►
From YouTube: 2021-02-09 Delivery team weekly rollbacks demo
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
Hey
everyone
so
another
exciting
tuesday,
so
actually
two
things
that
made
me
no
actually
three
things,
so
these
would
be
good
to
know
in
advance
for
future
rollback
demo.
So
the
first
one
is
most
significantly.
Reliability.
Teams
are
a
bit
swamped
out
with
incidents,
so
certainly
don't
want
to
risk
adding
any
more
work
to
their
place.
That's
the
main
reason
that
I
don't
think
we
should
do
a
full
test
rollback
today.
A
Second,
one
is,
and
maybe
an
interesting
discussion
point
is
whether
it
adds
any
risk
to
the
security
release
and
then
the
third
one
which
actually
I
just
certainly
noticed-
that
is
somebody
just
enabled
an
experiment
on
staging
about
half
an
hour
ago,
and
so
that
would
be
a
clash
as
well
so
also
in
advance
need
to,
I
guess,
like
book
out
staging
so
I'll,
make
a
note
so
that
we
do
that
next
week
and
like
an
hour
or
so
before,
we
actually
or
even
earlier,
maybe
give
heads
up
that
staging
is
going
to
be
unstable
at
this
time.
A
So
three
good
reasons
not
to
attempt
to
full
roll
back
today,
but
hopefully
we'll
still
get
some
good
value,
because
I
think
the
most
difficult
thing
that
we
saw
last
week
was
attempting
to
find
a
run
book
and
work
out
whether
a
deployment
was
safe.
So
your
ex
you
want
to
give
us
a
demo
of
how
that
looks
now
and
we
can
see
if
there's
any
further
improvements
we
can
identify.
So
for
everyone.
That's
not
uric!
A
I
put
together
an
mr
which
you're
ex
merged
in
which
is
absolutely
by
no
means
not
the
best
I've
ever
seen
or
even
close,
but
hopefully
it's
a
decent
first
iteration.
So
should
we
give
it
a
test.
B
B
Let's
see
so
roblox
shouldn't
be
tempted.
You
know,
let's,
for
the
sake
of
argument,
assume
that
we
do
have
to
roll
back
so
first
step,
we
go
to
the
security
project
and
then
it
says
on
the
right:
nfk
to
operations
and
environments.
B
I
think
probably
the
first
step
here
is
to
link
to
that
directly.
Let
me
write
if
somebody
could
write
that
to
the
google
doc,
because
we
can
just
link
to
this
directly.
That
might
be
a
little
easier
because
then
you
don't
have
to
go
through
all
these
menus
on
this
side
to
figure
out
which
one
it
is.
B
The
next
step,
then,
would
be
to
pick
the
environment
yeah.
Let's
do
staging
there,
you
go
that's
the
next
step.
You
see,
output
from
chelps
run
order,
deploy
status
to
find
the
required
deploys.
Okay!
Well,
let's
see
what
happens
if
we
do
that
that
can
go
away.
B
Oh
there
we
go
chip
staging-
let's,
let's
see
already,
checks
to
get
results.
Okay,
so
staging
so
we
have
to
branch.
Compare
with
I
let
me
see.
I
use
the
output
file
to
find
required
deployments
to
compare
yeah.
I
think
the
first
question
is
so.
Let's
say
we
are
on
staging.
B
In
the
sorry,
the
current
state
of
the
runbook,
it's
not
really
clear
what
I
would
have
to
compare
so
as
an
example,
this
will
spit
out
revision,
f6b
and
let
me
double
check
if
that's
actually
the
latest
one
yeah.
So
that's
the
latest
one
that
is
the
political
staging,
so
I'm
guessing
what
you
would
then
have
to
do
is
oh,
why
is
it
not
clicking?
Oh
there's
some
things
blocking.
B
D
Value
in
this
case,
if
I
remember
what
what
I
did
last
week
is,
I
said
something
like
that:
gives
you
a
quick
link
that
you
can
click
and
then
you
you
can
change
the
versions,
because
from
this
page,
there's
no
easy
way
to
tell
compare
the
difference
between
these
two
versions,
and
I
already
had
that
thing
opened.
So
I
think
this
is
what
amy
captured
from,
because
amy
you've
wrote
the
rumble
price
yeah.
D
D
B
Yeah
so,
in
other
words,
I
think
the
chat
up
status
on
in
its
current
state
at
least
is
not
useful
for
determining
if
we
can
roll
back
or
specifically
for
determining
the
the
range
of
commits
to
check.
Basically.
B
It's
an
issue
for
that
which,
at
this
point,
I
think,
is
about
a
year
old
to
redo
that,
but
it's
apparently
not
really
a
priority,
but
anyway,
I
think
probably
the
better
approach
for
now
then
would
be
to
say.
Okay,
you
know
step
one
go
to
this
staging
or
production
overview
step,
two
figure
out.
Okay,
what
is
the
deploy
that
we
actually
want
to
roll
back?
It's
probably
the
latest
on
most
of
the
time
and
then
say:
oh
then,
just
pick
the
previous
staging
deploy.
B
B
Yeah,
let's
see
so
I
want
to
compare.
B
These
two:
let's
see
if
this
will
work
yeah,
so
I
think
I
have
to
flip
it
around.
Actually,
oh,
we
used
to
have
a
button
for
that.
Yeah.
B
I
think
they
never
used
it
like.
Sometimes
when
you
use
gitlab,
I
feel
like
such
an
old
man.
It's
like
how
how
do
computers
work-
let's
see
so,
let's
flip
it
around.
B
So
let
me
see
so
we
are
comparing
four
nine,
oh
yeah,
and
then
here
it
shows
it
in
reverse,
like
that
that
didn't
work,
that's
what
I
entered
first.
Am
I
saying
that
correctly
actually
hold
on
no
okay,
so
the
source
should
be.
The
current
deploy
target
should
be
the
previous
one
very
convenient.
B
Let's
confirm
that
is
actually
the
case.
So
if
we
go
back
to
that
first
one,
so
we
do
source
f6.
That
was
the
latest
one
right.
Yes,
so
source
is
the
latest
targets
will
be
the
previous
one
enter,
let's
see,
okay,
so
that
seems
to
work
so
source
latest
target
previous
one,
and
then
we
end
up
here
on
this.
B
D
D
You
write
more,
it
gets.
It
sticks
to
what
you
write.
So
if
you
do
a
db,
slash,
post,
migrate
or
db
it,
it
starts
removing
stuff,
yeah,
it's
fuzzy,
so
here's
the
tv,
slash,
post
and
a
change
log
shows
up.
E
B
B
So
coming
back
to
the
searching
yeah,
we
probably
need
something
better
for
that,
because
the
fuzzy
search
I
mean,
I
guess,
if
you
type
migration
it
it
limits
it
enough.
But
it's
a
bit
of
a
hit
and
miss.
A
D
A
B
D
B
Guessing
what
you
could
do
as
an
alternative
if
I
start
a
terminal,
gitlab
gdk
gitlab
branch
in
my
own,
so
I
guess
what
you
could
do,
then
is
you
know,
take
this
range,
get
a
lock!
D
You
need
to
fetch
from
there's
another
thing:
you
need
to
fetch
from
security,
because
other
deploy
branches
are
only
on
security
yeah,
but
the
commit
should
exist
in.
D
B
Okay,
so
that
then
won't
work
and
guessing
what
we
could
do
eventually
in
chat,
ups
is
or
some
other
tool
add
a
command
that,
basically
you
give
it
a
deployment,
and
then
it
fetches,
the
previous
one
you
know,
gets
the
comparative
and
then
figures
out
if
there's
migrations,
but
for
now
this
is
a
somewhat
tedious
manual
process.
B
Okay,
so
we
have
to
div
we've
determined
we
can
in
theory
roll
back.
Let's
see
how
to
roll
back
after
completing
the
steps
to
check
for
migrations,
follow
the
deployment
drawback
steps.
So,
let's
open
that.
A
So,
let's
not
do
those
today
right.
B
D
B
Let's
see
creating
a
new
deployment
for
rolling
back
gitlab
important
before
initiating
that
I
did
the
dip.
Okay,
let's
pretend
we
did
that
application
rollbacks
are
initiated.
You
know,
let's
pretend
we
also
did
that
when
a
checkups
command
is
received.
B
Manually,
so
that's
if
I'm
thinking
here
maybe
this
is
just
perhaps
me
being
a
bit
pedantic
right
now.
If
you
quickly
scan
through
this,
it
appears,
as
this
section
with
the
environments
is
required
when
in
reality,
it's
an
option.
If
you
want
to
do
it
manually,
so
I
think
in
that
case
it's
best
to
perhaps
push
that
further
down
the
document
to
say
hey.
If
you
want
to
do
this
manually
here,
you
go.
B
B
I
want
to
be
able
to
scan
through
this
document
quickly
and
say:
oh
for
this
step,
I
have
to
do
this
and
then
I
have
to
do
that
and
not
this
stuff
in
between
it's
like.
If
you
want,
you
can
also
do
this.
It's
like!
No,
if
it's
optional,
you
know
put
it
at
the
end,
basically
or
collapse
it,
but
don't
break
the
the
sort
of
flow
as
you
go
through
it.
B
Let's
see,
it's
also
possible
to
just
start:
blobby,
yeah,
okay,
again
same
thing,
so
screenshots.
I
think
that
relates
to
the
optional
stuff.
Let's
see
cicd
pipeline
for
deployer
upgrade
pipeline
upgrade
pipeline
multi.
B
I
think
my
first
response
to
seeing
this
if
I
were
a
release
manager
and
I
had
to
roll
back,
my
first
response
would
be
basically
what
the
hell,
not
as
saying
it's
bad,
like
it's
actually
really
well
written,
but
I
think
it's
written
for
a
different
scenario
like
this
very
much
reads
like
how
do
rollbacks
work.
How
do
we
implement
them
and
that's
very
interesting,
but
not
I
think,
when
you're
actually
doing
a
rollback,
I
think.
A
That's
a
good
point
actually,
so
it's
something
that
mary
and
I
have
chatted
about
with
quite
a
lot
of
ad
documentation.
It's
almost
like.
We
need
a
run
book
style
of
literally
just
the
steps
that
you
go
through,
like
we
do
with
our
release
issues
and
then
link
off
to
the
kind
of
here's.
Why
it's
written
this
way
and
the
kind
of
the
actual
explanation
of
why
it's
set
up.
B
Yeah
and
that's
I
mean
I'm
gonna
sound
a
little
bit
indicate,
but
that's
basically
wrong
book
is
this.
I
think
if
I
look
at
this
document,
so
the
first
half
more
or
less
seems
reasonable.
Where
we
say,
oh,
if
you
want
to
do
this,
here's
the
commenter
run,
that's
it
and
then
here
at
the
end,
we
sort
of
added
this
whole
section,
which
I
would
almost
say
should
go
in
the
handbook,
not
in
a
run
book,
because
this
is
just
saying:
oh,
you
know
here's
how
it
works.
B
Right
so
we
need
the
rambo,
so
in
that
case
basically,
I
think
this
first
chunk
should
be
moved
somewhere
else
essentially,
but
either
way.
Let's
see,
is
there
anything
else
that
I
would
have
to
read
in
the
event
of
a
rollback?
I
think
that's
part
of
the
problem
here
like
I've.
Let's
say
I've
run
this
command.
B
It's
not
clear
if,
at
that
point,
the
rest
of
this
document
is
relevant.
Like
there's
no
indication
like
oh
now,
you
need
to
wait
for
xyz
and
then
run
this
step.
You
know
further
down
the
line,
so
that's
a
bit
confusing.
I
would
imagine
this
produces
the
output
that
would
then
sort
of
guide
you
through
the
pros,
like
you
know,
pipeline
and
when
a
rollback
fails,
it
spits
out
something
but
right
now
it
seems
a
bit
like
like
the
road.
Suddenly
just
stops.
B
It's
like
you
know
a
bridge,
that's
unfinished.
Essentially,
let's
see
so
does
the
so
it
says:
follow
a
rollback
with
the
rollback
complete
the
user
experience
is
restored,
but
deployments
will
remain
blocked
until
a
fix
or
revert
makes
it
through
the
auto
deployment
pipeline.
B
I
mean
that
makes
sense,
maybe
rollback,
because
there's
a
bug
and
somebody
should
be
fixing
it.
I
think
what
might
be
helpful
here
is,
if
we
link
to
the
process
to
follow
for
that.
B
If
there
is
one
documented,
I
would
imagine
most
of
the
time
people
sorry
release
managers
will
know
that,
but
for
the
cases
where,
like
me,
you
haven't
done
it
for
a
little
while
and
you're
coming
back,
it's
kind
of
nice
to
have
a
reference
to
sort
of
clear
up
your
mind.
So
you
refresh
your
mind
about
the
the
process.
B
So
the
process,
essentially
the
rollback,
has
finished-
has
succeeded
at
that
point.
You
need
to
go
through
the
process
of
making
sure
that
you
know
there's
a
fix
submitted
for
whatever
triggered
the
rollback.
B
A
A
Form
of
like
what
how
we
handle
a
revert
like
today,.
B
Let's
see
so
severity
one
incident,
I
would
imagine
it's
this
because
there
is
a
deployment
rollback
as
a
remedy,
but
again
this
is
this,
isn't
a
rollback.
This
is
more.
A
sort
of
reference
guide
like
a
rollback,
is
really
just
basically
a
list
of
bullet
points,
kind
of
like
a
release,
task
issues.
It's
like
you
know
on
the
20th
run.
This
command
then
run
that
command
boom
down
so
yeah
right
now.
B
D
I
have
a
comment
on
this,
so
the
way
we
are
kind
of
designing
this
process
is
that
you
don't
start
rollback
because
you
want
to
right.
So
probably
is
a
sub
rumba
run
book
within
the
the
other
one,
but
you
were
referring.
The
s1
probably
would
be
the
case
also
for
s2
or
s3
right
once
we
automate
it
and
we
know
that
it
works
well,
it
doesn't
have
to
be
an
s1
if
it's
something
that
maybe
it's
true,
it's
enough
and
we
run
it.
D
D
D
There's
another
point
which
is
still
important
in
that
link
that
we
were
looking
at
so
the
the
rollback.
It
just
mentioned
that
this
in
the
text
that
the
rollback
will
start
on
staging,
which
is
fine
for
the
thing
that
we
are
doing,
but
in
the
event
of
running
this
in
production,
you
need
to
remember.
You
need
dash
dash
production
to
it,
and
this
is
not
mentioned
as
well.
B
No,
I
think
that
kind
of
comes
down
to
like
this
document
that
we
have
it
in
it's.
It's
not
a
wrong
book.
It's
a
reference
guide
where,
but
it's
I
would
say
it's
an
architecture
guide
because
it
explains
you
know,
flow
of
pipelines
and
all
that
stuff.
B
But
let's
do
the
part
of
figuring
out
the
package.
Let
me
share
my
screen
again.
B
B
Let's
pretend
I
don't
know
where
to
get
that,
so
I
go
back
to
environments
staging
thinking.
Maybe
it
is
shown
here.
Well,
it's
not
so
then.
What
I
could
do
is
perhaps
go
to
the
commit
wait
for
it
to
load
which
will
take
a
while
observe
how
the
ui,
for
a
moment
there
looked
all
funny.
B
Then
I
would
say
browse
files.
I
I
know
I
could
have
clicked
the
branch
too,
but
I
just
want
to
make
sure
there's
no
newer
commits
on
the
same
branch.
B
Then
I
guess
I
could
go
to
version
somewhere
in
this
big
pile
there.
We
go
and
very
helpfully
that
just
says
390p.
In
other
words,
it's
useless.
So
then
my
guess!
Well,
you
would
see
here
update
version
files
two
weeks
ago.
So
that's
probably
not
it.
B
Then,
let's
see
what
the
process
will
be
normally
what
you
could
do
if
the
deploy
is
recent
enough
is
run
auto,
deploy
status
and
then
I
guess
yeah,
you
would
see
here
the
package.
D
B
B
B
I
can
wait
for
that
to
load
then
go
back
to
slack,
and
so
it's
where
was
the
previous
again
shoot?
Now
I
have
to
compare.
B
B
Let
me
cheat
by
looking
at
slack,
okay,
so
yeah
for
order
deploys.
We
can
then
use.
I
suppose
that
whole
part-
or
is
it
just
the
state,
dash
right?
But
this
isn't
this.
Oh,
this
is
the
omnibus
format,
not
the
form
between
the
shadows.
B
So
I
think
here's
a
step
where
or
here's
a
stage
where
we
need
to
do
something
either
clear
this
up
in
the
run
book,
because
here
in
slack
we
see
that
the
package
version
is
39
big
date
dash
x,
dot
y.
However,
here
it
is
date
plus
x,
dot
y,
so
I'm
guessing.
I
could
change
that
to
just
a
dash,
but
that's
completely
not
obvious.
B
D
Have
a
simpler
approach,
but
still
terrible,
which
is
searching
in
the
announcement
channel
same
thing
that
you
did
on
the
tags.
But
I
search
in
the
announcement
channel
right
because
you
have
the
package
but
you're
right.
It's
just
replacing
plus
with
dash,
but
is
that's
true
in
the
incident,
is
terrible
because
in
the
example
you
have
dot
ee
dot
zero
and
you
kind
of
expect
to
maybe
match
the
same
thing.
But
this
is
only
true
for
batch
releases.
B
I
can
see
this
being
tricky
if
people
start
mentioning
this
shot
or
this
version
multiple
times,
hopefully
not
in
announcements,
but
it
is
possible,
but
then
through
here
I
could
figure.
Oh
wait.
You
know
this
is
the
package
version,
and
then
you
know
copy
paste
that
and
use
it
here
to
do
the
roll
back
yeah
for
this
process.
We
absolutely
did
something
better,
because
if
I
have
to
do
this
twice,
I'm
going
to
be
flipping
my
desk.
Basically
I
I
know
the
post
like
hey.
B
B
B
B
B
B
Yeah,
and
even
if
we
did,
I
don't
think
we
ever
had
a
need
so
as
a
starting
point
that
doesn't
really
matter,
because
now
we
use
auto
deploy
status
and
it
would
show
you
here
in
staging
currently
we're
on
version
b.
I
know
the
previous
version
was
version
a
I
don't
know
if
you
can
get
that
info.
I
suspect
not
because
we
probably
read
this
from
chef
and
jeff
prolio.
B
Excuse
me
probably
only
knows
what
we
have
right
now,
not
what
we
had
yesterday,
the
yeah,
the
alternatives
you
go
through
these
tags
and
somehow
derive
the
package
version
from
that
or
you
would
have
to
go
to
package
cloud
and
somehow
figure
out
what
the
previous
package
is
out
of
the
50
that
we
built.
E
B
B
We
probably
could
do
as
well
as
we
have
that
metadata
project
on
the
ops
instance
and
the
tracks
we
currently
deploy.
We
could
in
three
I
think
I
might
have
created
an
issue
of
that
and
reached
that
to
also
record
some
info
about
the
previous
deploy,
and
then
you
can
build
a
slack
tool
that
just
reads
that
info.
D
I
was
thinking
about
using
the
environment
feature
here
as
well,
because
there
is
a
bug
that
works
that
we
were
hit
by
a
couple
of
days
ago.
So
I
can
tell
you
where
those
values
are
coming
from,
because
we
had
monday
morning.
I
think
there
was
something
something
funny
going
on
with
those
revision.
So
revision
comes
from
an
api
call
to
that
instance.
D
Package
is
coming
from
chef.
Comparison
is
just
the
comparison
link
between
the
api
code
on
the
previous
one.
Now
we
have
a
new
thing
which
is
cannery
is
also
part
of
the
production
fleet.
Now,
which
means
that
you
may
have
the
wrong
answer,
which
is
what
happened
to
me?
I
was
expecting
a
big
difference,
say
no
production
and
cannery
are
in
line.
It's
exactly
the
same
thing.
No,
this
can't
be
true,
so
we
dig
at
this
with
jarv
and
we
figured
out
that
yes,
the
cannery
is
part
of
the
production
fleet
as
well.
D
So
you
have
five
percent
chance
of
getting
the
wrong
version,
but
if
we
do
rely
on
security
from
gitlab
security
environments,
we
can
get
the
latest
deployment
and
show
that
and
for
the
package.
D
B
We
already
do
that
with
the
deployer
on
ops.
Let
me
quickly
double
check
that,
because
at
some
point
prior
to
us,
adding
tracking
in
release
tools,
I
added
that
in
the
deployer,
let
me
see
where
the
operations
environments
we
have
too
many
sub
menus
production
all
right.
The
first
annoying
thing
we
use
different
environment
names
here,
let's
say
staging
last
deploy
six
days
ago.
B
Don't
think
that's
accurate,
but
you
know
imagine
for
a
moment
that
it
is
created
one
week
ago
deployed
six
days
ago
and
then
priced
it
at
three,
so
yeah.
So
it
seems
this
isn't
actually
always
running
and
we
always
tag
master.
So
that's
not
particularly
useful,
unfortunately,
but
we
could
do
something
yeah
with
release
tools.
I
I
do
feel
it's
maybe
a
little
weird
to
start
adding
git
text
just
for
this,
because
we
will
be
tagging
something
in
release
tools.
B
B
D
Single
pipeline
that
work,
we
are
doing,
we
no
longer
tag
omnibus
or
cng
directly,
but
we
tag
release
tools
and
then
there
is
a
special
stock
pipeline
that
runs.
So,
oh.
D
It's
already
the
case,
but
yeah
I
was
referring
to
actually
using
the
environment,
feature
to
track
the
those
deployments,
because
in
theory
I
don't
know
if
we
have
those
very
those
values
available
as
a
variable,
we
could
use
this
with
plain
gitlab,
ci
and
yaml
configuration
because
you
can
add
the
environment
things
and
writing
all
the
information.
B
D
B
What
we
could
do
is
extend
the
deployments
api
so
that,
instead
of
just
saying
hey,
I
deployed
this
random
shar
ref
whatever
you
can
also
give
it
a
version
name,
and
then
you
know
we
display
it,
and
then
we
could
use
that
to
record
the
auto
deploy
versions.
B
D
B
B
B
D
B
The
current
list
of
texts
that
we
have
from
listtools,
it's
not
clear
how
I
would
determine
which
of
those
tags
is
from
the
previous
deploy
based
on
a
eesha
or
a
package
or
tag
name
of
whatever.
We
have
currently
deployed.
D
This
is
why
I
was
referring
to
using
environments
in
release
tools
so
that
we
can
track
the
the
tag
names
when
we
deploy
them
in
every
environment,
and
then
we
can
query
the
metadata
project,
because
it's
exactly
what
we
are
doing
with
a
single
pipeline,
because
we
tag
release
tools
with
the
same
format
of
the
release
metadata.
Then
every
job
in
that
pipeline
can
ask
information.
Like
tell
me
cng
version.
Tell
me
all
this
information
and
you
can
grab
them.
D
So
in
that
sense,
a
chat.
Ops
command
can
do
the
same
right,
just
go
to
release
tools.
Tell
me
staging
deployments,
and
it
just
gives
you
the
list.
You
picked
the
last
two
so
last
current
one
in
previous
and
then
you
go
to
release
metadata
and
say:
could
you
please
give
me
metadata
about
those
two
things
and
you
have
all
the
information
to
build
package
names
or
comparison
links
and
what
you.
B
E
Route,
would
you
st
would
we
still
end
up
with
a
problem
where
the
environments
page
may
give
us
incorrect
information,
like
you
just
told
us
about
billabodago.
D
E
Captured
everything
we
just
discussed
in
existing
issues
as
either
supplementary
comments
or
adjust
to
the
issue,
descriptions
where
appropriate,
which
leads
to
the
next
discussion
point.
I
created
an
epic
for
just
a
blank
canvas
to
hold
all
of
the
action
items
that
I
created
from
the
last
time
we
had
this
meeting
nothing's
prioritized
at
all,
there's
no
labels
on
anything.
I
just
created
the
issues
and
dropped
them
somewhere.
E
A
Awesome
thanks
for
doing
that,
scarbeck
yeah.
I
have
some
admin
to
tidy
things
up
and
like
label
things
and
link
them
into
the
okr
epic,
but
that,
like
this,
is
fantastic.
So
what
do
so?
We
kind
of
got
two
threads
that
we
can
maybe
prioritize
before
next
week.
So
one
is
another
round
of
improvements
to
the
runbook
and
the
other.
One
was
one
we
raised
last
week,
which
was
investigating
assets
and
whether
we
could
actually
they
had
something
we
could
try
and
skip.
A
Do
they
sound
like
they
would
still
be
two
useful
steps
to
take
to
before,
like
we
next
to
a
run
through.
E
I
think,
as
far
as
your
question
regarding
the
assets
job,
I
would
probably
try
to,
I
would
suggest
something
different.
I
recall
during
last
week,
unless
you're
brought
up
the
fact
that
the
migrations
job
for
rollbacks
may
not
be
set
up
correctly.
E
I
think
we
should
investigate
that
instead
and
that's
just
because
that's
if
we
need
to
do
a
rollback
in
production,
that
would
be
something
that
we
want
to
make
sure
it's
actually
going
to
work
as
desired.
The
assets.
Job
is
just
something
that
adds
time,
which
I
think
is
not
as
bad
as
missing
a
job
that
we
may
need
to
have
inside
of
our
rollback
pipeline.
E
D
Production,
okay
for
production,
the
problem
is
that
it
will.
What
is
the
problem
from
production?
There
was.
There
was
an
issue
with
with
italy
that
basically,
but
this
was
fixed
this
week,
so
there
was
a
bug
in
the
pipeline,
so
you,
if
you
roll
it
back
for
the
production,
it
would
have
brought
back
gizzali
on
staging
as
well
just
but
yeah.
So.
E
D
D
A
Awesome
sounds
good,
so
then
the
only
final
piece
I
wanted
to
just
discuss
was
I'm
curious.
Like
would
we
have
been
able
to
like?
Would
it
have
been
risky
to
roll
back
today
whilst
we're
preparing
security.
B
I
think
it
would
depend
on
what
state
the
security
preparation
is
in
like
if
we
have
already
deployed
a
security
release,
the
staging
and
technically,
you
could
roll
that
back.
It
might
be
annoying
if
people
are
still
doing
testing
on
staging
photo
security
fixes.
Although
you
know
with
the
current
process,
they
should
have
done
that
in
the
merch
quest,
but.
B
It's
I
think,
in
the
in
the
context
of
a
security
release,
it
kind
of
depends
on
like
how
bad
the
incident
is
right.
If
this
is
something
like,
oh
the
the
the
button
in
the
ui
is
broken.
You
know,
users
are
annoyed
by
it
and
you
could
say
well,
you
know
that
is
annoying.
A
lot
of
people
are
affected
by
it.
However,
we
still
have
this
big
security
release.
That
is,
you
know.
All
things
considered
gonna
be
much
worse.
B
If
we
don't
get
that
out,
so
you
know,
let's
do
that
first
and
then
I
know
not
roll
back
or
something
else.
On
the
other
hand,
if
this
is
an
incident
like
we
have
to
roll
back
now
or
we're
going
to
go
down,
I
would
basically
say
screw
the
securities
it's
kind
of
a
balance
between
your
security
and
availability.
A
Yeah
for
sure,
so
it
wouldn't
it's
just
literally,
it
would
block
security,
there's
no
extra
complexity
to
a
rollback
or
risk
to
the
security
release.
D
I
was
thinking
that
so
I
agree
with
yorik
idea.
It
completely
makes
sense,
and
I
think
that
your
question
amy
was
more
about
rolling
back
staging,
so
the
impact
of
our
testing
in
the
in
the
in
the
roll
out
of
a
security
release.
So
what
I'm
thinking
here
is
that
well
we
will,
unless
it's
a
remote
code,
execution
type
of
issue.
We
wrote
back.
Staging
people
are
not
using
staging
it's
just
for
yeah
testing,
so
it's
fine.
D
It
will
just
I
mean
release.
Managers
are
just
doing
this
instead
of
taking
care
of
the
of
the
release,
but
that's
another
story,
the
thing,
but
even
in
case
as
yorick
mentioned,
that
we
have
we're
risking
to
go
down
and
there's
a
big
outrage
and
we
roll
back.
We
still
have
to
consider
that
the
details
of
the
security
fixes
are
not
public
in
in
this
stage
of
the
of
the
rollout,
because
we
are
do
early
mergers,
so
we
are
rolling
out
security
fixes
and
yeah
people
don't
know
about
them.
A
E
A
Nice
and
thank
you
once
again,
alessio
for
sorry
scarborough
for
doing
our
notes
and
coordinating
on
the
issues
and
stuff
like.
I
really
appreciate
that
one
thing
I
did
actually
last
week
I'll
link
them
in
here.
I
find
a
approach,
be
text
files
I
ran.
There
is
the
way
I
got
the
run
book
together.
Last
time
was,
I
ran.
The
video
through
otter,
you've
got
a
transcript
and
then
it's
quite
easy
to
just
jump
to
the
relevant
bits
of
the
video.
So
you
can
watch
so
I'll.
A
Do
that
the
same
this
time
and
add
the
link
in
so
it
makes
it
a
bit
easier
if
you're,
like
specifically
trying
to
get
the
steps
for
like
oh,
how
did
we
do
grabbing
that
package?
You
can
just
see
that
in
the
transcript
it's
like
four
minutes
57
and
you
can
just
leap
in
there.
So
that
saved
me
a
ton
of
time
as
well.
A
It's
called
otter
and
it
you
upload
a
recording.
I
think
we
were
trialling
it
for
a
period,
but
I
think
we
must
have
stopped
because
I
was
hoping
to
get
recording,
but
I
didn't
get
one,
but
I
have
an
account
myself
so
yeah,
it's
you
upload
an
audio
file
and
it
transcribes.
A
It's
not
perfect,
but
it's
like
enough
for
you
to
find
like
if
you're
looking
for,
like
how
it's
actually,
when
wonder,
is
actually
prompting
you
yuric
to
say
like
which
way
around
did
you
have
those
packaged
like
things
is
actually
it's
helpful
if
you
actually
explicitly
say
like?
Oh,
when
I
look
up
staging,
I
go
to
announcements
because
it's
easier
to
find
it
later
on.
So.
D
I
think
zoom
does
the
same,
because
I
was
downloading
a
recording
yesterday
and
there
is
a
transcript
vtt
file
with
times
who's
speaking
and
in
the
transcript.
I
don't.
A
Know
it's
coming
through
either.
Maybe
it
doesn't
do
all
of
them.
It
didn't
do
last
week's
or
it
didn't
capture
everything
usefully
but
yeah,
it's
quite
good
when
you
get
them
like
they're,
not
great
transcripts
but
they're
useful
to
if
you've
got
an
hour-long
video,
it's
a
great
way
to
jump
into
like
a
relevant
point
of
it.
So.
A
Awesome
so
let's
assume
we
will
be
able
to
do
another
like
a
another
round
next
week
we
can
review
that
as
we
get
closer,
but
who
would
like
to
go
up.
A
Next,
I'm
going
to
nominate
you
myra,
because,
as
a
release
manager,
I
think
it
will
actually
be
the
most
risky
thing
next
week
will
be
impact
on
the
release
so
to
give
you
full
control
over
whether
you
are
comfortable
doing
that
or
whether
you
think
you'll
have
the
full
context
of
like
deployments
and
any
kind
of
uncertainties
around
that.
So,
if
you're,
okay
with
that,
I
will
nominate
you
to
take
the
lead
next
week.
A
Okay,
stuff,
good
stuff,
it'll
be
fine
right
because
by
then
you're
going
to
have
the
perfect
run
book,
I'm
just
going
to
keep
promising
this
like
every
week
for
like
the
next
like
six
months,
fantastic
great
thanks,
much
very
much!
Everyone
I'll
see
you
all
later
bye.