdown
procedure
so
and
also
its
start
shutdown
in
so
doing
and
signaling
from
the
serpent
threat.
So
it's
only
different
method
of
gathering
metrics,
but
it
doesn't
structurally
change
the
way
how
it
works.
So
I
would
expect
that
they're
gonna
see
the
same
problem
as
they
are
seeing
right
now.
B
D
A
A
D
So
what
I'll
do
is
I'll
I'll,
just
since
are
open
to
a
call
what
might
help,
and
you
guys
couldn't
obviously
quickly
if
I'm
wrong
here,
but
it
may
be.
We
have
a
call
just
to
me,
talk
to
them
and
I'm,
not
sure
if
you
happened
before,
on
tougher
support
and
and
and
sort
of
the
other
arm
them
with
what
commands
we
want
them
to
run,
make
sure
they
understand
them
actually
understand
the
parameters
on
which
we
want
them
to
run
it
or,
alternatively,
if
we
think
the
only
option
is
to
get
them.
D
A
call
when
this
is
happening
is
to
maybe
even
propose
that,
on
the
on
other,
you
know
either
an
email
through
support
or
or
in
that
call
but
I'm
open
to
other
way.
If
you
think
is,
this
is
the
only
option
we
think
is
actually
getting
the
call
when
it's
happening
or
do
we
think
we
can
maybe
try
and
arm
them
with
some
debugging
tools
or
talk
through.
C
When
it's
happening
mister
point,
so
he
posted
a
PS
output,
but
it
was
not
from
a
time
where
it
actually
happened
and
and
also
I
actually
asked
if
you
could
use
a
ps3,
because
that
gives
us
the
breakdown
of
the
threat
as
well.
But
he
has
some
kind
of
different
for
the
node
ran
on
a
different
version
of
ps3
that
didn't
support,
write
output.
So
it
wasn't
super
useful
yeah,
but
I
still
think
it's
a
good
idea
to
maybe
give
them.
C
B
Am
like,
from
my
perspective,
I
am
most
interested
about
ps3
from
the
sidekick
process.
Panel
open
FD
handles
on
the
process,
and
basically
this
is
the
most
interesting
things
to
me.
Yeah
the
simplest
way.
The
simplest
way
to
get
these
data
out
would
be
asking
them.
Can
you
create
a
snapshot
of
the
prog
directory?
B
Basically,
this
would
be
right
in
the
simplest
way
of
like
after
getting
snapshot
from
these
happening
because
I,
like
AI
I,
don't
know
like
because
different
organization
they
are
kind
of
reluctant
to
execute
external
provided
binaries.
They
also
a
literary
comes,
will
provide
potentially
sensitive
data
as
like
processing
and
things
like
so
I
have
no
clue
how
it
works.
We've
we've
done
part
right.
It
seems
from
the
discussion
that
they
are
very
cautious
on
giving
us
a
lot
of
details
so
right
so.
A
Maybe
we
give
them
a
list
of
things
that
we
would
like
to
see
and
if
they
say
now,
then
then
we
set
up
the
live
session,
so
they
can
just
show
us
so
that
they're
not
actually
writing
it
down
and
putting
it
into
an
issue
so
step
on
again
write
some
instructions:
hey.
Can
you
run
this
and
send
us
the
information
and
if
they
don't
want
to,
then
we
can
set
up
the
live
meeting.
It's
all
good
yeah.
C
So
the
the
problem
with
the
instructions
is
okay:
let's
ask
him.
First,
what
environment
exactly
they're
running
the
song
I
think
they're,
unlike
a
Red
Hat
Linux
and
like
the
so
the
tools
for
instance,
that
I
had
tried
to
give
them
instructions
for
what
they
could
run
when
it
happens,
but
it
wasn't
even
the
same
version
of
the
tool
so
that
some
of
the
command-line
options
I
had
provided,
didn't
even
exist
on
their
deployment.
So
so
that
was
a
bit
tricky,
so
we
first
maybe
need
to
figure
out
like
what
they
even
on
exactly
okay,.
D
F
D
We
can
work
to
get
that
one
account
established
and
then
also
when
we
would
you
speed
to
give
the
list
of
commands
I'm
going
to
run
and
then
just
tell
them
either
ass
for
the
environment,
running
on
or
also
just
tubs
what
we
need,
and
they
might
be
the
figure
out
on
their
side.
What
the
right
commitment
options
are
like:
hey,
we're
looking
at
our
file
and
animals
we're
looking
for
this,
so
your
seal
options
may
vary
on
your
environment,
but
this
is
kind
of
what
the
information
is,
that
we
need
yeah.
D
Okay,
move
on
yeah
thanks
all
right,
so
I'll
get
that
going
and
we
can
communicate
in
that
issue,
perhaps
or
I
just
and
slack
here
with
Latino
and
Philippe.
Okay,.
A
Then
I
got
notes
here.
I
can
put
a
note
in
the
confidential
issue.
If
that
there's
a
better
place,
to
put
it
see
so
I
think
we'll
probably
have
to
take
the
discussion
about
sidekick
and
import
in
synchronous
since
Alexi
left
can
use
not
here
I'll
write
up
some
notes
and
put
it
maybe
on
the
memory
team
board,
so
we
can
better
define
those
we
can
run
through
issues
now.
Will
this
keep
running
through
the
issues
we'll
take
the
other
part
asynchronous
so.
B
A
E
C
Just
yeah-
and
you
should
groom
this
out
a
bit
more
so
because
we
so
the
context
for
this
is
that
we
ended
up
for
12.5
shipping
like
the
really
the
Bears
minimal
change.
We
could
to
let
have
Redis
connection
pool
sizes
scale
with
a
number
of
threats
under
come
on.
Just
just
do
it
have
the
ability
to
do
that,
but
it's
like
quite
messy.
The
way
it
is
all
set
up
and
so
I
had
started
to
work
in
a
refactoring.
I
wasn't
really
going
quite
in
the
right
direction
and
it
wasn't
really
covering
all
cases.
C
C
C
B
So
like
as
we
as
we
discussed,
if,
if
we
can
achieve
that,
I
would
kind
of
conclude
that
part
of
its
story,
but
given
amount
of
the
hugs
and
like
behavior,
like
places
where
we
track
for
the
different
type
of
the
service
that
we're
running
I.
Think
that
we
are
kind
of
hitting
critical
mass
at
this
moment
that
we
should
grab
that
mic
properly.
Supported
because,
like
if
you
look
at
the
amount
of
the
defied,
comma
or
defined
unicorn
or
defined
sidekick,
it
kind
of
shows
you
like
a
ridiculous
amount
of.
C
G
C
Totally
agree:
that's
how
I'm
going
about
it
now
actually
and
that
that's
what
I
meant
that
story
isn't
really
super
I
can
update
it.
Okay,
we
can
repurpose
this
story
to
actually
focus
on
the
pinna
mentions
under
that's
just
like
a
side
note
under
approach
here,
so
we
should
also
consolidate
the
test
for
the
current
name,
so
we
can
make
it
all
about
that
first.
So.
B
C
B
C
B
B
C
Sorry
like
so
that's
the
wrong
window
focus
so
I
I
said
I
forgot
where
we
were
with
that,
but
I
think
we
came
out
of
the
saying
we
wanted
to
look
into
migrating
to
the
upstream
Prometheus
first
and
then
do
this
I.
B
So,
like
I,
don't
think
that
this
is
pretty
quick,
because
the
upstream
application
obscene
right
Mary
cannot
have
the
same
set
of
the
problem
because
we
lose
these
matches
because
we
constantly
clean
the
directory
on,
let
me
start
and
since
we're
on
the
sidekick
cluster,
because
it
happening
on
the
side,
lustre
configuration
that
made
me
lose
these
metrics
and
since,
like
this,
restart
is
kind
of
like
is
connected
and
ever
ever
process.
My
works
on
the
same
directory.
This
cleanup
should
happen
by
the
cyclic.
C
B
B
C
All
right,
fair
point,
so
I
guess
the
second
point
I'm
not
sure
about,
is
how
big
of
an
issue
this
actually
is
because
correctly
from
wrong
Camilla,
but
I
think
this
only
happens
to
a
very
small
window
of
time
right
because,
as
these
workers
come
and
go
eventually
one
will
be
the
last
winner
right
who
actually
crates
that
directory.
So
I
think
only
during
like
a
rolling,
a
restart
or
something
or
restarting
workers
in
any
other
way.
C
B
And
this
is
tricky
because
I
would
say:
I
don't
know
we
because,
like
if
you
look
at
the
comments
from
the
band,
we
actually
see
right
problems
on
the
production
of
the
site,
key
metrics
and
side
keys.
Finally,
the
cluster
mode,
because
this
issue
also
happens
on
the
start.
So
if
you,
if,
for
example,
your
psyche
process
takes
10
seconds
to
start
and
you
saw
the
site,
keeper
sister
sequentially
and
they
say
indeed
I
between
them,
you
may
lose
some
of
the
mingott
files
from
the
start
up.
B
It's
it's
inherently
the
problem
of
us
like
managing
this
directory
from
the
cyclic
process,
not
from
this
provider
process.
This
is
this.
This
is
actually
their
like
the
problem.
What
we
are
facing,
but
trying
the
only
way
to
really
solve
that
once
and
for
all
is
right.
Move
this
a
complex
project
to
site
big
cluster.
Ok,
like
we
not
gonna,
move
the
to
this
I
click
duster,
because
we're
gonna
have
to
maintain
separate
Rajic
for
the
core
and
slightly
faster
I
mean
like
another.
C
C
B
B
B
D
A
He
recently
asked
if
we
could
take
this
on
and
I
said.
Let
me
think
about
it.
Basically,
so
if
we
are
looking
at
building
proactive
tools
to
indicate
to
developers
like
to
see
ICD
or
wherever
you
know
their
overall
growth
of
memory
over
time,
then
it
might
make
sense
for
us
to
own
it,
but
we're
far
away
from
that
right
now.
So
that's
why
I
was
kind
of
on
the
fence
on
whether
or
not
we
should
own.
This
I
can
see
is
owning
it
in
the
future,
but
right
now
it
doesn't
make
any
sense.
So.
D
D
D
The
foundational
work
that
is
still
dry,
you
know
until
it
got
working
so
like
this
one
here,
I.
B
Tricky
part
about
migrating
is
not
so
straightforward.
It's
not
it's
not
like
a
drop
in
like
from
the
interface
perspective.
It's
a
drop-in
replacement,
and
maybe
this
is
Ray.
The
first
step
is
to
align
our
four
to
have
the
same
interface
as
upstream
library,
but
we
cannot
really
speak
today
to
the
upstream
library,
because
it's
not
performant
it's
this,
the
biggest
issue
and
all
of
the
problems
that
we
have
with
our
library
and
like
the
way
we
handle
that
they
are
not
solved
by
the
upstream
library
anyway.
B
D
B
I,
don't
think
that
this
is
accepted
with
it
like
we
switched
the
upstream
library
and
I
say
that
it's
like
even
like
ten
seconds
significantly
worse,
and
what
we
have
today
and
like
even
two
seconds
is
causing
us
like
some
troubles
with
running.
So
that's
that's
the
tricky
part
it
like.
If
we
gonna
move
to
the
upstream,
it
has
to
have
the
same
performance
benefits
as
ours.
It
doesn't
okay,.
A
A
A
Work
to
do
right
now,
yeah
yeah
I'm,
seeing
a
pattern
here,
like
maybe
a
sidekick
next
epic,
with
a
bunch
of
related
issues
that
could
include
this
one
as
well
so
yeah.
You
and
I
can
sync
up
later
this
week,
but
we
are
pretty
much
at
time.
So
what
I
would
ask
is
have
everybody
look
through
issues
that
are
assigned
to
them
before
12
6
make
sure
that
they
make
sense
that
you're
not
overloaded?
A
And
if
you
are,
then
we
can
have
any
synchronous
conversation
about
it
or
just
move
things
that
you
know
you're
not
going
to
touch
in
twelve
six
and
then
I
will
spin
up
some
issues
to
talk
about
the
goals
for
import/export
in
sidekick.
I'll
probably
do
them
separately
in
the
memory
team
project
so
that
we're
not
spamming
everyone,
that's
watching
those
other
issues
and
then
once
we
actually
get
the
goals
and
objectives
for
those
two
epics,
then
we'll
all
call
copy
them
over
to
them
to
the
main
epics.