►
From YouTube: 2020-04-07: High availability Gitaly 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).
C
A
A
D
B
C
A
E
A
Told
no
we're
just
talking
about
this
flag
here,
like
what
happens
if
you
were
to
set
not
set
get
early.
True,
a
primary
true
on
any
of
the
get
lean
eyes
and
just
say
like
here:
are
the
nodes
like
I?
Don't
care
which
one's
primary?
What
would
happen?
I,
pretty
sure
people
won't
start
up
I
think
we
have
validation,
I'm,
config,
yeah,
I!
Think.
A
E
Don't
think
it
should
be
necessary,
yeah
I
mean
I,
think
in
the
figure
we
could
just
remove
that
and
if
you
especially
will
be
default
to
the
right.
You
know.
If
we
do
the
failover
election
process,
then
it
can.
We
can
just
write
ignore
that
whole
just
let
the
discovery
work.
So
that's
a
good
point.
Do.
A
E
D
A
C
A
A
A
A
B
A
B
D
A
A
E
A
E
E
A
E
C
E
A
C
C
E
Like
when
I
configure
demo
last
way,
I
did
the
internal
eyepiece
just
cuz,
I
was
parent,
I've
been
paranoid
and
I,
don't
like
you
know,
people
potentially
accessing
the
house
from
the
outside
I
will
get
an
example.
I
was
doing
a
test.
The
other
day
and
I
was
trying
to
get
I
was
somehow
I
configured
my
load
balancer,
so
that
people
could
access
Redis
from
the
outside.
So
someone
actually
start
messing
with
my
Redis
instance,
and
like
did
some
weird,
you
can
do
some
random
attack
from
Redis
from
external.
C
E
D
D
A
A
D
A
D
B
A
B
E
A
A
B
E
E
D
B
A
D
E
A
A
A
D
C
C
D
E
A
B
A
E
D
E
E
D
D
A
So
I
think
now
we
should
pull
up
Griffin.
Is
that
right.
D
A
C
E
C
A
D
C
That
top
left
one
is
the
one
I
need
feedback
on
cuz
I,
don't
know
if
this
is
confusing
or
if
this
is
better
than
we
had
before
it
could
be
before
we
had.
You
know
lines
going
up
and
down
now.
It
just
shows
the
most
recent
entry
for
a
virtual
storage
where
the
counter
is
greater
than
zero.
So
if
you
look
at
the
actual
query,
the
query
is
really
simple:
it's
just
greater
than
zero
because
we
have
all
these
different
gauge
values
coming
in
from
different
giddily
storages.
C
E
C
E
C
E
Alright,
so
it's
there
yeah
in
the
cuckoo
console
like
this
is
my
demo
last
week,
I
just
because
we
have
control
over
our
logging
on
postcards.
I
just
enabled
full
statement.
Logging,
so
I
could
see
every
okay
kokkuri,
but
obviously
I
don't
know.
We
can
do
that
here.
Maybe
maybe
that's
config.
It's
configurable
to
do
that.
E
B
E
D
D
B
B
D
D
D
E
Just
this,
this.
E
B
E
Yeah
I
think
you
can
complete,
go
in
the
repository
settings,
just
go
to
the
project
and
I
think
it's
under
repository
or
is
in
general
I
forget
one
or
the
other
might
be
in
general
notice.
I
think
it's
in
general
and
there's
visibility,
features
I
think
is
underneath
yeah
large
file,
storage,
yeah,
I,
think
I
know
doing
something
like
that
and
I
was
trying
I.
D
Copy,
how
do
I
set
this?
What
why
they
left
us
back?
D
B
B
C
D
A
E
B
E
E
E
D
E
D
E
A
D
A
E
You
sorta.
A
E
A
A
E
A
C
B
E
A
A
A
C
Nice
I
don't
know
if
there's
a
way
to
do
this
with
Prometheus
or
Griffin.
Oh,
it
would
be
really
cool
if
that
table
instead
of
an
absolute
time
they
had
last
seen
or
last
update
on
there.
So
you
could
see.
Oh
I
haven't
seen
a
change
in
this
storage
like
a
day
ago,
her
five
minutes
ago
or
ten
seconds
ago.
Something
like
that.
But
I
don't
know.
If
there's
a
way
to
do
that,
I
don't
have
that
Griffon
a
lizard,
your
wizardry
get
that
done.
C
E
E
C
A
E
A
E
E
A
C
Think
the
the
the
the
fact
that
there's
multiple
rows
means
there's
multiple
locks.
I
think
the
entire
table
is
acting
as
a
law
from
what
I
understand,
because
each
one
of
those
is
a
different
job,
so
I
think
you're
only
permitted
to
take
one
job
at
a
time.
It
has
to
be
some
kind
of
lock
state
right.
Do
we
see
anything
here
that
indicates
the
entire
thing
is
locked
well.
E
E
E
Yeah
I
think
I
think
that's
the
idea
and
then
the
replication
key
lock
is
about
locking
on
the
repository
right.
So
yes,
as
each
of
them
finishes,
and
it
goes
and
processes
that
can
you
kendo
stay
where
you
just
you're
just
sitting
on
locks
like
your
whole.
Queue
is
just
filled
with
locked
repositories
that
take
a
long
time
to
complete.
D
A
E
A
E
C
E
That's
a
good
point.
We
should
actually
do
query
the
database
occasionally
for
these
statistics
right,
like
I,
want
to
know
how
many
have
are
in
the
pending
state,
for
example,
right,
like
whatever
state
we're,
calling
it
ready,
I
think
it
is
yeah
and
I
put
a
question
on
indexes
because
I
don't
see
indexes
on
the
table.
So
if
we
want
to
count
the
number
of
ready
over
time,
this
tables
can
get
really
big
and.
A
The
other
thing
I
was
thinking
is,
it
would
be
really
nice
to
say,
like
its
throughput,
like
the
number
of
upper
like
replication
jobs
being
processed
like
per
second.
A
So
you
can
just
see
that
things
are
running
like,
as
you
can
see,
that,
like
there
are
jobs
in
flight,
but
I
mean
there's
only
one
at
a
time,
but
like
it's
not
actually
telling
you
the
right
at
which
they
don't
have
to.
We
can
actually
get
that
from
the
because
we're
you
have
a
histogram
of
the
latency,
so
we
have
it
with
a
count.
A
Oh
cool
yeah,
okay,
yeah
and
it'd,
be
really
cool
to
say
that
Prefect,
because
I
think
the
other
thing
we
like
Stan
once
yours
comes
is
merge
like
we
want
to
be
doing
this
with
three
prefix.
So
we
need
to
like
understand
the
replication
to
behavior,
not
just
across,
like
one
note
but
like
all
three
yeah.