►
From YouTube: 2020 04 20 Database Sharding Working Group
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
You,
oh
so
want
to
check
in
on
the
Postgres
11
rollouts
see
Jose
starting
testing
this
week
and
its
knee
blocker
Brett
will
announce
the
date
of
the
upgrade
and
then
there's
one
follow-up
item.
We
want
to
make
sure
that
we
re
enabled
the
9:6
testing
on
CI,
because
I
think
they
were
disabled
and
reduced
down
to
weekly,
since
we
thought
we
were
gonna,
be
rolling
out
on
13.0.
A
There's
some
comments
in
there
on
the
issue
too,
for
follow-up
discussions,
I
think
Remy
was
commenting
on
it,
but
in
absence
of
any
other
DRI
I
put
Mac
on
there
make
sure
that
this
happens
so
Tanya.
You
are
the
quality
representative
right
now,
so
something
to
follow
up
on
to
make
sure
that
until
we
actually
upgrade
the
PG
11,
we
need
to
make
sure
that
we
are
continually
testing
on
nine
point.
Six.
B
A
A
All
right
kick
those
No.
So
moving
on
what's
been
done,
met
with
legal
regarding
situs
community,
they
had
an
a
GPL
license
and
basically
the
feedback
from
legal
was
it's
a
non-starter.
We
just
didn't
even
pursue
it,
so
we're
not
going
to
cite
us
enterprise
licensing,
also
going
to
be
a
non-starter,
so
we
will
close
out
our
situs
work.
We
looked
at
finishing
up
the
testing
that
we
were
doing
on
situs,
but
it
would
be.
The
estimates
were
take
a
couple
weeks
fully
dedicated
to
finish
the
testing.
A
C
I
have
one
sing
on
the
show.
We
actually
keep
the
cluster
until
we
reach
our
conclusion
of
whether
we
are
going
to
abstract
at
a
service
or
still
investigating
sharni
via
the
sharding
discussion,
because
if
we
still
want
to
kind
of
I
mean
this
could
be
it's
going
to
be.
A
data
for
us
to
justify
a
few
charting
is
actually
a
worse
doing
at
the
moment.
Right,
we
don't
know
yet
and.
A
C
Think
right
now
the
primary
concern
right
here
is
charting
and
partitioning
will
give
us
more
scalability
resolve
more
to
more
scalability
concerns.
Then
the
performance
comes
on.
I
mean
performance
comes
second
to
the
skin
immunity.
So
that's
what
I'm
thinking
we
shall
keep
the
environment
conclusion
whether
we
go
with
shardene
or
now.
A
A
So
we'll
leave
it
and
then
also
moving
on
to
the
next
point
so
met
with
various
stakeholders.
So
in
the
meeting
Camille
Stan
Andrew
database
team
talked
about
concerns
with
respect
to
charting
and
talked
about
some
of
the
alternatives
and
based
on
that
hour-long
conversation
with
everybody.
It
was
involved.
Service
extraction
was
a
preferred
path
and
there's
the
dot
link
to
the
doc
and
there's
a
recording.
We're
gonna
have
a
follow-up
meeting.
D
E
E
G
E
Yeah
I
was
this
from
authority,
cut
perspective,
okay,
so
my
question
is:
this
is
a
path
that
I
think
is
very
interesting,
especially
if
they're
gonna
have
post
criticism
back
in
anyway,
because
that
does
not
not
draw
dry.
The
operational
cost
exponentially
now
is
there?
Is
there
any
intention
to
also
consider
the
custom
charting
approach,
because
service
extraction
will
take
significant
development
effort
and
time
to
implement
and
I'm,
not
sure
if
that
would
be
higher
or
less
than
a
custom
charting
alternative
by
the
way
both
can
be
also
combined
at
develop
at
the
same
time?
A
So
yeah
there's
there's
lots
of
different
approaches
and
Christopher
asks.
Can
we
get
a
ruffle
proposal
of
which
service
extraction?
We
should
consider
first?
So
there's
there
are
some
that
are
in
flight
like
contain
a
registry
and
prefect
that
are
already
they
already
have
their
own
database
and
I
think
they're
looking
at
service
extraction,
and
then
there
are
existing
services
that
we
are
talking
about
pulling
out
and
extracting
into
their
own.
Like
the
CI
CD
Damon
proposal,
that's
floating
out
there
and
you
know
there's
another
one.
A
That's
slipping
my
mind
right
now,
but
there
are
various
proposals.
We
haven't
taken
a
rough
cut
on
how
much
effort
any
of
these
would
take
and
I
know
that
the
sharding
versus
service
extraction
there's
going
to
be
strong
opinions
on
both
sides,
so
we'll
keep
iterating
on
this.
We'll
get
some
feedback
later
on
this
week
in
the
follow
up
conversation,
but
that's
just
one
recap
on
what's
been
done
this
week
and
we're
not
in
a
final
decision
yet,
but
this
is
the
one
that
the
stakeholders
that
were
involved
in
the
conversation.
A
D
The
yeah
that
my
my
previous
docket
on
sharding
methods
included
service
extraction
and,
yes,
I.
There
are
lots
of
downsides
also,
when
considering
what
service
to
extract
you
have
to
consider
how
much
of
the
database
load
is
actually
caused
by
the
extraction
as
would
be
removed.
So
how
much
are
you
actually
gaining
on
that
surface
extraction?
D
A
E
I
I
don't
have
any
any
preference
actually,
if
I
would,
if
I
would
ask
for
the
ideal,
I
would
say
both
at
the
same
time
would
be
the
best
way
to
go,
and
now
I'll
read
the
document
in
and
with
Tom
and
I
access
and
I'll
drop.
Some
notes,
if
appropriate,
I'll
just
wanted
to
say
that
one
of
the
main
problems
to
solve
with
service
extraction,
which
I'm
sure
is
over
there,
I'll
drop
a
note.
E
E
H
Very
good
and
the
practical
way
of
doing
that
could
be
to
start
with
partitioning
for
those
services
that
we
extract
already,
and
that
gives
you
the
same
idea
of
design
for
that
application
and
later
on,
you
can
use
that
too
to
do
sharding
or
whatever.
Yes,
the
applications
already
like
sliced
correctly
or.
E
F
G
F
Up
another
interesting
point
which
be
really
important
to
get
this
in
the
document.
My
opinion,
just
because
I've
seen
in
previous
lysis
become
super
critical,
which
is
kind
of
auditing
and
reconciliation,
to
make
sure
that,
like
we
have
consistency
because,
there's
generally,
you
know
cruft
left
over
from
an
infrastructure
perspective.
Android
do
we
have
anything
like
that
going
on,
or
you
know
like
we're
double
checking
on
things
right
now.
F
I
might
be
a
good
one
to
add
Craig,
just
a
kind
of
kalaiy
I
mean
it
becomes
a
bigger
issue
with
services
when
you
do
service
extraction
or
services
with
separate
DBS,
but
in
general
I
think,
as
assistant
gets
larger.
We
need
that
anyways.
Like
a
good
example.
Is
you
know
if
you
have
things
moving
over
nest
from
sd,
d2
or
SSD
to
normal
hard
drives,
because
you're
trying
to
save
money,
you
could
end
up
with
them
being
in
both
places,
because
you
know
processes
fail
and
those
kinds
of
things.
A
F
F
A
Yeah
so
I
have
a
canoe
and
I
know.
A
lot
of
people
want
to
join
the
meeting,
but
I
have
a
concern
that
you
get
too
many
people
in
a
room
and
it's
hard
to
come
to
a
decision
and
a
direction
going
forward.
So
I
will
I
will
do
what
I
can
do,
invite
the
right
folks
and
no
offense.
We
don't
get
invited,
but
I
want
to
make
sure
it's
an
effective
meeting.
So.