►
From YouTube: Geo Group Conversation
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
B
B
So
I
had
two
questions
on
that.
Like
first
of
all,
my
assumption
there
was
that
geo
is
currently
designed
around
the
idea
that
you
trust
all
the
secondary.
So
though
I
haven't
really
introspected
on
what
I
mean
by
trust
there,
and,
secondly,
do
we
have
customers,
we
don't
need
to
name
them
obviously,
but
do
we
have
people
already
who
are
using
geo
who
have
nodes
not
as
trusted
as
other
nodes,
I
think.
A
Trust
might
be
a
I
think.
I
may
have
not
chosen
the
best
word
there
when
I
say
a
trust
issue
with
the
owner
of
the
road.
I
think
it
is
more
around
trusting
the
people
that
have
administrator
access
of
that
node
not
to
go
and
access
the
information
in
the
database
that
they
don't
need
to
know
to
do
their
job.
So
for
the
example
where
they
may
have
a
consultancy
with
administrator
access
for
whatever
reason
down
to
the
database
level
and
because
geo
replicates
absolutely
everything
in
the
database.
A
A
Yes,
so
I
think
maybe
Trust
is
yes,
it's
not
the
correct
word
there,
but
we
don't
actually
have
any
situations
at
the
moment
where
we
have
deployments
where
they're,
not
trust
it.
If
that
makes
sense,
I
think
it's
much
more
important
to
be
thinking
of
it
in
terms
of
the
legal
ramifications
of
having
data
where
it's
not
supposed
to
be.
C
C
On
the
things
that
didn't
go
well
and
how
we
should
do
it
differently
next
time,
but
we
don't
focus
things
to
go
well
and
consistently
across
the
board.
For.
A
A
A
D
Yes,
Rachel
I
have
a
question
for
you,
so
we
recently
went
through
a
horse
security
assessment
of
Geo
with
a
third-party
firm
and
I
would
like
to
get
some
of
your
thoughts
as
a
stakeholder
in
that
assessment
about
whether
you
think
that
process
worked
well
and
if
not,
what
could
be
improved
next
time.
I
think.
A
So,
while
that
was
already
available
on
the
PDF
I
appreciated
that
the
security
team
then
made
the
issues
for
us,
it
just
was
one
less
less
step
that
we
needed
to
do
and
just
verified
that
we
were
going
to
be
working
on
the
correct
things,
because
ultimately,
we're
doing
this
to
make
the
the
product
more
secure.
But
it's
important
to
work
well
with
the
security
team
to
get
that
in
order
to
get
through
their
process.
A
There
and
I
mean
I'm
I
totally
understand
people
go
on
vacation,
so
I'm
not
saying
I'm,
not
asking
for
I'm
trying
to
know
quite
how
to
say,
but
yeah
I'm
not
putting
pressure
on
people
that
take
vacation
at
all.
But
we
do
have
things
that
are
now
waiting
in
review
and
we
just
want
to
make
sure
that
those
get
released
as
soon
as
possible.
Okay,
yeah.
D
E
A
Any
issue
that
comes
up
now
that
does
have
some
security
implication.
Israel
is
required
to
go
through
a
security
review
before
the
merge
request.
That's
been
a
very
good
addition
from
the
security
team
into
the
into
the
code
review
cycle
and
into
the
code
review
process.
So
there
are
certain
issues
outside
of
this
review
that
the
Geo
team
will
have
to
have
reviewed
and
that's
just
part
of
the
standard
process
going
forward.
A
A
That's
right:
these
are
people
in
Kathy's
team
that
will
perform
their
final
reviewed
from
a
security.
You
just
work
with
a
security
focus
of
it.
So
while
we
try
to
catch
everything
inside
of
the
team,
it's
nice
to
have
people
that
will
focus
on
the
security
side
of
things.
Just
looking
for
anything
that
we
may
have
missed
is.
A
Would
be
wonderful,
I
think
there
was.
There
was
talk
of
having
some
sort
of
stable
counterpart,
the
same
way
that
there's
a
stable
counterpart
of
the
front
end
UX
product
management,
etc,
but
I'm
not
sure
how
far
discussions
are
with
it.
If
we
could
have
a
stable
counterpart
for
geo
I'd
be
very
happy.
A
F
I
can
answer
that
sidekick
by
default
use
Redis
as
that
they
store
and
whenever
you
ask
for
a
job
from
Redis,
it
deletes
from
the
queue.
So
if
something
goes
wrong
between
the
job
is
fetch
from
the
queue
and
process
it,
there
is
a
chance
that
it
you
can
lost
a
job.
So
with
the
reliable
fetch,
we
use
a
different
operation
from
ready.
So
basically,
what
we
do
is
remove
the
job
from
one
queue
to
a
temporary
queue
and
that's
an
atomic
operation,
and
we
keep
that
there
until
the
job
is
being
processed.
E
A
What
is
what
we
were
trying
to
depict
here
is
the
fact
that,
on
a
secondary,
if
you
perform
a
get
push
that
the
push
is,
is
it
is
accepted
by
the
push
proxy
and
that
is
then
proxied
over
to
the
primary
node?
That
is
what
we
were
trying
to
depict
in
the
bottom
in
in
the
bottom
section
there.
So
when
so
on
a
secondary
the
push,
while
it
is
accepted
by
the
primary
it
has
been
proxy
there
through
the
secondary
feeding.
E
That
makes
total
sense
and
state
that
you
that
you
show
that
and
I
think
that
makes
it
much
more
understandable.
I
think
you
might
want
to
consider
yeah
get
full
and
push
as
this
same
walk
on
the
left
and
it's
different
blocks
on
the
right
and
that
might
be
confusing
for
people.
Then
I
do
you
have
a
dotted
line
weather
where
there's
no
error
or
you
have
the
arrows
that
are
accurate,
but
right
now
I
think
there's
arrowheads
that
that
are
not.
E
A
G
A
Sure
so,
some
of
the
things
that
we're
exploring
are,
as
you
would
have
seen
on
the
on
one
of
the
slides.
We
do
talk
a
little
bit
about
service
discovery
and
having
an
easy
way
to
set
up
a
complete
cluster.
So
that's
one
of
the
things
that
we're
looking
at
is:
how
can
we
actually
make
it
easy
to
install
a
secondary
node
in
a
clustered
format
without
having
to
do
a
lot
of
configuration
changes?
A
One
of
the
other
things
that
we
are
exploring
is
that
some
customers
have
said
it's
difficult
to
promote
a
node
from
a
secondary
state
to
a
primary
States
in
a
very
easy
way
and
in
the
disaster
situation.
The
last
thing
you
want
to
do
is
be
reading
pages
and
pages
of
documentation
and
hoping
that
you're
making
all
of
the
right
changes.
Ideally,
you
want
to
push
a
button
and
say
this
is
primary,
so
we
need
to
figure
out
from
the
background.
How
do
you?
How
would
you
do
that?
A
There's
a
number
of
different
of
different
aspects
there,
but
ultimately,
we've
had
feedback
that
it's
slightly
cumbersome
to
install
cluster
and
it's
not
completely
straightforward.
So
we
want
to
trust,
try
and
make
all
of
those
processes
easier
for
our
customers
and
for
for
ourselves
as
well,
since
we'll
be
running
that
on
kake.com
soon
now,.
G
G
A
I
think,
if
you're
using
the
omnibus
installers,
really
it's
pretty
straightforward,
I
think
as
soon
as
you
start,
making
small
changes
to
what
is
or
was
it,
what
is
provided
by
default?
It
becomes
more
and
more
complicated
and
I.
Think
in
some
in
some
scenarios
you
can
get.
You
can
be
following
I
think
it
was
either
four
or
five
pages
of
instructions
to
make
the
changes
that
you
need
and
that's
probably
more
instructions
that
we
want
people
to
be
following
to
install
something
like
this
in
in
in
not
so
recent
history.
A
There
was
an
issue
about
having
one
command
to
say
that
geo
per
server,
which
is
ideal,
but
that
specific
issue
I,
think
we've
moved
slightly
past,
and
so
what
we
do
want
to
focus
on
is
just
having
a
look
at
this.
This
idea
of
service
discovery
and
how
we
could
potentially
utilize
that,
but,
as
I
said
on
the
slide,
one
of
the
challenges
we
have
is
figuring
out
how
to
make
this
work
for
all
customers,
not
just
for
a
small
subset
of
them.
H
I
was
just
looking
at
the
slide
on
the
Geo
and
a
bigger
nutshell
and
look
there's
noticing
the
object.
Storage
transfers
I
was
just
curious.
Are
we
transferring
just
the
reps?
They
were
like
the
contents
of
the
buckets
I.
Could
I
could
sort
of
see
two
use
cases
where
either
one
would
be
beneficial
like
if
you're
an
s3
or
GCS?
Maybe
you
only
want
to
copy
like
the
ref
septic
storage
like
like,
where
they
are,
since
it's
already
like
geographically
redundant
and
super?
H
You
know
super
durable,
but
if
you're
like
on-premise
and
using
like
like
an
Atmos
appliance
or
something
like
that,
where
it's
like
in
a
datacenter,
you
might
actually
want
to
copy
all
the
contents
of
object.
Storage.
Long
I
was
kind
of
curious
with
what
what
actually
happens
in
that
transfer
link
as.
A
F
Maybe
so
forget
to
get
operations
we
just
use
get
and
that
works
well,
but
for
everything
else
like
mud,
blobs
and
I,
don't
know
page
assets,
etc.
So
there
is
a
process
that
downloads,
from
the
secondary
point
into
the
primary
and
just
doing
regular
HTTP
downloads.
There
is
some
education
in
between,
but
that's
how
we
replicate
that
they're.
F
E
E
A
At
the
time
that
that
comment
was
written
that
was
a
week
ago
and
then
at
the
in
the
middle
of
last
week,
another
another
idea
came
up
of
leaving
it
in
making
slight
changes
to
it.
I
haven't
gone
into
the
details
of
that
of
the
new
of
the
new
proposal,
but
I
do
want
to
I,
do
want
to
say
on
the
call
that
I
was
very
happy
that
Camille
also
put
together
a
proof
of
concept
of
the
new
proposal.
A
I
I
can
jump
in
a
little
bit
on
this.
We
have
been
kind
of
kicking
around
a
number
of
different
ideas
for
how
to
approach
this.
We
did
talk
about
rolling
in
into
rails
and
then
the
the
latest
proposal
to
keep
it
and
go
and
get
it
from
there.
It
feels
like
a
more
iterative
approach
and
so
I
think
we're
going
to
pursue
that
a
little
bit
more
and
get
a
little
more
details
on
that,
but
it's
still
we're
still
kind
of
open
in
the
discussion
to
how
we're
going
to
approaches
Thanks.
E
E
If
the
events
on
Redis
are
lost,
thank
you
for
not
we're
not
doing
this
at
second
to
second
backup,
etc.
Everyone
gets
locked
out
it's
father,
some
some
caching
is
lost,
but
it's
not
essential
and
I
felt.
The
design
from
Geo
was
so
that
in
the
end,
the
database
is
a
single
source
of
proof.
You
can
just
regenerate
all
those
psychic
events
to
make
sure
you
play
him
back,
but
the
part
I
quote
here
kind
of
makes
me
feel
like
that's
no
longer
the
case,
so
it's
very
essential.
E
E
There's
a
lot
of
talk
about
geofencing
things
and
making
sure
data
only
lives
in
one
location.
That's
not
really
what
we
do
with
geo
geo.
We
have
a
central
place
where
everything
is
and
then
we
can
selectively.
Sync
at
some
point,
which
degree
do
you
think,
there's
a
future
where
there's
stuff
that
just
there's
some
stuff
that
lives
in
location
only
in
location
in
some
stuff
that
only
this
and
location
B
I,
don't
see
that
with
the
current
GeoDesign.
But
I
was
wondering
whether
people
have
ideas
well,.
E
A
A
Think
there
are
some
customers
that
are
interested
in
primarily
in
the
in
the
git
repository,
but
also
the
issue
information
connected
to
why
the
code
changes
were
made,
and
there
is
one
particular
customer
who's
also
only
interested
in
having
the
user
information,
so
they
get
lab
users
only
existing
on
specific
node.
So
a
user
on
node
a
wouldn't,
have
wouldn't
have
a
user
of
profile
on
node
B,
for
example,
yeah.
E
I
think,
like
the
technical
changes
we
have
to
do
to
do,
the
database
split
up.
The
database
are
like
an
order
of
magnitude
more
than
what
we
have
to
do
to
split
up
the
git
repos
I
I,
wonder
what
the
relative
demand
is
there
and
is
there
certain
issue
were
recollecting.
You
should
feedback
about
this
I'm.
A
Collecting
user
feedback
in
a
separate
doc
I
need
to
figure
out
how
to
take
that
documents
and
make
it
into
an
issue
without
exposing
user
information,
which
is
an
action
that
I
do
have
for
that
at
the
moment.
But
it
is
interesting
what
you
say
about
relative
need
for
the
different
features,
because.