►
From YouTube: 2021 04 20 Database 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
There
we
go
so
this
is
the
database
team
weekly
20th
of
april.
Today,
craig
is
on
pto
this
week,
so
we
will
dive
right
into
those
topics,
starting
with
the
intro
dev
issues.
I
don't
think
there
is
anything
new
as
far
as
I
could
see.
A
A
All
right,
then
we
we
can
dive
into
those
next
points.
So
basically,
I
think
craig
has
assigned
a
lot
of
the
issues
to
everybody
like
the
open
work
that
we
want
to
do
on
the
database
testing
side
and
also
primary
key
migrations.
A
If
you
see
those
assignments,
I
think
it
was
meant
to
be
to
indicate
that
this
is
the
work
that
we
have
ahead
of
us.
So
if
you
don't
feel
like
picking
that
up,
I
think
it
would
also
be
good
to
understand
you
also
just
that
it's
clear
who's
working
on
what
and
we
do
need
to
define
the
work
better,
especially
in
database
testing.
A
I
think
we
are
very
much
in
an
mvc
kind
of
state,
so
I
think
the
immediate
next
step
is
that
we
need
to
do
is
really
define
what
we
are.
What
we
want
to
work
on
and
a
lot
of
these
issues
don't
exist
yet
so
that's
that's
something
we
need
to
do
for
sure.
B
Alex
and
ali
any
questions
about
this-
maybe
I
don't
know-
and
just
have
you
shared
a
little
bit
of
background
of
why
we
are
doing
this
with
ali
and
alex.
A
Yeah
we
were
chatting
already,
but
we
have
another
meeting
schedule
for
this
week,
the
three
of
us,
so
we
can
dive
into
the
database
testing
topic
more
and
starting
from
there.
I
think
we
should
talk
just
talk
about
the
next
steps
that
we
want
to
do
and
get
those
issues
defined
and
then
start
working
on
that.
A
Cool,
okay
and
the
the
other
big
topic
that
we
have
is
the
primary
key
migrations
and
I
would
like
to
make
a
checkpoint
today,
sort
of
seeing
where
we
are
at
and
figuring
out
what
we
are
focusing
on
going
forward
and
I'm
not
entirely
sure
when
heinrich
is
going
to
be
able
to
join
us
or
not.
Krasner
is
already
picking
up
work,
so
he's
working
on
the
multi-column
support,
and
I
think
that
is
even
nearly
done
as
far
as
I
could
see.
A
The
the
the
last
mr
is
in
review
right
now,
and
this
is
also
what
we,
what
we
really
need
going
forward
is
the
multi-column
support,
because
all
the
tables,
the
large
tables
that
are
that
require
this.
So
once
we
have
that,
we
have
much
a
lot,
a
lot
of
tables
that
we
can
tackle.
Basically,
and
what
I
would
propose
to
do
is
right
now
we're
basically
sitting
sitting
idle
with
migrations.com.
So
that
means
we
have
finished.
A
I
think
the
three
migrations
that
we
were
running
previously
and
we
don't
have
something
scheduled
yet
so
right
now
we're
not
doing
anything,
I
would
recommend
we
we
go.
We
try
not
to
have
this
because
you
know
we
should
use
the
time
wisely
that
we
still
have
right
now.
This
is
a
single
threaded
process,
keeping
things
simple
and
if
we
use
the
time
wisely,
we
might
not
make
need
to
make
it
more
complicated
later
on.
A
So
what
I'm
trying
to
do
is
getting
getting
two
tables
out
and
scheduled
for
for
that
primary
key
backfill
which
isn't
working
out
yet.
So
I
appreciate,
if
anybody
has
time
to
to
look
into
that,
but
this
will
be
those
two
tables
see
I
build
needs
and
see.
I
built
trade
sections,
and
maybe
we
can
discuss
some
later
in
the
meeting.
B
Yeah
a
circle
back,
so
I
just
mentioned
that
you
are
not
sure
when
appendage
will
be
available.
Basically
everyone,
all
the
four
engineers
should
be
available
this
week,
so
I
should
be
working
on
this
working
group
this
week.
If
you
have,
you
got
a
chance
to
talk
to
him
or
I
think
craig
talked
to
him,
I'm
not
quite
sure,
but
if
you
haven't
talked
to
him,
I
will
try
to
reach
out
directly
to
henrich
to
see
when
he
can
pick
up
the
work
I
mean.
B
Basically,
he
should
be
available
this
week.
Okay,.
A
Good,
I
didn't
know,
I
haven't,
had
a
chance
to
talk
to
him.
I
will
I'll
also
reach
out
to
him
and
we
can
have
a
chat.
B
Yeah,
crass
and
henrich
they're.
In
the
same
time,
zongish
I
mean
they're
both
in
apex,
so
they
can
pair
up
very
nicely
to
work
on
this
topic.
So
please
reach
out
to
planet.
A
B
A
Cool,
so
this
is
the
the
multi-column
support.
This
is
basically
the
the
next
step
and
then,
at
the
same
time,
there
is
quite
a
few
other
tables
that
we
need
to
address
that
can
be
picked
up
individually.
This
is
like
the
smaller
tables
that
we
need
to
tackle.
A
So
if
anybody
feels
like
spending
time
on
that-
and
you
want
to
pick
one
of
these
tables,
I
put
a
link
to
the
the
list
of
tables
that
we
plan
to
tackle.
A
Janice
maintains
that
also
with
the
current
state
of
things,
and
if
you
want
to
pick
one
of
these
put
your
name
on
on
that
list,
and
then
there
is
an
example
or
actually
two
two
migration
examples.
What
we're
looking
for
is
basically
adding
adding
a
schema
change,
that's
column
as
triggers,
and
the
second
one
is
scheduling
the
backfill
for
those
tables
and
I'll
put
a
few
of
those
links
in
below.
C
C
C
We
are
going
to
wait
for
that,
because
we
want
to
also
build
all
the
other
tooling
and
release
those
at
the
same
time
on
self-hosted
nd
club.com,
because
the
rationale
there
is,
we
don't
want
the
schema
for
self-hosted
instances
to
deviate
from
the
canonical
schema
from
the
schema
we
have
instruction.sql
or
in
gitlab.com,
because
that
will
cause
various
issues
for
us
and
it's
not
only
that
they
will
deviate.
They
will
deviate
also
between
all
the
self-hosted
instances.
C
Gitlab.Com
and
new
surf
coastal
distances,
because
all
new
surf
coaster
instances
use
the
structure.sql
to
to
build
their
database,
and
then
we
have
a
different
schema
from
other
self-hosted
instances.
We
don't
want
to
do
that,
for
so
we
have
a
couple
of
issues
I
have
created.
I
have
not
created
for
all
the
satellite
ones.
I
think
that
the
canonical
implementation
now
is
the
one
that
address
has
added
for
ci
build
needs.
A
I
just
closed
dmr
in
favor
of
what
crescendo
was
working
or
what
patrick
has
had
prepared.
I
didn't
see
that
before
so.
C
We
can
use
the
same
helper
now.
C
Yeah
so
yeah
the
then
the
mr
for
events
should
be
okay.
If
it,
it
has
only
one
column,
it
should
be
okay,.
B
Yeah
so
question
so,
are
we
asking
I
mean,
are
we
asking
also
alexa
and
ali
if
they
are
available
to
help
migrating
this
table
or
is
pretty
much?
We
are
asking
chris
and
henrich
to
to
help
this
migration.
A
A
I
think
there
is
a
lot
of
work
to
do
that
can
be
picked
up,
especially
those
those
smaller
tables.
We
know
how
to
do
it.
We
have
all
the
helpers
in
place,
so
it's
a
matter
of
picking
the
table
up
and
shipping
that
migration.
Basically,
if
you
have
time
to
do
that,
alex
and
ollie
totally
appreciate
you
picking
up
that
if
you
know,
but
perhaps
I
think
it's
good,
that
we
sort
of
have
a
focus
on
database
testing
and
the
other
group
focuses
on
the
primary
key
migrations
in
general.
A
But
if
you
feel
like
spending
time
on
that,
it
would
be
greatly
appreciated.
I
think.
B
Okay,
so
let
me
just
refresh
so
ali
and
alex
your
primary
task
will
be
helping
the
database
automation,
testing
task,
but
if
you
get
free
cycles,
welcome
to
help
migration
migrating
the
smaller
tables,
because
we
have
a
mature
process
where
we
fix
the
recipe
to
do
that.
B
So
welcome
to
help
that
if
you
are
available,
you
have
other
free
cycles
to
to
pick
up
some
of
the
smaller
tables
and
the
class
and
heritage
will
be
mainly
working
on
the
migration,
including
the
tooling
and
the
migration,
the
large
tables,
but
let's
let's
they
can
help,
of
course,
on
an
automation
patch.
I
guess
they
will
probably
not
be
able
to
but
anyways
yeah.
C
C
Whatever
is
added,
there
means
that
we
can
pick
it
up
and
I
will
add
an
example
there
of
you
know,
that's
the
mr
that
we
can
check
either
way
we
will
address
or
we
can
review
so.
The
first
steps
will
be.
You
know
we
can
do
an
onboarding
there,
so
I
will
add
the
issues
for
all
the
tips
that
are
ready
to
be
converted.
B
Thank
you,
okay,
alia
and
alex
are
so
basically
where
you
feel
comfortable
when
you
have
the
free
cycle
help
pick
up,
but
you
know
this
is
a
volunteering
based.
A
It
works
and
what
I
think
so.
This
is
a
this.
This
project
has
many
aspects
like
there's:
there
is
gitlabcom,
which
is
basically
at
risk
of
overflowing
at
some
point
with
primary
keys
and
there
is
self-hosted,
and
so
I
think
at
the
moment
we
need
to
pick
on
what
we
are
focusing
on
on
this
project.
What
I
would
propose
to
do
at
this
time
is
focus
on
keeping
the
migrations
alive,
so
keep
make
sure
that
there's
always
a
migration
running
on
google.com.
A
Basically,
because
it's
gonna
be
taking
a
lot
of
time
and
the
earlier
we
can
get
it
get
it
done.
I
think
the
more
confident
we
are
also-
and
we
don't
have
to
complicate
things
in
the
end,
worrying
about
concurrency
and
all
that
so.
A
That's
the
I
or
the
implementation
right
now
only
supports
one
process
at
a
time.
Okay,
we
have
ideas
of
how
to
do
that
concurrently,
but
it's
it's.
If
we
don't
need
to
do
that,
I
think
it's
much
better
so
using
the
time
seeing
of
how
we
can
get
that's
the
idea.
A
No,
that
is,
that
is
unrelated.
You
can
totally
merge
it.
It's
gonna
be
enqueued,
but
it's
only
going
to
be
picked
up
when
when
it's
time
basically
so
we
can,
we
can
always
merge
those,
mrs,
it's
also
good
to
merge
them
early,
because
this
is
a
two-step
process.
A
One
is
basically
adding
a
new
column
and
and
the
trigger
that
basically
keeps
the
old
and
the
new
column
in
sync,
and
once
you
have
shipped
that,
basically
all
the
new
data
that
comes
into
the
table
doesn't
have
to
be
migrated
later,
because
it's
already
in
the
shape
that
you
wanted
to
be
in,
and
the
second
phase
is
the
backfill
that
will
that
can
also,
like,
I
said,
can
be
merged.
It
will
be
enqueued
picked
up
later,
perhaps,
but
it's
it's
always
good
to
merge,
merge
those
things
early.
C
And
because
we
depend
from
if
something
happens
and
deployments
are
blocked,
we
cannot
deploy
a
migration
so
because
we
need
the
migration
in
order
to
to
enqueue
that
backfilling
and
we
don't
care
because
we
serialize
everything
after
we
run
it,
it's
better.
Let's
add
as
many
as
we
feel
comfortable.
D
C
A
C
C
To
queuing,
that's
what
happens
right
now,
also,
so
our
background
migrations.
Right
now
everybody
emerges
background
migration.
It's
not
like.
We
are
doing
it
all
the
time,
but
we
can
let
the
10
different
developers
merge
10
mars
with
background
migrations.
They
are
serialized
in
in
the
sidekick
queue
with
this.
D
B
I
think
both
angels
and
iannis
made
it
clear
that
the
we
always
want
some
migration
work
is
cranking.
We
always
want.
The
system
is
cranking
on
the
migrations
and
there
is
a
recipe
to
do
that.
The
honest
will
be
the
contact
point.
If
you
have
question
and
and
of
course
this
is
important
because
we
are
we're
racing
against
our
head
room-
the
primary
key
moving
it
from
inch
to
begin.
B
So
it's
also
tracking
our
daily
dot-com
daily
stability
stand
up.
B
A
C
Okay,
that's
great,
because
that's
important
now
that
we're
going
to
work
on
all
those
aspects
on
the
database,
for
example,
it
will
be
a
nice
thing.
You
can
use
the
replica
to
check
the
progress
of
the
migrations
and
various
other
stuff
and
I
will
add
some
links
on
how
to
monitor.
B
Yeah,
if
you
discover
you,
don't
have
a
permission
to
access
anywhere
at
an
access
request.
Issue
ping
me:
I
will
get
it
going
as
soon
as
possible.
A
Okay
and
then,
let's
make
a
checkpoint,
perhaps
next
week
or
in
the
beginning
of
may,
and
real
and
check
what
we
what
we
still
need
to
get
into
the
release,
so
that
things
are
also
happening
on
self-hosted,
but
right
now
I
would
rather
focus
on
filling
up
a
pipeline
for
the
migrations
as
good
as
we
can
and
then
go
to
self-hosted
area.
Does
that
make
sense,
yeah,
okay,
anything
else
on
primary
key
migrations.
C
For
the
workbook
logs
webco
clocks,
we
are
addressing
it
in
a
different
way,
so
we
are
not
only
converting
the
primary
key.
We
are
partitioning
the
table
because
they
it
had
other
issues
as
well
and
at
the
same
time,
we're
converting
the
primary
key
and
all
the
migrations
of
step.
Three
step.
Three
was
adding
indexes
and
foreign
keys
and
made
it
for
13.11,
that's
great
for
us.
That
means
that
in
this
milestone,
we're
going
to
swap
the
tables
now.
C
C
C
Amazing
that
now
the
we
have
the
tooling
at
the
level
where
we
are
trusting
our
tooling
to
just
just
go
and
add
the
migration
map
migration,
not
migration.
A
I
think,
to
some
extent,
we
still
need
to
do
a
little
bit
of
writing
on
the
documentation
like
explaining
explaining
how
to
approach
it.
Talking
about
access
patterns
very
similar
to
what
we
discussed
on
on
the
shouting
call
or
the
looking
group.
C
But
the
good
thing
is
now
with
after
the
audit
events
and
now
with
the
the
fiber
march
for
the
webhook
loads.
We
have
the
canonical
example.
C
A
D
C
We
asked
we
asked
the
group
ecosystem
to
turn
off
their
bloom,
their
root
voter.
So
for
for
everyone,
it's
one
terabytes
of
data.
We
are
adding
160
gigabytes
of
logs
every
month.
It's
it's!
It
goes
so
fast.
We
had
the
prone
worker
that
was
going
deleting
all
the
loans.
So
we
have
irritation
policy
to
keep
only
90
days
of
our
loss
and
our
worker
could
not
keep
up
with
the
rate
that
the
blocks
were
created.
C
So
we
were
creating
something
like
a
2.5
3
million
records
per
a
day
and
the
worker
will
delete
only
one
million,
so
web
looks
kept
on
increasing
and
increasing
increasing
in
size,
and
nowadays
because
it
has
reached
the
point
of
one
terabyte,
it
cannot
not
only.
It
cannot
keep
up
it's
also
times
out.
So
what
we
are
going
to
do.
C
We
are
adding
the
monthly
partitions
and
when
we
are
done
next
milestone
after
we
do
the
swap
and
we
drop
the
old
table,
a
new
worker
will
be
added
that
will
go
back
and
every
month
we
drop
a
non-partition
and
only
keep
90
days,
90
days
in
the
current
month.
C
So
yeah
we
have
talked.
We
discussed
that
with
a
group
ecosystem,
that's
the
point
where
we
will
hand
off
the
table
to
them
back
to
them
a
clean
signing
new
partition
table
and
then
we
implement
the
or
I
don't
know
andreas.
Maybe
we
could
yeah.
We
will
have
to
think
about
the
implementation,
but
maybe
they
they
can
help
us
now
that
we
are
pretty
much
our
hands
are
tired.
A
Yeah,
that
would
be
nice,
but
it's
it's
kind
of
an
impactful
area.
Isn't
it
like
takes
10
of
the
total
database
sizes
webhook
logs,
which
is
like.
A
C
A
Okay,
we
need
to
finish
it
up
retention
strategy
and
then
we
might
even
need
to
think
about
cutting
the
partition
smaller
in
the
future.
A
A
Right:
okay,
we
talked
about
database
testing
in
the
beginning.
Should
we
jump
to
the
blueprints
john
yeah.
B
Just
a
quick
question
touch
base
where
we
are,
it
looks
like
I
guess
you
already
replied
on
hold
given
other
priorities.
Yes,
I
agree,
but
we
cannot
wait
too
long,
so
hopefully
the
we
can
run
the
alex
crash
and
the
heritage
as
soon
as
possible.
Let's
shoot
for
like
a
one
week
in
about
a
week.
They
can
get
up
to
speed
for
what
we
have
them
to
help.
Then,
let's
resume
our
focus
on
the
on
the
blueprints,
because
we
still
need
to
make
progress
there.
B
Although
the
main
discussion
is
around
charlie
but
the
other
side,
we
we
must
make
progress
as
equally
important
because
yeah
the
sharding
group
already
discussed
this
morning.
What
else
we
are
doing.
A
Yeah
yeah,
I
totally
agree
the
for
for
the
mostly
blueprint
I
think
it's
in
a
good
state
where
it
basically
summarizes
what
the
problem
is,
how
to
solve
that.
A
But
the
question
is:
do
we
want
to
go
further
and
sort
of
propose
a
new
layer
propose
how
to
make
that
even
easier,
or
is
that
you
know
what
what
is
the?
What
is
the
scope
of
the
blueprint
that
we're
that
we're
aiming
for?
That's
something
that
I
need
to
find
out,
but
I
right
now
I
just
don't
have
time
to
pick
it
up,
but
we'll
do
so
as
soon
as
possible.
Yeah.
B
Thank
you.
Let's
yeah
yeah.
C
And
for
time,
decay
I
have
to
spend
one
or
two
more
days,
so
I
have
posted
right
now.
I
have
some
great
feedback
by
andreas
and
gregors
and
I
want
to
add
the
two
big
examples
that
we
so
web
cloaks
and
the
audit
events.
This
is
missing
there.
So
there
is
the
theoretical
discussion
and
all
the
nice
comments
by
addressing
gregors
that
I
will
incorporate
them
and
also,
I
want
to
add
those
large
sections,
because
those
will
be
the
actionable
part
of
the
blueprint
without
those.
This
is
just
a
discussion.
C
B
Thank
you.
Thank
you
to
both
of
you.
So
we
are.
I
don't
think
we
can
merge
it
by
the
end
of
the
quarter,
but
hopefully
we
should
for
the
first
several
weeks
of
the
q3
q2,
to
merge
those
blueprints.
B
Then
we
can
start
think
about
building
the
backlogs
to
work
on
that,
I
think
so
also
for
ali
and
alex.
So
there
is
isn't
a
shortage
of
work,
I'm
asking
too
much.
I
know
I'm
asking
too
much,
but
I
imagine
a
majority
of
this.
The
database
team
will
be
also
working
on
the
these
two
blueprints.
We
will
pick
likely
we
are
going
to
pick
one
of
those
blueprints
to
work
on
to
start
with,
but
still
even
one
blueprint.
There
will
be
a
lot
of
work
here.
B
This
is
also
under
that
working
group.
So
a
lot
of
interesting
stuff
to
happen.
B
Cool,
thank
you
so
just
touch
base
where
we
are
about
the
blueprints
and
the
next
one
is
actually
just
a
task.
Pg12
upgrade
right
now
scheduled
at
may,
8th,
roughly
the
same
time
as
cigarette
starts
on
9
am
utc
and
goes
into
10
a.m.
Utc,
that's
the
actual
starting
and
ending
time.
The
window
probably
is
going
to
be
three
hours
one
hour
earlier
one
hour
post,
so
that
window
might
be
a
three
hour
window,
but
the
actual
upgrade
happens
in
one
hour,
so
the
infrastructure
team
is
asking
for
two
bees.
C
Yeah,
I
can
do
it.
I
have
no
problem.
B
Yeah
and
also
they,
they
also
need
the
older
staff
plus
bes
available
or
by
so
I'm
going
to
coordinate
that
and
send
that
send
out.
Another
notice
to
those
people
give
them
a
heads
up
to
be
prepared.
So
yeah,
that's
just
a
heads
up.
What's
coming
up.
A
Yeah
sounds
good.
Is
there
the
last
time
we
talked
about
like
support?
I
wasn't
quite
sure
about
the
the
expectations
or
if
there
was
any
like
expectations
to
be
on
call
for
an
extended
period
of
time,
but
it
sounds
like
we
were
basically
looking
to
have
people
around
during
the
upgrade
once
we're
confident
that
things
are
running
on
saturday,
then
yeah.
B
I
see
yeah,
I
think
during
that
window,
the
three
hour
window
we'll
know
more
in
that
issue,
but
if
that's
a
schedule
like
a
three
hour
window
like
a
one
hour
before
the
actually
kick
off
the
job
and
one
hour
later,
just
to
watch,
we
want
to
be
shadowing
there.
B
We
want
both
of
you
to
shadow
there
to
in
case
something
happens,
and
if
we
two
of
you,
is
very
specific
about
the
database,
I
think
it's
going
to
be
very
specific
about
database.
The
reason
jerry
is
asking
for
other
bus
is
like
something
else
may
happen
like
a
sidekick
or
something,
so
they
really
want
some
other
bs
to
be
available
to
troubleshoot
other
issues,
not
database,
directly
related,
but
more
about
other
components.
Sidekiq
and
jerry
just
mentioned
the
sidekick.
But
if
we
see
other
things,
we
want
bees
to
help
troubleshooting.
C
Oh
yeah,
that's
a
very
small
dog
update.
I
have
even
I
think
that
it's
what
I
have
in
the
description.
Maybe
I
will
take
the
time
to
to
do
it.
But
if
anyone
wants
to
pick
this,
I
will
greatly
appreciate
it.
It's
just
for
helping
surfboard
instances
address
self-serve
and
address
some
minor
issues
when
going
through
minor
versions
of
13
gitlab
13..
So
the
sooner.
C
Because
this
is
the
documentation
update,
it's
just,
it
will
remove
any
if
they
have
a
similar
problems.
We
have
seen
some
of
those
problems.
If
they
have
similar
problems,
they
will
find
the
solution
in
the
documentation
and
not
on
our
issues.
But
okay.
We
also
have
solutions
in
our
issues
if
they
search
google.
C
A
C
C
C
D
A
E
D
A
Typically,
walk
the
billboards.
Should
we
should
we
do
that
today
we
can
go
to
that
quickly.
B
Yeah
do
the
regular
process.
A
You
are
the
driver,
yeah.
Okay,
let
me
share
my.
A
A
There
we
go
so
if
you
haven't
seen
those
before
basically
have
two
boards
one
is
the
build
board,
basically
captures
all
the
the
work
that
is
currently
in
flight.
You
can
say,
and
the
other
one
is
the
the
triage
board,
which
is
all
the
issues
that
is
in
an
early
state
where
we
are
basically
looking
to
explore
like
what
is.
Is
it
even
the
initiative
you
want
to
work
on?
Does
it
have
all
the
details
and
once
it's
done,
we
basically
and
we
decide
to
pick
it.
A
Then
we
basically
move
it
to
the
billboard.
So
if
you
want
to
see
what
is
currently
in
flight,
basically
look
at
the
billboard.
A
This
is
what
we
basically
saw.
We
stopped
from
the
back,
but
I
think
we
talked
up
a
lot
about
this
already
like
this
is
the
weblogs
partitioning
that
was
that
was
added
with
the
foreign
keys.
Is
it
something
you
want
to
add?
Giannis,
no.
C
A
Okay
sounds
good
thanks.
The
only
thing
I
want
to
touch
on
is
this
one,
so
I
don't
know
if
everybody
remembers
that
we
had
a.
We
have
a
database
lab
instance,
and
we
we
were
running
that
with
a
very
small
instance
type
for
some
time
and
then
realized
that
we
need
to
upgrade
it.
Then
we
upgraded
it
to
what
we
expected
to
have
in
production
after
the
pg-12
upgrade.
A
But
then
during
the
testing
for
the
pg-12
upgrade,
we
realized
that
the
performance
wasn't
so
great
for
those,
for
that
instance
type
so,
and
that
was
noticeable
on
the
on
the
database
lab
instance
as
well.
It
was
basically
burning,
so
we
were
restarting
the
database
elements
instance
using
a
different
type,
so
it
should
be
in
better
shape.
A
Now
there
is
a
couple
of
things
that
are
that
I
think
we
need
to
address
as
a
team
is
that
we
typically
reach
out
to
nikolai
for
this
kind
of
maintenance
or
helping
us
changing
things,
fixing
fixing
stuff
on
the
database
lab
instance,
and
I
would
I
would
like
us
to
have
more
insight
into
into
the
operational
side
of
that
as
well,
because
it
powers
the
database
testing
pipeline.
It
powers,
database
lab
and
everybody
like
exploring
database
issues,
and
we
should
make
that
a
part
of
our
work.
A
I
think
for
the
database.
Testing
is
also
being
able
to
operate
this
instance
and
control
the
environment
better.
A
Cool
we
have
another
work
stream
for
the
the
background
migrate
or
the
primary
key
migrations
in
flight,
which
is
the
metrics
for
the
migrations.
A
So
right
now
we're
basically
going
into
the
database
to
figure
out
the
progress
of
a
migration
and
the
the
various
tuning
opportunities
that
we
that
we
want
to
have
by
by
looking
into
the
database
and
and
logs,
and
all
that-
and
there
is
two
changes
in
flight
that
basically
add
prometheus
metrics
to
that,
so
that
we
we
can
easily
reason
about
the
progress
of
a
migration
and
also
the
the
you
know
the
efficiency
like.
Do
we
use
time
wisely?
A
Do
we
need
to
increase
the
batch
size,
for
example,
and
that
should
all
come
to
prometheus
soon?
We
will
only
see
that
happen
if
we
actually
run
migrations
again
or
once
we
run
migrations.com.
But
that's
that's.
The
one
is
one
issue.
Is
this
one.
A
We
still
have
the
database
testing
to
database
maintainers,
I'm
basically
closing
that
soon
we
have
rolled
it
out
to
database
maintainers.
I
think
even
more
people
are
using
it
that
has
been
used
as
an
as
a
feedback
issue.
So
that's
why
I'm
not
yet
closing
it,
but
we
should
let's,
let's
just
talk
about
that
on
wednesday
tomorrow,
when
we
catch
up
on
a
testing
pipeline
talked
about
read
mostly
then.
A
This
is,
I
think,
what
what
krasner
is
working
on.
Basically,
the
support
for
the
multi-column
primary
key
migrations.
A
Exactly
and
it's
in
review
already-
and
this
is
again
prometheus
metrics-
also
in
review.
A
C
I
think
the
one
is
for
I
don't
know
yeah,
so
we
had
the
two
yeah.
Maybe
they
are.
A
Issues,
okay
time
decay,
we
talked
about
ci
build
needs.
This
is
one
of
the
next
tables
that
we
would
like
to
tackle.
I
got
feedback
from
carmel
about
the
change
that
we
needed
to
make
and
I
think
we
need
to
make
that
a
bit
different
before
we
can
merge
it.
So
that's
I
hope
to
get
that
done
tomorrow,
but.
C
Yeah
a
reminder
there
is,
whenever
we
add
the
new
columns,
we
should
be
remind
remember
to
check
if
those
tables
are
auto,
backed
up
or
whatever
and
those
columns
affecting
something.
So
it's
not
like
the
columns
are
not
hidden
and
and
sometimes
in
rails.
We
are
going
and
you
know
automatically
do
something
with
all
the
columns
of
the
table.
We
should
remember
that
sometimes
doing
we
should
hide
those
comments.
A
Especially
when
now
that
we
have
triggers
in
the
database
which
do
their
thing
on
the
back
end,
basically
and
and
the
application
doesn't
know
about
it,
so
that
that's
why
it's
surprising
in
this
case,
for
for
the
ci
application.
A
Okay
and
then
this
is
a
minor
issue.
I
think
it's
a
follow-up
from
from
a
review
that
we
had.
We
would
like
to
make
the
the
rollback
step
for
the
primary
key
migration
helpers
a
bit
more
concise,
basically,
so
that
you
don't
have
to
repeat
yourself
all
the
time
and
I
might
need
to
push
that
back
or
put
that
back
onto
the
backlog.
I
had
a
proposal,
but
it's
I
think
we
should
do
it
differently,
I'll
revisit
that
today,.
A
Okay
and
then
this
is
something
that
I
can't
speak
to
the
admin
panel.
I
think
this
falls
under
the
things
that
we
need
to
do
for
self-hosted
yeah,
so.
C
The
first
thing
we
need
to
do
one
of
the
things
we
need
to
do
in
order
to
roll
out
those
migrations
other
than
making
sure
that
our
framework
works
correctly
and
can
recover
from
issues
and
whatever
is
to
allow
self-hosted
instances
administration
traders
to
know
that
those
migrations
run.
So
they
have
to
know
so.
At
the
moment
there
is
an
admin
panel
where
they
can
check
all
the
background,
migrations
that
are
running.
Why
is
that
important?
C
So
we
ask
our
administrators
to
to
check
that
everything
has
finished,
and
that
means
that
some
migrations
may
run
for
a
week
or.
C
So
we
have
to
provide
them
with
a
panel
where
they
can
see
that.
Oh
you
know
I
cannot
upgrade
to
14.0
or
I
cannot
upgrade
to
14.2
before,
because
there
are
two
migrations
still
running
and.
C
D
C
C
Position
etc
so
yeah.
This
is
one
of
the
things
that
we.
C
Yeah,
but
we
added
this
is
a
simple,
except
this
specific
one
does
not
have
any
options
to
pause
or
start
or
whatever.
This
is
a
self-contained
issue
to
pick
up
at
some
point.
That
only
adds
a
view
on
the
bat's
background,
migrations
and
batch
background
migration,
jobs
tables
on
the
admin
panel.
So
it's
a
pretty
self-contained.
C
A
Finalizing
means
basically
yeah,
basically
swapping
the
columns
in
the
end,
and
then
only
when
that
is
done.
We
basically
benefit
from
or
we
have
basically
removed
the
risk
of
the
overflow,
because
only
then
we
are
using
integer
8
primary
keys
and
we
stop
using
integer
4.
C
D
C
A
A
Yeah
janice
can
we
can
we
just
do
that
after
the
meeting?
If
you
have
time.
C
C
Already
so
this
is
the
problem
and
the
fortune
of
being
also
an
operational
team.
We
have
some
back
reports
where
we
support
end
users
and
they
have
issues
and
we
try
and
they
stay
around
for
weeks
or
maybe
sometimes
months,
and
we
have
those
back
and
forth
back
and
forth
and
we
are
working
on
them
or
well.
C
So
they
stay
in
the
board
for
maybe
a
month
or
two
months
like,
for
example,
the
one
on
the
dev
that
class
has
could
not
find
the
function
name,
the
table
seek
function.
C
We
have
back
and
forth
for
more
than
a
month
on
that
one.
A
A
Something
that
I
wanted
to
ask
behind
have
created
that
issue
for
an
emergency
plan
for
the
integer
4
overflow
problem.
So
basically
we
were
thinking
about.
What
can
we
do
if
we
don't
manage
to
get
the
migrations
done
in
time?
I
think
we've
spelled
out
a
couple
of
ideas.
I
wanted
to
ask
if
we
still
feel
like
this
is
something
that
we
want
to
discuss
at
this
point
or
if
that
is
something
that
we
can
put
to
the
backlog
and
spend
the
time
on
the
on
the
actual
migrations.
A
I
was
a
bit
surprised
to
be
honest
by
by
the
speed
of
the
migrations
that
we
were
seeing
for
the
ones
that
we
already
finished.
That
was
a
very
good
sign.
I
think.
So.
What
do
you
think
about
the
emergency
plan
discussion.
B
So
sunday,
like
we
are
moving
faster
than
we
expected.
Do
we
have
this?
Do
we
have
a
refresh
of
the
projection,
the
the
the
trend
when
we
are
going
to
hit
the
wall?
Do
we
have
a
refresh?
Are
we
going?
Are
we
still
on
track?
I
mean
basically,
we
are
not
moving
towards
the
exponential
exponential
curve
right.
We
are
still
far
away
from
there.
A
We
have
refreshed
that
forecast,
I
think
two
weeks
ago,
and
we
are,
there
is
no
surprises.
In
that
sense,
the
only
takeaway
that
we
were
seeing
last
month
was
that
there
was
a.
I
think
it
was
20
to
30
percent
increase
in
the
number
of
jobs
that
we
were
running
in
march
compared
to
the
previous
month,
and
that
was
attributed
to
abuse
mostly.
So
we
think
that
that
we
are
still
like
on
on
track
with
with
what
we
have
forecasted
okay,
correct.
B
Then
I
think
we
can
put
it
on
code
a
little
bit,
but
maybe
not
move
back
to
the
backlog.
Let's
just
monitor
the
trend.
I
think
this
will
just
stay
here,
but
we're
not
going
to
take
immediate
actions
but
continue
to
monitor
the
trend
to
see
if,
if
the
trend
turns
out
to
be
not
so
favorable
that
we
need
to
turn
this
on,
so
putting
it
back
to
the
backlog,
probably
will
lose
visibility.
A
Yeah,
that
makes
sense.
Maybe
this
is
also
something
that
we
can
checkpoint
on
in
may
or
may,
when
we
have
the
the
larger
migrations
running
on
the
very
large
tables
that
weird
that
we
have,
we
should
have
a
good
impression
of
how
how
fast
we
can
expect
them
to
complete
and
then
go
back
to
the
emergency
if
we
need
the
emergency
plan
at
all,
yeah,
okay,
good.
A
Okay,
I
don't
think
we
need
to
talk
about
the
other
issues,
perhaps
aside
from
the
pg-12
upgrade,
but
this
is
something
that
I
would
do.
I
think
we
can.
We
can
wait
for
a
week
before
the
upgrade
and
then
see
what
the
state
of
things
are.
So
the
concern
was
basically
do.
We
need
to
pause
the
migrations
that
we're
running
while
we
did
the
pt
type
upgrade.
A
That
all
right
anything
else,
we
should
be
discussing.
B
No,
I
last
of
thing
is
I
in
the
dark,
so
I
made
I
massaged
the
organization
of
the
the
epics
added
one
top
level,
epic
for
the
for
all
our
epics,
because
we
are
all
of
our
our
epics
are
under
the
the
engineering
allocation.
So
we
need
to
track
that.
So
I
did
one
top
level
epic
and
wrote
all
everything
beneath
that.
So
we
have
a
dashboard
to
monitor
our
progress,
but
that's
what
I
did
last
night.
B
E
B
A
C
Do
we
keep
the
label
also
for
issues
that
andreas
or
me
are
working
on.
B
Yeah,
because
we
are
a
hundred
percent,
our
team
is
100.
The
engineer
allocation,
so
everything
we
work
on
issues
and
imrs
should
be
labeled
by
the
engineering
allocation
and
lily
is
going
to
build
a
dashboard
to
monitor
the
issues
on
mrs
under
engineering
allocation.
B
A
Okay,
great
and
thanks
everybody
have
a
great
day
and
janice,
and
I
we
probably
catch
up
on
the
on
the
issues
fleshing
out
that,
after
the
call,
that's
good.
Okay
have
a
nice
one.
Everybody.