►
From YouTube: 2020 05 18 Database Team Weekly Follow Up
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).
B
A
A
A
A
A
D
A
D
D
A
B
We've
done
this
for
a
while,
but
from
the
infrastructure
side
you
know
we
have
access
to
data
anyways
and
this
sort
of
tripped
over
into
development.
All
so,
basically
we're
asking
as
a
development
team
can
we
can
be
used
at
theta4,
database,
side,
testing
and
I.
Think
the
yes,
the
process
exists.
We
might
want
to
adjust
that
a
little
bit
for
our
team
and
the
the
access
method
is
pretty
much
the
same
as
you
as
we
already
have
for
production
data.
So
we
would
be.
B
B
D
C
C
A
Migrations
yeah.
We
talked
about
migration
partitioning
design,
so
we
got
the
two
topics
on
the
table
for
partitioning
audit
logs,
whether
it's
created
that
or
a
top-level
namespace-
and
maybe
that's
something
we
can
talk
about
next
since
josh
is
on
here
and
then
implementing
like
cascading
deletes
without
foreign
key,
because
that
did
that
close
out.
A
Okay,
so
it
is
the
start
of
thirteen
one
milestone.
Andre
is
rightfully
asks
what
what
are
we
committing
to?
What's
the
goal
for
the
milestone
and
I
actually
added
comment
in
there,
so
partitioning
there's
a
lot
of
eyes
on
it
right
now,
so
partitioning
is
far
and
away
our
top
priority
and
having
a
couple
of
you
pair
up
on
it.
So,
like
everybody
focus
where
they
can
on
this,
and
we
can
get
into
specific
here
in
a
sec
database,
growth
and
sizes
popped
up
again
Josh.
You
called
that
one
out
screen.
D
If
there
is
in
the
running
question
in
the
goldman
sachs,
don't
this
is
public
right,
government
sex
issue
tracker,
so
basically
there
I'm,
not
sure
India
actually
fails.
I
just
happen
to
us
clipping
from
the
Goldman
Sachs
select,
channel
and
I
saw
some
of
their
alerts,
and
then
so,
basically,
it
sounds
like
they're
deleting
the
pipelines
to
try
and
save
database
space,
not
sure
exactly
why
or
what
constraints
they
have
in
the
database
size,
but
they
weren't
a
fan
of
it.
D
You
know
this
seems
like
it
might
be
helpful
for
them,
because
I
can
simply
just
run
a
new
pipeline
if
you
need
to
right
as
opposed
to
hitting
retry.
If
you
need
to
recreate
something,
I
think
that
was
a
major
concern
with
just
deleting
stuff
is
that
they
might
not
be
able
to
recreate
certain
environments
anymore.
They
can
run
the
pipeline's.
D
A
D
I
think
this
is
I
think
this
is
underway.
I
just
want
to
highlight
that
that
in
general
database,
size,
at
least
from
one
customer
is,
is
they're
having
problems
with
it.
I
would
like
to
learn
more
about
that.
D
I
have
made
comments
about
some
of
the
data
retention
in
archival
benefits,
who
can
get
by
cleaning
up
data
from
the
database
and
I've
gotten
some
comments
that
it
doesn't
matter
once
we
have
sharding,
because
we're
just
have
a
constant
time
cost
to
encrypt
a
fan
out
across
the
partitions
and
therefore,
like
you
know
it
doesn't
matter
anymore.
We're
not
gonna
do
some
table
scans,
but
so
I
think
understanding
more
the
challenges
our
customers
are
having
and
how
building
a
case
around
why
baby
size
is
important
might
be
valuable.
D
So
we
can
better
address
some
some
some.
We
know
what
can
what
will
purchasing
solve
and
what
what
partitioning
won't
solve
and
what
we'll
have
to
make
sure
we
tackle
other
projects
to
turn
sure
we
have
a
performance
and
scalable
database.
That's
all
I
think
so.
Take-Home
series
I'll
try
and
get
more
detail
with
the
customer
issues,
so
they
have
understanding
of
what
they're
experiencing
and
why
and
sort
of
what
maybe
I'm
idea
of
how
common
this
is
across
customers
and
I.
D
A
So
yeah,
since
we
talked
about
a
customer
I'll,
make
sure
this
is
marked
as
private
makes
me
kind
of
cringe
and
I
saw
at
least
one
or
two
faces
kind
of
react
to
your
comment
about
it.
Doesn't
matter
shardene
I'll,
take
care
of
this
problem,
we're
still
a
ways
away
from
production
level,
sharding,
rollout
and
I
think
there
are
probably
still
some
things
that
we
should
fix
in
our
database
growth
and
current
sides,
and
we
have
fortunately
have
epics
for
both
of
those
and
to
ondrea's
question.
What
should
we
prioritize
for
thirteen
one?
A
We
have
a
couple
individual
issues
in
there
and
we
know
we
need
to
focus
on
partitioning.
We
don't
know
specifically
what
would
have
the
most
impact
on
the
database,
growth
and
overall
size.
So
maybe
we
can
spend
the
next
several
minutes
talking
about
that
into
the
partitioning
side.
So,
right
now
we're
approaching
audit
events
on
the
created
by
so
partitioning
on
a
date
range.
But
the
question
that
CID
brought
up
last
week
was:
why
aren't
we
focusing
on
namespace
and
ondrea's?
A
B
Mm-Hmm,
so
maybe
just
as
a
quick
recap
of
why
we
chose
time
as
a
partitioning
key
for
audit
log
with
the
existing
features
that
is
sort
of
ideal,
because
if
we
can
assume
that
we
always
have
a
date
feature
than
the
date
based
partitioning
that's
kind
of
a
deal,
it
goes
with
everything
it
works
for
the
admin
view
instance
wide,
and
it
also
works
for
the
project,
views
and
the
API.
It's
all
good.
As
long
as
we
always
have
this
beta
filter
in
place
now,
I
think.
B
The
argument
that
was
made
was
assuming
that,
in
the
broader
sense
we
want
to,
we
want
to
go
after
namespace
base
charting
and
that
in
that
sense,
you
can
think
that
they
bays
partitioning
scheme
as
a
local,
optimization
right
it
optimizes
for
that
problem,
but
nothing
more
now.
If
we
wanted
to
explore
the
global
problem
more
and
go
after
namespace
charting
that
I
think
that
raises
products.
Our
concerns
with
even
with
the
audit
audit
log
feature
an
example
for
that
is.
B
There
is
two
ways
to
access
that
data.
One
is
on
the
project
level.
That's
all
fine,
but
that's
inside
a
namespace
goes
well
with
the
Charlie
idea
of
my
namespace.
The
other
one
is
on
the
instance
level,
which
is
instance
wide
gives
the
admin
a
way
to
look
into
the
audit
log
and
that
doesn't
play
nicely
with
the
namespace
based
routing
idea.
So
if
we
want
to
explore
that
more,
there
are
certainly
product
questions
that
are
going
to
come
up
because
those
features
that
don't
go
well
with
the
namespace
base
charting.
B
They
are
either
going
to
break
if
we
keep
it
the
same
way
or
become
much
more
expensive
because
we
have
to
change
them
so
that
they
still
work
in
the
namespace
base.
Sharding
world
I
think
this
is
very
valuable
to
explore.
Even
for
the
audit
log,
because
those
things
are
likely
to
happen
much
more
across
the
application.
D
Yeah
I
think
we
need
to
have
a
discussion
forum
or
clothes
that
are
instance
wide
which,
as
you
know,
it's
like
some
of
the
dashboards
Center
Project
SEARCH,
workflows
things
like
that
audit
logs
is
another
one.
That's
wide
right.
If
you're
an
admin
I'm
the
guy
comm
you're
gonna
want
to
have
some
kind
of
instance,
Y
access,
so
that
is
important.
I
think
a
question
would
be.
D
Can
we
partition
on
on
two
things
like?
Is
that
an
option?
Can
we
have
a
one
like
I'm,
not
sure
how
that
work,
with
Charmin
you're
gonna
shirt
on
one
thing
right
and
you're
counting
partition
on
two
things:
either
I
don't
know,
but
anyways
a
bit
curious
one.
That
I
think.
One
of
the
reasons
trying
I
wasn't
for
the
previous
conversation,
but
on
on
the
namespace,
is
that
it
provides
potentially
some
isolation.
D
If
they,
if
someone's
abusing
a
service,
then
the
database
should
be
relatively
not
impacted
for
people
outside
of
that
particular
namespace
and
secondarily
it
could
could
potentially
have
some
help
on
data
residency
unclear
what
that
would
be.
But
if
you
wanted
to
have
potentially
a
you
know,
database
somewhere
with
your
projects
that
have
to
live
in
Europe,
that
could
be
helpful,
so
I
think
that's
part
of
what's
it's
driving
at
as
far
as
I.
D
My
previous
context,
so
those
are
probably
some
pretty
good
benefits
to
have
so
I
opened
up
an
issue
to
explore
that
partitioning.
A
key,
miss
release,
I'm,
not
sure
I,
have
any
insights
right
now.
The
two
competing
things
right,
one
is
the
per
workflow
and
what's
the
best
Christian
key
for
that
work,
though
the
otherwise
sort
of
these
high-level
goals
of
isolation
and
potentially
into
residency,
part
of
it
might
be
speed
exploring
do
we
actually
get
the
residency
benefits
or
not?
Is
isolation
actually
going
to
be
beneficial
or
is
the
like?
D
The
initial
manager
layer,
coordinator
layer
still
gonna
go
down
and
have
it
and
be
impacted
and
then
on
the
audit
events,
you
know
if
we
do
go,
things
makes
ID.
What
are
our
options
to
try
to
continue
scaled
risk
in
other
similar
types
of
workflows,
because
you
know
isolation
is
a
thing
and
we're
by
not
gonna
figure
out
all
the
limits
or
implement
all
the
limits
that
people
can
potentially
abuse
soon
right,
and
so,
if
you're,
a
major
kind
of
rise
customer
get
lab.
D
B
What
I'm
not
clear
about
at
the
moment
is
if
we
actually
want
to
go
in
in
the
direction
where
we
sort
of
isolate
users
to
to
a
namespace,
we
can
find
them
into
a
namespace
that
sort
of
has
come
up
as
an
as
the
reason
why
namespace
base
sharding
works
right
because
everything
is
inside
a
namespace,
but
the
reality
is
that
users
are
not
and
I
I
wonder.
Do
you
know
if
there's
any
like
efforts
going
in
that
direction
or
exploring
that
idea
of
confining
users,
the
namespace,
yeah.
D
So
we
have
the
idea
of
the
method.
Manage
team
is
working
on
this,
the
concept
of
spaces.
It's
basically
like
awesome.
Our
word
rooms,
like
group,
managed
accounts
where
your
account
is
tied.
The
company.
The
reason
that
is
happening
is
because
of
a
number
of
corporate
data
controls.
Then
you'd
be
implemented,
for
example.
One
event
that
really
did
happen
is
that
someone,
forked
a
company
private
project,
interpersonal
nice
space
and
once
that
happens,
that
company
loses
all
rights
over
the
personal
namespace.
A
A
D
Think
ride
to
evaluate
the
key
right
now
so
to
the
degree
that
we
can
write,
maybe
run
some
additional
tests
and
and
estimate
what
growth
would
look
like.
I
think
that
would
be
good
right
if
we
can
say
look
without
partitioning
an
audit
of
events
on
time
scale
on
time,
stamps
we're
going
to
run
into
major
problems
here
or
something
like
that.
I
think
that
would
be
helpful
right
to
counterbalance
some
of
the
I
would
say
strategic
benefits
we
might
get
with
shorting
on
namespace,
around
isolation
and
potentially
get
a
residency
like
I.
D
Think
there's
two
completely
concerns.
One
is:
does
the
audit
events
workflow?
Can
that
continue
to
scale
given
the
access
patterns
of
that
workflow
versus
we
want
to
have
isolation,
finished
data
residency
in
a
state
of
isolation?
Interesting
could
get
us
there
so
that,
like
one's
a
global
thing
and
one's
a
local
optimization,
but
if
audit
events
is
gonna
fall
over
without
time
stamp
conditioning
I'd
be
good
to
know,
because.
A
I
think
the
difference
here
is
if
we
we
move
forward
with
partitioning
on
timestamp,
we
pretty
sure
we'll
get
some
performance
improvements
on
what
we
know
today
and
it's
something
that's
actionable
now
right,
whereas
with
the
name
space
partitioning
on
audit
events,
we
know
we're
gonna
break
some
things
based
on
current
functionality.
It's
still
if.
A
It
could
be
both
so
by
breaking.
It
means
just
functionally:
it's
probably
not
gonna
return
in
a
time
frame
that
either
the
user
wants,
or
it's
just
gonna
timeout
right
if
they
are
under.
It's
then
probably
answers
better
access
patterns,
if
they're
in
their
own
namespace.
When
they're
looking
at
odd
events,
maybe
it'll
work
if
it's
coped
down
or
it
might
require
some
product
changes.
I
can't
remember
what
andre
has
told
me
like
20
minutes
ago.
No.
B
It's
exactly
like
that
we
would
optimize
for
them
by
namespace
case,
and
that
might
mean
nothing.
For
example,
the
admin
UI
I
would
guess
that's
likely
to
break
on
good
luck
as
it
looks
now,
but
I
really
don't
know
if
that's
working
at
the
moment
at
all
like
it
could
also
be
broken
and
nobody
knows
or
are
we
even
using
that
isn't?
Is
it
important
to
support
that
for
a
good
luck?
That's
another
question.
D
I,
don't
know
what
the
other
comments
in
it
might
be
irrelevant,
given
we
have
cabana
and
other
tools
that
we
utilize
instead
of
instead
of
that,
so
might
be,
but
no
one
actually
goes
there
and
looks
at
it
at
the
instance
level.
Let's
apply
a
question.
We
should
ask
the
comm
the
m14
with
a
security
team.
Whoever
might
use
that
thing.
My
guess
is.
In
general,
you
would
like
people
to
to
operate
gitlab
as
our
customers
have
operated
and
so
providing
in
in-app
tools
to
do
security
investigations,
and
things
like
that
is
probably
good.
D
That
way,
you
don't
our
tools,
but
it
spice
I
question
we
should
dig
into
and
in
the
issue,
and
maybe
this
petitioning
key
one
as
R
exploring
like
the
challenges,
but
it
sounds
like
either
way
from
then
printed
purchases
on
namespace.
We
need
to
figure
out
what
the
path
forward
is
an
auto
of
events
and
to
your
point,
maybe
it's
irrelevant
on
the
instance
little
workflows,
but
even
in
a
namespace
is
it
still
scalable
I,
don't
I,
don't
know
for
some
of
our
larger
enterprise
customers
I
kept
you
get
50
dozen
users.
D
A
Different
ways:
anybody
waiting
for
decisions
on
whether
we
teams
to
Tuckahoe
namespace
on
partitioning
audit
events
versus
agree
with
created
at
partitioning
yep,
your
block,
okay
and
that
might
get
yeah.
We
can
I
get
the
next
half-hour
to
talk
about
it
in
working
group
and,
like
I,
said
I'm
free
after
that.
If
we
want
to
continue
this
conversation,
because
we
are
down
to
five
minutes
for
this
meeting,
we
haven't
answered
the
questions
yet
and
I
have
some
concerns
about
what
Josh
said
about
isolation
residency.
A
That
seems
to
further
complicate
partitioning
strategies,
but
you
leave
that
for
a
later
conversation.
So
thinking
in
the
recap
on,
if
the
partition
on
created
that
it's
not
going
to
break
anything-
and
we
feel
like
it's
gonna-
benefit
the
application
immediately
with
partitioning
on
namespace
on
audit
events,
we
do
feel
like
it's
gonna
break
and
probably
require
a
lot
of
changes
that
have
enough
high-level
summary.
But.
A
Especially
when
we
try
and
transition
that
into
other
areas
and
in
the
database
where
we
want
to
partition
that
table
all
the
names,
please
right
so
I
guess
that's
a
product
question
right.
Do
we
want
the
immediate
performance
improvements
on
audit
or
do
we
want
to
focus
on
the
long-term
goals
of
this
charting
implementation?
Why.
D
I
think
if
it's
a
difficult
to
wade
or
write
if
it
go
down
and
partition
on
art
events,
if
we
do
end
up
partitioning
on
main
space,
what
is
that?
What
does
that
mean?
Do
we
have
to
essentially
drop
at
a
drop
table
like
I,
don't
know
what
that
what
happened
there
I
think
that
would
be
interesting
to
know
more
on
if
it's
a,
if
it's
a
one-way
door
to
somebody,
it's
fine,
not
a
one-way
door,
but
how
much
time
you
missing
would
it
be
to
switch
over.
A
A
Works
I
will
set
up
a
meeting
then
so
we
can
continue,
because
we
still
need
to
look
at
thirteen
point,
Oh
see
what's
still
in
there.
Maybe
folks
can
move
them
in
half
an
hour
if
you're,
not
in
the
meeting
and
we'll
have
some
more
information
from
the
chart
in
life.
No
all
right
over
to
the
other
meeting.