►
From YouTube: Memory Team Weekly
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
It
is
monday
march
29th,
and
this
is
the
memory
team
weekly
meeting,
not
verbalized.
I'm
gonna
go
ahead
and
change
this
topic,
because
we're
gonna
talk
about
all
these
things,
so
retro
feedback's
due
this
week,
I've
already
seen
some
comments
going
back.
So
thanks
to
those
who've
already
contributed,
I
think
the
live
stream
is
actually
sometime
next
week.
A
A
B
Okay,
yes,
so
no!
This
is
just
a
small
issue,
like
I
found
out
that
we
have
some
broken
specs
if
the
load
balancer
is
enabled.
So
I
just
pushed
the
mr
that
fixed
it.
In
fact,
everything
was
working:
fine,
because
on
local
environment
and
rci,
the
load
balancing
is
not
configured.
So
the
specs
were
like
passing
just
fine,
but
in
case
that
someone
like
configured
for
tests,
those
holes,
the
specs,
will
be
broken.
So
I
just
just
fixed
the
spec.
C
Yeah,
that
was
so
so.
This
was
the
issue
where,
like
whenever
there
was
a
pipeline
abuse
incident,
the
block
call
in
the
admin
ui
would
frequently
time
out,
because
it
was
just
particularly
those
users
that
have
used
pipelines
that,
just
when
we
had
to
like
cancel
a
lot
of
the
work
they
did
and
unfortunately
that
was
then
exactly
when
it
tended
to
time
out.
So
it
was
very
annoying.
D
C
The
folks
who
like
handle
these
incidents,
because
they
would
have
to
go
in
manually
and
fix
that
so
so
yeah
that
worked.
We
turned
that
into
just
a
bulk
update.
It
was
basically
a
really
expensive
one
by
one
update
before
yeah.
It's
just
like
there
is
something
like
this.
C
Some
follow-up
work
to
do,
because
the
reason
we
are
able
to
both
update
that
is
because
we're
not
skipping
active
record
handlers,
so
there's
some
like
posts
like
cancellation
books
that
are
not
firing
right
now,
but
we
said
it's
like
okay
to
not
do
that.
Work
for
now,
just
to
make
sure
that
things.
C
A
Oh,
this
is
great.
It
looks
like
it
aligns
closely
with
the
database
initiative
that
I
was
describing
earlier.
So
thank
you
and
then.
D
C
And
the
next
one
that
had
been
open
for
a
while.
I
that's
just.
C
A
kind
of
boy
scouting
thing
I
just
noticed:
we
have
like
this
function
that
we
use
to
limit
result
sets
in
extra
record
calls
and
it
was
kind
of
basically
an
alias
for
something
that
already
existed
and
it
was
bypassing
the
cop.
So
instead
we
changed
the
gitlab
style
rules
to
actually
allow
limit
clauses
outside
of
active
record
models,
because
that
seems
sensible.
You
should
always
be
able
to
live
in
a
service
structure.
A
Thank
you.
That's
great
camille
yep,
this
one's
done.
I
think
we
need
to
talk
about
this
one.
Then
blog
posts
need
to
release
post
eventually,
all
right
and
I'll
make
same
one
yeah.
C
Yeah
I
mean
like
we
talked
about
this,
so
much
like
this
was
just
like
verifying
that
our
own.
E
Thanks
for
for
pushing
this
through
by
the
way,
I
I
mean
I
don't
know,
was
there
any
feedback
in
a
checkpacker
one?
That
was
nothing
anybody
anything
we
received
on
it
yet.
A
No
camille
put
it
in
the
ceo
channel,
I
don't
know
camille.
Did
you
see
any
feedback
on
that.
C
E
Yeah
and
apologies
again,
I
actually
like
we
were
late
with
the
release
post
and
then
there
is
obviously
a
new
process
for
merging
things
into
the
release
post,
and
I
think
I
followed
that.
But
then
nobody
really
and
including
me.
I
didn't
really
announce
that
I
had
done
that
and
I
think
that
caused
some
confusion.
A
B
My
turn
yeah,
so
we
enable
load
balancing
for
sidekick
currently
on
staging,
like
we
created
two
separate,
mrs
to
enable
it
on
kubernetes
and
on
vms,
and
it
will
require
as
well
to
enable
it
later
on
production
when
we
roll
it
out.
But
I'm
not
sure
should
be
like
split
this
task,
because
this
is
one
more
general
one
to
enable
this
feature
flag
to
different
environments.
But
currently
it's.
D
C
Yeah,
that's
a
good
question
as
well
yeah.
I
would
like
it's,
I'm
not
a
big
fan
of
this
change
management
process.
To
be
honest,
it's
like
half
of
the
questions
in
that
long
issue
template.
I
could
not
even
answer
myself,
but
that's
what
they
use.
So
I
would
really
recommend-
and
they
also
asked
us
to
please
use
this
going
forward,
so
I
would
recommend
to
use
that
to
change.
It's
called
change
management.
There's
an
issue
template
for
this
on
the
infrastructure
tracker
and
then
you
need
to.
C
They
would
basically
ask
you
put
in
all
the
like
links
to
observability
dashboards
and
what
are
your
expect
expectations
you
know,
and
what
do
we
do
if
it
fails
and
then
you
can
you
and
then
sre
will
pick
this
up
and
then
that
is
actually
great,
because
then
you
have
a
point
person
that
you
can
do
this
together
with
that's
what
I
did
for
what
was
it?
B
F
We
will
so
you
have
that
in
works,
or
this
is
something
that
you
are
not
foreseeing
after
me
asking
about
this
loving
issue.
What
is
like,
you
are
talking
about
the
future
or,
like
you
are
talking
about
the
present.
B
The
future
the
the
future
like
do
we
want
all
of
that,
because
there
there
are
some
separate
issues
that
will
probably
introduce
some
of
the
logging,
because
I
I
noticed
that
stan,
who
is
changing
like
the
code
up,
how
it
works
and
it
will
introduce
probably
some
of
the
login
there.
So
I
don't
want
to
duplicate
the
work,
so
I
will.
F
F
F
Trend
management
issue
for
the
production
gonna
require
us
to
post
this
metrics,
and
we
want
to
see
like
the
full
of
behavior
of
that
how
many
times
we
make
a
decision
to
use
or
delay
to
use
primary
or
not.
We
can
like
right
now
to
some
extent,
observe
db.
Radicals
and
db
primary
counts
on
the
side
key,
but
it's
not
very
clear
information
to
like
to
use
that
for
the
production.
So
we
don't
really
need
a
lot
of
metrics.
F
C
Okay,
yeah,
and
that's
actually
that's
actually
why
I
like
that
we
have
this
change
management
process,
even
though
it's
it
could
be
more
lightweight.
I
guess,
but
it
kind
of
forces
you
to
think
about
when
when
can
I
actually
call
this
successful
right
or
how
do
I,
how
do
I
decide
that
it
should
move
from
staging
to
canary
or
when
it
should
move
from
canary
to
production
and
so
forth?
So
it
kind
of
helps
as
well.
Yeah.
C
B
Also,
this
change
will
not
like
did
this
environment
variable
will
still
not
enable
workers
to
use
the
sidekiq.
We
have
the
separate
roller
process.
C
C
F
C
F
C
F
Maybe
that's
like
our
future
roll
out
in
place
got
a
little
more
verbals.
Like
a
few
minutes
ago,
I
met
a
chancellor.
C
Do
it
yourself
is
what
I
mean
but
like
if
I
want
to
change?
Let's
say
I
want
like
like
when
we
did
the
gc
thing.
I
think
no.
Actually,
that
was
for
the
the
the
m
advice,
the
setting.
We
actually
enabled
it
first
on
a
particular
cluster
of
the
like
a
yeah
like
a
sub
portion
of
the
entire.
C
Api
fleet,
I
think
it
was
or
the
websockets
fleet.
This
is
something
you
can't
do
that
yourself.
You
need
to
work
with
an
sre
because
they
that
goes
through
the
pipeline
that
manages
these
overrides
for
our
health
charts
that
we
apply
to
sas.
So
that's
where
the
stuff
really
helps.
I
find
that
helps.
D
Yeah,
I
have
an
update
for
this
query
if
it
turned
out
to
be
more
complex
than
it
looked
like.
It
looked
like
a
lot
of
hanging
fruits,
but
they
are
not
so
we
paired
with
adam
last
week
and
today
this
fabian
and
to
be
shipped
one
small
index,
but
unfortunately
it
provides
rather
marginal
improvement.
C
Yeah,
that's
like
one
of
the
mega
issues
in
this
in
this
epic.
They
were
like
at
this.
This
end
point
it's
just
there's
like
no
end
of
problems
with
that
and
like
we've
made
steady
progress,
I
think,
and
that
wasn't
just
all
me.
I
think
I
think
there
were
like
five,
mrs
attached
already
to
that
issue.
I
just
opened
another
one.
Today,
yeah
like
this
empire
was
like
right,
but
I
actually
begin
to
think.
C
Maybe
it's
getting
a
little
convoluted,
because
maybe
we
should
have
broken
out
some
issues
or
something,
but
that's
why
it's
been
open
for
like
over
a
month,
but
it's
still
ongoing
like
there's
still
more
we
can
do.
I
think
we
just
need
to
find
a
good
cutoff
point
with
this.
One,
like
you
know,
kind
of
what
is
good
enough
because
it
just
is
very
complicated.
C
It's
like
this
massive
tree
of
things
that
we
load
and
then
ee
behaves
a
bit
differently
from
ce
based
on
like
the
kind
of
licenses
you
have
and
whatnot,
and
it's
it's
quite
complicated
but
yeah.
I'm.
C
Some
progress
there,
but
I
think
at
some
point
maybe
the
return
on
investment
isn't
that
great
anymore.
So
maybe
I'll.
C
Get
a
stopping
point
this
week.
I
think
okay,
there's
also
like
one
really
expensive
query
like
with
these
build
artifacts
like
I
talked
to
camille
about
this
already.
C
I
actually
think
at
this
point
because
in
parallel,
there's
there's
some
traction
now
behind
an
issue
where
we
actually
suggested
to:
let's
just
not
do
that
anymore
and
make
that
completely
lazily
from
a
user
perspective,
so
that
we
don't
even
have
to
worry
about
this
because
it
currently
preloads,
sometimes
thousands
of
build
artifacts
or
the
metadata
for
those,
because
it
needs
to
render
a
download
button
because
you
might
click
this
right.
C
So
there
was
some
back
and
forth
like
we
don't
really
know
how
often
that
is
actually
clicked,
but
pms
seem
to
be
on
board
with
just
saying:
let's
just
make
that
lazy
lazy
so
so
that
there's
maybe
an
additional
click
necessary.
But
then
it's
just
another
request
to
fetch
the
artifacts.
Only
for
this
one
pipeline,
you
clicked
on
which
I
think
makes
so
much
sense,
and
if
we
do
this,
then
we
don't
need
to,
and
this
query
will
just
not
run
as
part
of
the
tab.
I
think
that's
the
best
solution,
honestly.
A
That's
great
yeah.
There
were
some
other
options
where
they
discussed
moving
to
lazy
loading
in
some
of
those
query,
optimization
research
issues
that
were
covered
so.
C
And
the
nice
thing
is,
this
would
be
great
from
a
memory
perspective,
not
just
database
pressure,
because
we
found
that
I
think
I
think
it
was
70
or
so
of
the
json
render.
It
was
just
a
built
metadata
so
that
this
will
be.
This
will
shrink
memory
usage
like
massively
for
com
on
the
on
the
on
the
big,
not
all
pipelines,
obviously,
but
like
the
gitlab.
F
C
F
C
C
B
A
All
right,
camille,
yep.
F
F
F
E
And
I
I
didn't
do
very
much
right,
but
I
I
consolidated
all
of
the
comments
in
there
and
I
also
created
sort
of
the
first
like
sample
issue
right
here
for,
like
api
version,
internal
allowed.
I
think
that
was
one
of
the
candidates.
A
F
F
Heavy
endpoints
I
mean
especially
it
can
be
useful,
given
that
there
is
like
a
bounce
of
the
focus
on
the
performance
right
now.
So
it
may
be
also
like
the
right
time
to
push
that.
E
Yeah,
I
think
what
would
be,
I
think,
the
sort
of
maybe
medium
term
if
we
can
create.
I
mean
there
was
a
discussion
recently
with
sort
of
infradev
issues
right
if
we
have
automation
at
some
point
where
we
can
say
hey.
This
is
a
query
that
or
an
end
point
that
uses
more
memory
than
we
want
right
and
it's
new.
E
F
Okay,
see
next
one
I
I
have
to
drop.
I
have
another
meeting
so
just
continue
without.
A
B
Yeah
the
next
one
is
mine,
like
this
is
the
follow-up
item
from
like
introducing
the
read-only
database
replica
for
sidekick
there.
There
are
some
open
discussions
like
to
provide
the
better
retry
mechanism
and
to
introduce
some
caching
so,
but
before,
like,
I
continue
my
work
on
this,
I
will
probably
first
address
those
metrics
that
camille
mentions.
So
I
will
probably
pause
this
until
I
finish
the
the
metrics
first.
A
That's
it
that's
the
board.
I
don't
know
if
there's
anything
on
here.
B
I
just
want
to
mention
that
like
allow
sidekick
jobs
to
use
really
only
database
replica
is
currently
in
the
open
column,
but
actually
it's
like
currently
workflow
is
set
to
feature
flex,
so
I
guess
it
should
be
closed
soon.
Currently
we're
verifying
it
on
the
staging,
but
it's
still
not
gonna,
be
deployed
to
the
production,
so
it's
somewhere
in
between,
but
it's
currently
it
shouldn't.
B
C
B
C
Wanted
myself,
because
I'm
also
working
on
a
rollout
issue
and
someone
else
moved
it
into
workflow
verification
which
I
had
never
used
before.
I
think
that's
actually
the
label
we
use
for
something
that
is
done
and
it's
like
being
verified
in
production.
I
guess,
but
you
it's
not
closed
yet
because
you're
not
100
sure,
yet
that
it's
working.
I
know,
I'm
not
sure
if
we
have
this
column
but
like
that's,
maybe
where
they
all
are
now.
A
B
F
C
E
Don't
believe
so
I
think
you
have
to
edit
manually.
I
think
we
use
it
sort
of
sparingly
in
geo
for
exactly
what
you
just
said.
It's
like.
B
I
just
wanted
to
ask
about
this
specific
item,
because
we
have
a
rollout
issue
and
we
will
have
also
the
separate
issue
for
like
that
environment
variable,
and
I
know
that
there
was
an
open
discussion
when,
like
we
have
the
separate
rollout
issue,
do
we
close
the
original
one
how
we
track
the
progress
like?
Is
everything
stays
open
until
the
rollout
issue
is
closed
or.
A
E
B
A
E
No,
I
think
I
think
it
sounds
like
maybe
converting
it
into
an
epic
with
some
issues
is
the
right
thing
to
do.
Okay,
yeah,
I
mean
it's
like
this
is
also
gitlab.
Has
support
for
modeling
these
sort
of
issue
networks
right,
which
is
actually
able
to
display
more
complex
relationships
between
things,
but
I
personally
find
that
it
is
a
lot
easier
to
just
provide
a
hierarchical
bucket
organization
which
is
like,
let's
put
stuff
in
here.
E
A
A
D
D
D
A
D
A
A
Okay,
if
you
find
yourself
with
work
that
you
need
to
do
like,
if
you
don't
have
anything
on
the
board
or
you
can't
contribute
to
what
nikola
is
working
on,
then
yeah,
you
can
jump
over
here
and
start
grabbing
from
these,
and,
if
you
see
especially,
if
you
see
a
manager
assigned
to
any
of
these,
feel
free
to
just
take
it
from
them,
because
they're
probably
still
trying
to
find
an
individual
to
work
on
it
and
just
let
them
know
like
hey,
I
haven't
been
with.
I
can
pick
this
up
from
you:
okay,
okay,
thanks.