►
From YouTube: 2021 05 04 Database Group 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
Let
me
share
my
screen,
so
it
is
tuesday.
May
4th-
and
this
is
the
database
team
weekly
meeting
jose
is
not
here.
I
looked
at
the
infradeb
issues.
There
was
nothing
new,
so
I
think
we're
in
good
shape.
He
did
have
a
request
and
it's
probably
mostly
directed
to
you:
can
we
andreas?
Can
we
pause
re-indexing
during
the
upgrade
the
upgrade's
happening
this
weekend.
A
Migrations
and
then
andreas,
you
had
the
next
topic.
B
Yeah
I
saw
that
issue
about
the
cybertech
training
coming
up
for
a
question
about
what
should
be
the
agenda.
I
was
wondering,
as
far
as
I
remember
the
last
time,
cybertech
did
a
lot
of
stuff
on
postgres
itself,
like
normal
postgres
stuff.
B
I
was
wondering
if
it
would
be
worth
looking
for
something
beyond
that.
Perhaps
in
addition
to
that-
and
I
didn't
really
know
what
kind
of
budget
we
have
or
if
that
was
a-
was
an
option
at
all
like
go,
and
particularly
look
for
something
that
gives
us
a
little
bit
insight
into
how
you
would
approach
sharding
with
postgres,
for
example.
Perhaps
that
is
something
that
we
can
even
ask
cybertron
tech,
what
they
would
suggest
yeah,
I'm
a
little
bit
afraid
they
might
not
have
something.
But
I
don't
know.
A
Yeah,
so
I've
only
had
one
conversation
with
the
same
people
at
cybertech,
so
everybody
that
worked
with
us
last
year
is
available
this
year
and
I
specifically
asked
for
so
lawrence
was
the
trainer
last
year
he's
available
he'll
there
there's
times
of
summer,
when
he's
not
available,
so
we'll
just
need
to
make
sure
we
coordinate
with
when
he's
available-
and
I
indicated
like,
let's
start
with
last
year's
agenda,
there
are
some
things
we'd
like
to
drop
and
some
requests
that
we
will
have
for
the
new
agenda
so
and
patrick
was
totally
open
to
it.
A
But
it's
probably
going
to
depend
on
how
comfortable
lawrence
is
with
training.
So
I
will
put
together
an
issue
I'll
find
last
year's
agenda
and
content
I'll
put
it
in
there
and
then
we
can
just
kind
of
say,
keep,
modify
and
drop
and
then
suggestions
to
add
new
topics.
And
then
we
can
present
that
cyber
tech
and
see
what
they
can
do.
They
were
pretty
open
to
modifications.
Last
year,
like
we
added
in
the
ruby
on
rails,
section
active
record
and
they
just
had
to
find
someone
within
their
staff.
A
They
could
actually
do
that
piece
of
presentation,
so
they
are
open
to
it
just
caveat.
We
haven't,
got
the
budget
approval
yet
so
it's
still
we're
still
hoping
to
get
it.
I
I
think
there
is
support
at
christopher's
level
to
actually
do
this.
It's
good
timing
to
ask
for
it
with
database
issues
that
we've
been
having,
so
we
don't
have
approval
yet,
but
I'm
gonna
continue
to
assume
that
we
will
eventually
get
approval,
we'll
get
the
agenda
together
and
reach
out
to
cybertech.
B
Primary
immigration,
something
that
carmilla
pointed
out
was
that
there
are
certain
kinds
of
things
we
want
to
also
look
at
for
for
the
primary
key
migrations
or
models,
because
we
have
been
looking
at,
for
example,
the
ci
builds
table
and
looking
at
foreign
key
constraints
that
are
well
defined
in
the
database
and
then
figuring
using
that
to
figure
out
what
other
tables
do
we
need
to
touch
to
migrate
for
the
primary
key.
B
What
kami
pointed
out
was
that
there
is
at
least
two
or
three
already
known
models
that
are
that
have
a
polymorphic
association
to
our
ci
bolts.
In
this
case,
I
think-
and
that
means
we
don't
have
a
foreign
key
constraint
defined
on
a
database.
So
there
is
more
than
more
things.
We
would
need
to
look
at
make
sure
that
we
catch
all
those
things
that
are
using
pointing
towards
ci,
builds
and
other
tables,
and
I
wasn't
quite
sure
if
we
are
in
a
sort
of
a
good
position
to
do
that.
B
I
think
we
can.
We
can
look
around.
A
B
But
was
wondering
if
we
should
perhaps
also
have
somebody
in
the
verify
group
looking
through
that
because
they
know
their
model
best.
I
think-
and
there
is
coming
from
that
as
well,
if
we
don't
catch
all
the
the
places
where
we
need
to
to
migrate
the
primary
key
done,
this
is
going
to
stop
working
as
soon
as
there
is
a
value
that
is
high
enough,
basically,
which
might
not
even
happen
right
away,
but
happens
sometime
later,
which
might
come
in
as
a
surprise.
B
So
we
should
be
rather
like
diligent
with
that
and
fabian.
You
asked
about
a
specific
ask.
I
think
it
would
be
great
if
we
could
have
someone
sort
of
review.
What
tables
do
we
need
to
migrate
to
avoid
this
overflow
happening.
C
Yeah
no,
like
my
my
point,
is
and
sorry
my
airpods
are
charging,
so
my
I
hope
you
can
hear
me
all
right.
So
if
we
can,
if
we
can
be
very
crisp
on
you
know,
we
need
to
you
need
to
look
at
these
tables
right,
because
we
need
that
kind
of
information
for
the
primary
key
migration.
If
we
can
complete
that
by
this
date
right,
we
have
an
issue
with
our
our
database
toppling
over.
I
think
we
have
a
pretty
strong
case
for
saying.
C
Please
go
look
at
this
right,
but
the
more
concrete
we
can
be
in
terms
of
what
we
want
people
to
do
and
by
when
we
want
them
to
do
it,
the
easier
it
will
be
for
them
to
to
say
yes,
we'll
find
a
person
in
the
team
and
do
it.
So
I
think
also
I
just
checked
there
is
the
engineering
and
location
thing
right
for
stage
groups
and
I
think
ci
verify
is
supposed
to
allocate
at
least
10
percent.
B
I
don't
think
that's
a
scalability
question
I
mean
ultimately
it
is,
but
it's
an
existential
question
like
if
we
miss
those
things,
then,
as
soon
as
the
overflow
happens,
it's
gonna
break
like.
C
C
C
B
I'll
create
an
issue
later
today
and
pinging
on
that
messaging
wise.
I
think
it
would
be
just
great
to
have
somebody
with
with
the
domain
inside
in
as
a
second
second
or
as
a
pair
of
eyes.
Basically.
A
I
lack
of
better
terms
so
we'll
create
an
issue
per
table
being
migrated
to
have
an
expert
within
that
stage
review
to
make
sure
that
we're
not
missing
anything
so
like
verify.
I
would
tag
craiger's
on
that,
because
he's
gonna
have
the
most
knowledge
and
be
able
to
jump
right
in.
I
know
he's
overloaded,
but
as
as
andreas
put
it,
this
is
an
existential
problem
and
we
don't
have
a
lot
of
time
to
mess
around,
so
it
should
probably
be
prioritized
pretty
quickly.
A
A
And
let
me
let
me
know
if
you
need
help
on
creating
those
issues
andreas
I
have
some
time
later
on
today.
I
can
help
crank
those
out
thanks.
A
A
question
came
up
so
moving
on
to
number
four.
If
we're
done
with
number
three
on
the
migration
testing
prioritization,
there
was
a
question
on
whether
the
issues
in
the
epic
are
in
priority
order.
It
came
up
when
alex,
and
I
were
talking.
A
B
Oh
sorry,
I
was
on
the
wrong.
I
thought
it
was
about
the
primary
case.
Migration.
Sorry!
No!
I
don't
think
this
is
in
priority
order,
there's
any
like
priority
defined
on
that,
probably
also
not
that
there
is
any
like
dependencies
between
those
things.
E
Yeah,
my
understanding
of
the
issues
that
are
left
in
this
epic
is
that
we
would
like
to
do
all
of
them
for
v1
and
I
don't
think
any
of
them
depend
on
each
other.
So
I
guess
I'm
just
re
restating
what
andreas
just
said.
So
sorry.
A
A
So,
looking
at
the
billboard
across
that's
the
first
one
that
one's
underway
andreas,
you
got
the
next
one.
B
Optimizing
batch
sizes,
that's
been
running,
that's
been
adjusting
the
batch
size
automatically,
depending
on
how
much
we
can
do
in
two
minutes.
Basically-
and
sometimes
I
like
to
think
it
works
quite
well
and
sometimes
the
database
load
is
just
very
random.
So
it's
it's
like
it's.
It
keeps
adjusting
stuff,
but
we
don't
know
how
good
it
is
really,
and
sometimes
it's
just
the
assumption
breaks
that
that
you're,
you
know
you're
seeing
more
consistent
database
load
and
that's
not
always
the
case.
It.
B
I
think
very
useful
because
we
can
just
use
it
for
self-hosted
and
they're
gonna
like
increase
their
batch
size
as
much
as
they
can.
So
we
don't
have
to
figure
that
out
and
there
was
an
additional
request
coming
recently
over
the
weekend.
B
So
in
addition
to
basically
trying
to
maximize
the
throughput
for
the
migration,
we
should
also
be
mindful
about
the
state
of
the
database
in
a
sense,
so
we
had
a
situation
where
we
basically
had
an
alert
going
for
the
dead,
tuples
ratio
on
the
table.
So
we,
you
know
we
keep
updating
a
lot
of
tuples
on
the
table
and
in
some
situations,
vacuum
might
not
be
able
to
catch
up
with
that.
We
don't
know
why.
That
was
the
case.
B
It
was
working
fine
before
that,
but
over
the
weekend
it
wasn't,
and
we
can.
We
can
actually
tell
when
that
happens,
so
we
will
be
able
to
maybe
just
pause
the
migration
when,
when
we
reached
that
threshold,
so
basically
to
avoid
the
the
noise
for
the
on-call,
there
is
an
issue
for
that
somewhere
in
the
in
the
world.
Already:
okay,.
A
Next
one,
I
close,
because
it's
not
a
consideration
to
worry
about,
I'm
just
going
to
pause
for
the
next
upgrade,
not
sure
what
this
one
was
actually.
A
B
I
was
wondering
like
there
was
a
suggestion
to
basically
expand
things
and
describe
a
caching
layer,
and
I
need
to
figure
out
how
important
that
is
for
us
right
now,
because
otherwise
I
think
it's
it's
ready.
A
B
Yeah-
and
I
think
we
have
a
lot
of
the
things
that
we
described
there
available
like
the
caching
mechanic
mechanics,
we
were
talking
about
making
it
easier,
making
it
more
convenient,
more
consistent,
which
will
make
sense
but
yeah.
Perhaps
that's
the
second
iteration
online
yep.
A
Okay,
I'll
have
a
comment
in
there
on
the
mr
later
on,
heinrich
the
background
migration
to
the
admin
panel.
F
Yeah,
I'm
working
on
the
first
iteration
to
be
for
review
soon.
I'm
not
sure,
I'm
not
sure
when
what
the
mvp
is
for
this
issue
like
do.
We
need
the
action
buttons
like
allowing
admins
to
retry.
I
guess
we
do
right
for
self-hosted
when
they
get
to
a
failed
state
or
something.
So
I
guess
that's
part
of
that
issue,
but
yeah
I'll
still
do
it
in
smaller,
mrs
though
we
could
still
deploy
the
first,
mr,
which
is
just
take
a
view
and
then
go
from
there.
E
Yep
same
same
thing:
there
I
was
just
gonna
start
looking
at
a
new
issue
today.
While
this
one
goes
to
review,
I
think
it's
hopefully
it's
good,
we'll
see
it
depends
on
if
ali
emerges
before
me,
because
if
he
does,
then
I
think
we
have
to
fight
or
something
I'm
not
quite
sure
how
that
works.
B
Yeah
we've
crescendo
described
how
we
can
or
what
the
customer
can
do
to
fix
that
problem,
I'm
still
to
review
that
which
is
taking
a
little
bit
longer.
Okay,.
A
And
there's
a
topic
later
on
the
agenda
crash
asked
like
at
what
point
do
we
just
send
it
over
to
like
ben
our
support,
stable
counterpart,
say:
hey
you've,
seen
something
like
this
before?
Can
you
triage
this
before
it
comes
straight
to
the
database
group?
Does
it
seem
like
one
of
these
issues
in
here
bypassed
support?
B
Right,
the
dashboard
for
the
migrations
that
was
waiting
on
a
another
metric
that
we
needed
to
add
and
we
need
to
do
a
little
bit
of
prometheus
foo
to
actually
properly
render
those
dashboards
and
the
metric
got
deployed
yesterday,
so
that
that
should
be
unblocked
in
a
sense.
So
next
step
will
be
prometheus,
foo
and
then
adding
those
networks.
Hopefully
that
works
out
and
andrew
is
helping
along
with
that.
So
that's
that's!
Pretty
good,
see
I'll
build
trace
sections.
B
That
is
an
interesting
one,
because
it
doesn't
have
a
single
column
primary
key
and
we
use
that
a
lot
for
the
batch
migrations.
So
there's
a
composite
primary
key
and
we're
basically
wondering
how
to
how
to
go
with
that.
We
can.
We
can
use
part
of
that
composite
primary
key
and
just
use
the
stem
batching
mechanics
as
we
have
now.
But
then
you
don't
have
a
lot.
B
You
know
you
kind
of
look
like
like
varying
batch
sizes
or
amount
of
data
that
you're
actually
touching,
because
we're
not
batching
by
the
primary
key
itself,
and
that
is
a
question
and
janis,
and
I
wanted
to
discuss
that
when
he's
back,
it
is
sort
of
prepared,
but
we're
not
really
sure
if
we
want
to
go
this
way
yet
so
that's
an
ongoing
discussion,
there's
also
not
no
rush
with
that
or
with
even
with
other
tables.
I
think
the
ci
builds
migration.
B
That's
currently
running
supposed
to
take
more
than
two
weeks,
and
so,
if
we,
if
we,
you
know,
if
we
add
more
tables
right
now
to
the
queue
it's
just
going
to
be
queued.
So
maybe
there
is
other
stuff
that
we
focus
on
right
now.
B
And
the
last
one,
there
is
a
basically
there's
a
problem
in
production
where
we're
being
paged
for
quite
a
lot,
which
is
query
that
falls
over
every
now
and
then,
and
that
is
also
slow.
When
it
doesn't,
it
hasn't
fallen
over,
so
it's
either
slow
or
super
slow,
and
there
is
perhaps
a
chance
to
add
a
very
targeted
index
to
speed
that
up,
although
like
looking
at
the
amount
of
indexes
on
that
table
on
ci
builds,
that
is,
that
is
totally
only
a
band-aid.
A
Right
thanks
so
ready
for
development.
Is
this
in
the
right
order?
A
B
A
A
Okay
sounds
like
we're
in
good
shape
on
this
board.
Close
it
go
back
to
the
agenda
triage
board.
I
know
giannis
was
on
perpetual
triage
for
a
while
there,
so
we'll
get
back
into
our
normal
rotation.
The
way
we
did
it
before
was
we
just
started
with,
I
think
andreas,
and
then
we
went
alphabetical
through
first
name,
so
you
welcome
alex
and
ali.
You
guys
get
skipped
by
us,
starting
with
andreas
and
we'll
go
through
the
rotation
after
that.
So
you
guys
are
a
couple
weeks
out
for
triage.
A
B
I
was
starting
to
look
through
that
yesterday
there
was
a
question
I
wanted
to
ask
you
about
it.
The
database
usage
by
name
space
at
the
bottom
is
that
something
that
you're
that
we
still
need.
You
know
there
was
a
question
like
what
are
the
top
10
namespaces
by
database
storage
size?
A
Nope,
it
has
not
so
it's
it's
fine
where
it
is
okay,
we
don't
need
a
prioritizer
right
now.
A
Now
I'll
revisit
it
later
see,
if
the
folks
that
asked
for
it
still
want.
A
A
Yeah-
and
there
was
I
referred
to
this
earlier
so
questions
on:
when
can
we
defer?
I
think,
I'm
not
sure
how
the
database
label
got
added
to
this
one.
I
nearly
added
it.
Okay,.
A
B
And
second
question:
after
that
perhaps
there
is
a
beckon
team
that
owns
owns
the
table
or
the
migration
or
knows
more
about
that.
B
Actually,
I
thought
it
would
be
nice
if
we
had
migrations
annotated
with
their
owners
like
we
do
in
other
cases
where
you
basically
can
tell
which
group
to
reach
out
to
that
might
even
be
helpful
later.
When
we,
you
know,
we
talked
about
documenting
the
schema
better
than
you
already
know,
which
team
introduced,
what
what
parts,
and
perhaps
that
helps
us
even
generate
things
more
automatically
later.
A
So
it
seems
like
there's
a
common
pattern
of
these
support
questions
where
a
customer
like
skipped
migrations
right.
They
just
went
from
one
build
like
three
beyond.
Do
we
not
have
anything
in
our
migrations,
it
says:
hey.
You
can't
run
this
until
the
prior
one
has
run.
A
Problem
for
future
us
right
do
we
have
an
issue
in
there
somewhere
to
investigate
that
it's
I'm
gonna,
overly
simplify.
This
seems
like
it'd,
be
pretty
simple,
say
I
can't
run
this
until
the
previous
one
has
run,
and
each
migration
would
only
have
to
know
the
previous
version
to
solve
that
problem.
Right.
You
gotta
start
somewhere,
okay,
I'll
put
a
very
simple
issue
together
to
talk
about
that.
F
F
They
run
in
the
background
right,
and
this
is
what
we
kind
of
discussed
during
the
in
the.
What
do
you
call
this?
The
release,
the
primary
key
migrations
to
self-hosted
thing,
where
we
also
you
know,
discussed
where
we
need
to
find
a
way
to
like
enforce
that
all
the
background
migrations
are
done
before
we
do
the
swap
so
it's
kind
of
like
the
same
thing,
because
right
now
we
only
have
docs
for
it
right.
We
tell
customers,
we
can
only
upgrade
after
migrations,
but
we
don't
do
anything
to
enforce
it.
A
Okay,
oh
yeah
I'll
put
a
very
simple
issue
together
or
we
can
talk
about
ideas
on
that
one
because
it
seems
like
that
would
solve
a
lot
of
problems.
I
don't
know
how
simple
it
would
be
to
implement,
but
okay
next
topic,
so
the
postgres
ai
question,
I
was
actually
asked
once
we
upgraded
our
hardware,
we're
going
to
revert
sounds
like
we.
Andreas
is
saying:
we've
already
reverted
back
and
we're
using
the
same
hardware
as
production.
At
this
point.
B
Yeah
we
had
troubles
with
that.
Then
we
we
went
back
to
what
we
had
before
my
understanding
is
that
the
upgrade
is
not
going
to
change
any
of
that,
like
we
have
the
same
instance
type
after
the
upgrades
are
very
good.
I
think
okay.
A
B
Oh
yeah,
there
was
a
disc
running
out
of
disk
space
sometime
this
week
last
week
and
that
comes
from
like
keeping
clones
you
can,
you
can
like
mark
them
as
permanent,
so
they
don't
get
cleaned
up
automatically
and
there
is.
You
can
check
the
disk
space
when
you
look
at
the
pusco's
ai
interface,
but
there
is
no
monitoring
for
that
matter
and
also
no
no
alerting.
That's
part
of
the
sort
of
the
make
make
things
more
mature
operationally
like
they
were
just
running
an
instance
that
nobody's
monitoring
that
automatically.
A
Okay,
on
a
semi-related
note,
we're
in
negotiations
for
renewing
contract
with
them
right
now,
so
current
contract
ends
at
the
end
of
the
month,
and
we
have
budget
approval
to
continue
for
at
least
six
more
months,
so
we're
we're
in
good
shape.
Right
now,.