►
From YouTube: 9.5 Retrospective
Description
Welcome to our Product team's retrospective of our 9.5 release!
Follow along in our retrospective doc: https://docs.google.com/document/d/1nEkM_7Dj4bT21GJy0Ut3By76FZqCfLBmFQNVThmW2TY/edit?usp=sharing
A
All
right,
it's
time,
welcome
everybody
to
the
retrospective.
We're
talking
about.
One
went
well
what
went
wrong
and
what
we
can
do
better,
especially
in
regards
to
get
that
line
is
fine
and
today
I'm
starting
off,
because
this
month
you
know
it's
vacation
time
for
the
most
of
the
world,
meaning
we
are
low
capacity.
We
cannot
get
as
much
done.
Get
this
amazing
community
contribution
of
GPG
commit
verification.
B
You
yeah
one
thing
that
went
really
well
with
some
caveats
that
I
think
as
a
teacher
in
the
retro
is
that
we
did
actually
process
to
more
background
migrations.
This
is
cool
because
last
time
we
get
a
first
one.
This
time
we've
done
two
more
and
these
were
kind
of
I,
don't
wanna,
say
more
complicated,
but
they
were,
they
were
different.
They
won't
SQL
only
migrations.
They
also
involved
Ruby
code,
unlike
Ruby
models.
So
it's
really
awesome
that
we
were
able
to
process
both
of
those
we
estimated
them
to
take
about
a
week
each.
B
C
It's
been
enormous,
ly
helpful
for
myself
and
I'm
sure
for
the
product
team
and
I've
seen
a
lot
of
folks
engineers
on
the
collar
right
now
who
actively
moved
their
issues
through
these
two
stages
and
development
and
in
review,
and
it
helps
a
lot
with
our
process
and
getting
that
visibility
and
being
able
to
manage
issues
ahead
of
time,
knowing
what's
what's
already
in
what
is
still
in
review
and
so
on
and
so
forth.
So
that's
that's
been
a
huge
when
10
Europe
yeah.
D
But
when
well
was
whole
performance
improvements
that
we
have
done
on
one
hand
on
the
front
inside,
so
we
have
improved
a
lot
with
different
techniques.
But
what
was
really
serious
that
we
have
so
many
performance
improvements
to
put
even
in
the
release
post,
that
there
was
a
very
long
list
and
it's
really
great
to
see
all
the
different
performance
improvements,
making
an
actual
measurable
effect
on
all
levels
in
all
pages
and
I'm,
really
looking
forward
to
all
that
improvements
on
the
next
releases
Joshua.
Next,
thanks.
E
F
E
Again,
give
it
a
whirl
again
we're
running
it
ourselves
on
get
lab
comm,
and
it's
been
running
great,
so
Gerber
to
the
team
and
were
excited
to
to
get
that
out
to
more
folks
as
well.
The
last
item
is
that
we
did
ship
automated
monitoring,
which
means
that,
if
you're,
taking
advantage
of
unemployment,
you
can
now
seamlessly
and
without
any
work
on
your
behalf,
start
to
get
a
response.
E
G
A
lot
yeah,
so
we
had
put
in
process
some
new
ways
of
writing
code
in
this
release,
because
we
had
a
couple
of
really
big
issues
that
we
were
working
on.
This
process
is
where
one
person
is
writing
the
code
and
one
person
is
writing
the
tests,
which
worked
really
well
as
a
pair
programming
situation,
which
is
something
that
we
haven't
done
very
much
agate
lab
and
we
learned
a
lot
in
the
process
and
I
will
link
the
docs
shortly
here.
B
Thanks
Jacob,
so
group
issue
boards
didn't
make
it
in
mind
if
I,
if
we
were
hoping
to
get
it
in
9.5,
but
we
made
the
call
I
think
on
the
3rd
there's
a
retrospective
document
link.
There
I
think
a
couple
of
headlines
from
that
document,
though
probably
just
we
as
lead
to
do
a
great
job
of
picking
people
to
work
on
it.
B
B
G
B
My
normal
accent
he's
asking
if
we
have
to
disable
sidekick
jobs
for
the
entire
deploy,
because
when
psychic
jobs
aren't
running
we're
still
up,
but
there's
a
bunch
of
stuff,
you
can't
do
so
for
a
bunch
of
things
we
may
as
well
be
down
because,
like
you,
can
merge
a
load
request
or
you
can't
update,
merge
request
so
having
them
down
for
a
long
period
means
that
we
can't
do
our
work,
which
was
it
means
that
people
who
are
using
get
our
dog
home
can't
do
their
work
either.
James
Lu
linked
to
an
issue.
E
Okay,
I'll
jump
in
James.
If
you
want
to
chat,
certainly
taught
me,
but
one
thing
with
9/5
is
that
some
would
be
matter
my
sister's
continued,
and
so
we
really
do
need
to
figure
out
a
better
method
of
testing
these
right
now
we're
getting
some
bugs
into
the
final
packages
which
isn't
good
and
we're
finding
out
only
after
people's
submitted
issues.
So
we
do
need
a
better
way
of
identifying
and
finding
these
before
we
ship,
and
so
hopefully
we
can
improve
that.
Your
going
forward,
also
on
things
that
didn't
go
great,
is
a
pivotal.
H
E
F
Yeah,
so
a
couple
things
that
we
mentioned
database
migrations
but
exposed
underlying
problem
with
our
infrastructure,
with
slow
discs
on
Gil
optic
column.
So
because
these
migrations
are
doing
a
lot
of
work,
they
caused
a
lot
of
load
on
our
database,
which
therefore
brought
down
get
lab
comm.
So
we
had
a
number
of
those
adages
that
that
were
fixed
with
the
new
failover
and
a
new
discs,
but
it
was
a
problem
for
a
week
or
so
in
the
second
main
point
I
want
to
bring
out
is
the
LDAP.
F
We
continue
to
have
problems
where
our
release
breaks,
LDAP,
logins
from
one
reason
or
another.
In
this
case
it
was
a
minor
issue
that
could
be
worked
around,
but
it
was
still
people
expected
to
upgrade
and
half
log
ins
work
and
it
stopped
working.
So
I
think
we
need
to
really
put
some
effort
into
this,
because
it's
happening
will
these
after
release
yep
Caleb
QA?
That's
when,
let's
later
on
and
I'll
discuss
that
Tim.
I
I
H
Yeah,
so
the
CCD
team
was
working
on
auto,
deploy
improvements
and
those
slipped
again.
Actually
it's
been
two
months
now
and
so
we've
discussed
with
the
CCD
team
elsewhere.
I
don't
have
a
link
handy,
unfortunately,
but
one
of
the
key
pieces
there
was
having
some
automated
testing
capabilities
for
end-to-end
deployments
using
auto
deploy.
We
don't
have
that
yet,
and
we
need
to
do
that.
Another
thing
is
capacity
and
people,
and
so
we
did
actually
add
Alessio
to
the
teams
that
are
Jacob
s.
G
G
F
J
F
Yeah
absolutely
and
please
remember,
to
create
an
issue
or
something
around
it,
because
these
flaky
specs
can
be
fixed.
You
know
people
created
them
and
we
can
actually
allow
them.
It
can
be
fixed
in
a
simple
way.
So
please
identify
them.
Don't
you
know
retry,
if
you
retry
in
more
than
once,
then
it
probably
means
we
need
to
investigate.