►
From YouTube: 2021-04-29 GitLab.com k8s migration EMEA
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
C
Hey
andrew
thanks
for
joining
us
again,
today's
always
fun
day,
because
we
do
a
pack
demo
in
the
morning
and
then
emia
now
so
pressure's
really
on
andrew
I'm
expecting
like
good
progress
compared
to
what
we
saw
a
couple
of
hours
ago.
C
Awesome
so
was
there
anything
that
anna
wants
to
demo.
B
B
I
haven't
looked
into
the
status
of
where
we
are
with
the
service
discovery
situation,
so
this
could
just
be
a
lack
of
understanding
where
we
are
with
that
particular
stage,
but
I'm
a
little
worried
about
service
discovery
in
canary
right
now.
So
we
have
this
desire
to
try
to
get
432
events.
B
We
had
this
desire
to
get
canary
taking
traffic
right,
I'm
concerned
that
we're
not
getting
any
response
and
if
you
look
at
the
counts,
they're
the
same
for
every
minute
that
this
occurs
because
we
were
like
every
minute
and
a
half
it
looks
like,
and
that's
because
we
run
five
pods
and
this
check
happens
every
60
seconds.
B
So
this
is
happening
for
all
of
canary.
It's
not
sporadic
it's
happening
for
every
single
time.
We're
making
this
request
so
on.
My
to-do
list
today
was
to
try
to
figure
out
where
we
are
with
this
particular
bit,
because
otherwise
canary
will
always
be
using
the
primary
database
instead
of
any
of
the
secondaries.
B
C
He
has,
he
has
ideas
that
he
is
oh,
we
talked
about
this
morning.
He
has
an
approach
that
he
wants
to
get
out
into
production,
but
he
agreed
that
we
probably
shouldn't
put
traffic
on
canary
until
we
have
that
so
so
I
think
that
kind
of
fits
in
he
was
talking
about
andrew.
Can
you
summarize
this
better
than
I'm
about
to,
but
he
was
talking
about
the
complexity
of
getting
the
chance
changing
right
like
it's
an
upstream.
D
C
D
He
was
talking
about
using
and
again,
I
was
also
only
kind
of
following
kind
of
a
little
bit,
but
there's
a
mechanism
in
helm
where
you
can
patch
the
effectively
patch
the
jason
and
that
he
was
thinking
of
using
that
rather
than
getting
the
upstream
patch
in
for
a
change
to
do
with.
It
was
something
to
do
with
shutdown,
though
I
can't
remember
what
the
original.
A
A
That's
how
I
understood
it
yeah
once
we've
kind
of
proved
that
that
works
yeah,
but
right
now
we
can't
change
demon
sets
to
to
because
the
problem
he
identified
as
that
demon
sets
specifically
when
they
are
turned
down,
they
will
not
wait
for
first
stopping
traffic.
So
in
this
case
we
we
lose
requests,
and
this
is
causing
problems
still
and
there's
no
way
in
our
chart
to
fix
this
at
the
moment.
D
But
my
question
is:
how
does
that
tie
in
with
that?
With
that
dashboard,
you
showed
right
at
the
beginning
skybeck,
because
that
looks
like
it's:
it's
not
a
shutdown
problem
that
looks
like
it's
an
ongoing
all
the
time
problem,
not
not
like.
Oh
we're,
gonna
shut
down
this
node
now.
B
Sure
if
it's
the
same
yeah
yeah,
I
think
there's
something
else
at
play.
So
I
guess
what
I'll
do
later
today
is
look
at
where
graham
is
with
that
issue,
because
I
do
recall
him
making
a
change
in
our
staging
and
pre-environment.
But
I
don't
believe,
that's
been
pushed
to
production
and
I
wonder
if.
B
Production:
okay,.
B
Missing
in
the
canary
environment
that
hasn't
been
applied
because
I
think
ports
changed,
and
if
that
is
the
case,
then
maybe
I
just
need
to
validate
that.
We
don't
have
a
network
policy,
that's
blocking
wherever
these
requests
are
being
sent
to,
because
I
think,
instead
of
going
to
our
dns
port,
we're
going
to
port
8600,
so
it
maybe
I
just
need
to
validate
that.
We've
got
our
configurations
and
canary
then
I'll
I'll.
Do
that
I'll
certainly
be
working
on
this
today.
C
I
mean
we
can
use
this
time
if
you
want
to
go
back
like
you
know,
if
you
want
to
spend
the
next
30
minutes,
whilst
we're
all
here
working
out
what
the
status
is
and
what
next
steps
will
be,
then
that
would
be
a
good
use.
C
B
B
Okay,
so
this
is
what
I
thought
so
quickly
I'll
share
my
screen,
so
we
added
this
change
where
we
changed
the
end
point
that
we're
using
changing
it
to
a
headless
endpoint
and
we
also
change
the
port
number.
So
what
I
would
like
to
validate
next
is
whether
or
not
we
made
the
same
change
for,
because
this
is
targeting
oh
console
yeah.
This
is
targeting
sidekick
and
what's
this
woman
down
here
below
psql,
so
that's
our
generic
stuff
for
rails.
I
believe
that's
the
global
configuration.
B
C
B
B
E
B
D
E
B
B
B
B
B
Which
doesn't
work
on
port
53?
It
only
works
on
port
8600.
Does
anyone
off
the
top
of
their
head
recall
how
to
tell
dig
to
look
at
a
tcp
port,
a
specific
psp
port
instead
of
the
default
53.
G
E
A
G
D
If
you
go
in,
if
you
just
do
nslookup
with
no
arguments
and
do
it
that
way
and
then
set
port
busybuzzy
busybox
nslookup,
I
don't
know
if
it's
got
the
interactive
mode
just
enter
no
okay,
because
that's
how
I
would
do.
G
F
I
D
Just
try
go
to
that
ip
address
and
port.
Sorry,
the
host
name
and
port,
the
of
the
of
the
console
port
860.
D
G
Protocol,
so
you
have
to
which
is,
basically,
you
have
to
do
a
real
lookup
using
something
like
netcat
to
get
sorry
tcpdump
to
get
the
binary
requests
and
pipe
it
back
with
echo.
So
it
doesn't
look
like
something
it's
easy
to
do
right.
It's
not
a
text
based
protocol.
D
Can
can
I
ask
if
this
this
is
going
to
be,
the
demon
set
console
is
going
to
be
bound
on
the
node
as
well?
Isn't
it
like?
D
B
We
can
my
problem
with
that
method,
though,
is
we
have
network
policies
that
restrict
pod
traffic
to
other
pods,
and
if
we
just
hop
into
the
node,
we
kind
of
bypass
that.
D
B
A
The
error
message
about
getting
an
empty
list,
but
not.
G
Yeah,
we
could
definitely
be
a
network
policy
because
there
is
no
connection,
it's
just
udp
messaging,
so
I
was
thinking,
there's
also
the
console
api
that
should
expose
you
the
same
information,
but
I
I
guess
that
we
are
past
the
point
where
we
know
that
the
information
is
on
the
on
console
is
just
why
it's
not
coming
back
right.
That's
the
question
so
question
also.
A
A
A
I
think
is
giving
the
ports
a
list
of
things
to
query
right
and
I'm
not
sure
if
this
list
is
correct.
A
G
A
Is
not
correct,
we
are
trying
to
reach
name
servers
that
we
got
in
some
kind
of
list
and
we
can't
reach
them.
So
I
wonder
what
kind
of
name
servers
trade
are
we
trying
to
reach.
D
D
Actually,
I
suppose
it
doesn't
really
help
very
much.
But
sorry,
I'm
I'm
confusing
myself
here,
because.
B
B
J
G
D
D
So
so
one
just
on
that
point,
is
there
any
chance
that
we
are
connecting
dns
to
the
http
port,
because
if
you
look
at
the
stack
trace
from
the
sentry
error
that
java
posted
it's
like
a
protocol,
it's
like
the
protocol's
broken
is
the
is
why
we're
throwing
that
exception.
It's
like
this
is
not
the
answer
we
were
expecting,
and
that
might
happen
if
we
were
querying
like
an
http
port
with
a
dns
query,
for
example,
that
just
you
know
that
could
be
a
thing.
There's
8600.
D
G
B
E
D
G
B
D
D
D
F
B
F
B
G
A
console
agent
should
accept,
should
should
answer
api
requests
unless
it
is.
D
It
is
it
not
just
that
we
have
not
map
that
port.
G
G
G
G
B
B
D
And
this
sorry
sorry
if
this
is
a
red
herring,
but
this
is
the
console
console
demon
set
and
it's
got
a
whole
bunch
of
config
in
there,
and
I
just
happen
to
notice
that
there's
no
dns
on
this,
but
I
don't
know
if
that
automatically
comes
through
or
not.
Can
you
see
what
I'm
looking
at.
B
G
I
D
J
G
Oh
back
in
your
question:
if
you
wanted
to
see
data
related
to
services,
the
query
is
console
catalog
nodes
and
then
you
do
dash
service
equal
name
of
the
service
and
will
tell
you
ip
address
of
the
nodes
that
are
offering
that
service.
D
H
G
B
D
D
Yeah,
which
is
why
it
kind
of
and
remember
we
needed
a
curl
to
8600,
and
it
got
it
just
it
just
said:
no
response
from
server
that
could
I
guess
that
could
be
because
console
just
hung
up,
because
we
were
sending
the
wrong
thing,
but
it
could
also
be
that
that
whatever's
on
that
port
is
just
hanging
up.
You
know,
as
soon
as
you
send
it
a
request.
It
just
kind
of
kicks
you
off.
B
Alessio,
since
we
have
rails
available
to
us,
could.
C
G
There
was
something
coming
so
I
was
thinking
about
this
regarding
the
networking
part
right.
So
this
is
a
udp
connection,
so
we
are
this.
Are
we
shooting.
B
G
G
C
G
Yeah
but
tcp
is
a
handshake
based
right
so
before
you
get
third
package
before
you
can
actually
send
your
request.
So
does
it
mean
that
we
established
the
tcp
connection?
So
demon
is
working
but
is
counsel.
D
That
are
given,
if
you
so
yeah,
because
if
you
look
at
the
exception,
it's
like,
if
not
ants,
which
presumably
is
answer,
then
throw
that
exception
right.
So
it
got
back
an
empty.
It
didn't
get
an
exception,
connection
refused
or
connection
rejected.
It
got
back
no
answer
it.
Just
got,
kicked
off
right,
yeah.
D
D
Exception,
yeah
yeah,
it
just
was
like
see
ya.
Yes,
okay,.
D
D
G
D
B
G
G
Repeat
your
first
part
of
that
question.
So
I'm
curious
about
how
we
instruct
the
pods
to
use
this
special
name
space.
So
are
we
providing
this
as
a
system
name
space
within
the
pod
or
it's
the
application
that
has
a
special.
B
Cd
config,
it's
gonna,
be
inside
of
our
database
cat
database.yaml.
This
is
gonna,
have
a
secret
in
it
hang
on.
Let
me
stop
sharing
real.
J
D
That
looks
about
right,
so
I
didn't
know
what
those
oh,
I
didn't
have
those
variables
in
my
copy
paste
buffer.
So.
A
B
C
G
G
F
G
G
G
F
G
G
D
D
Okay,
sorry,
I
thought
that
that's
from
the
require
resolve.
D
Let
me
just
check
I
stole
that
from
service
discovery
from.
Let
me
try
if
I
can
find
this
as
well.
It's
it's
in
discovery.
D
D
D
D
D
Ip
address
result
from
sorry,
that's
probably
on
me
resolver.start,
that's
that's
me,
cupping
copying
and
pasting
the
wrong
version
of
that.
Try
this
one
and
again.
D
D
B
B
G
G
D
G
G
B
D
Is
it
possibly
that
the
warning
is
kind
of
rate
limited.
A
Or
did
you
get
over
the
package
size
for
dns
udp
and
breakfast
yeah.
B
F
D
B
D
D
D
D
Yeah
or
I
think
you
just
put
a
because
yeah
it's
we've
got
an
ip
address
in
that
in
that
format.
So
I
can.
F
B
I
D
I'll
say
that
again
with
my
microphone
on
just
say:
resolver.search
db,
replica,
comma,
a
and
leave
out
the
record
type
colon.
D
G
G
D
Do
you
want
to
try
different,
like?
Is
there
any
other
service
that
we
can
try
just
to
see
if
we
get
this.
D
G
The
console
agent
gives
us
the
list,
so
we
can
do
console
catalog
services.
G
B
And
what
was
that
command
that
we
ran
to
like?
I
don't
wanna
try
to
ping
some
that
has
empty
data
they
they
should
have.
G
B
G
D
B
B
So
I'm
not
seeing
any
messages,
there's
no
messages
when
I
do
a
request
and
it
fails.
G
Can
can
we
ask
the
full
name
for
dns
name
instead
of
just
the
short
version
should
be
something
like.
D
Ask
the
catalog
for
the
full
names
rather
than.
D
B
G
B
Server,
let
me
where's
where's
the
pause
recording
button.
B
G
G
G
D
G
Don't
want
it
to
have
the
default
name
domain
name
of
the
machine.
G
D
That's
me
again,
sorry
that
dns
resolver.
D
G
Net
dns
solver
yeah,
sorry.
D
G
B
D
And
you
take
out
the
record
type.
Sorry
I
totally
cocked
up
the
it
was
this.
B
B
We
configured
the
port
as
part
of
the
the
whatever
the
a
variable
technically
is.
The
resolver
is
looking
at
port
8600,
no.
G
B
Well,
if
we're
looking
for
srb
records,
help
me
understand
this
output,
because
we're
outputting
this
same
information
that
we're
creating
as
an
srv
response,
but
like
the
only
time
I
see
hosts
is
a
record.
So
I
wonder
if
I'm
supposed
to
see
srv
records
in
this
section
as
well
or
if
I'm
supposed
to
see
petronio
2
in
this
big
section
up
here.
B
B
K
B
D
H
D
Went
through
all
the
answers-
and
it
was
like-
I
don't
understand
how
to
deal
with
this
or
something
it
was
like.
I
didn't
get
an
answer.
Could.
G
B
Well,
we
should
be
able
to
prove
that
right.
Isn't
there
a
fancy
irb
for
hey
give
me
your
configuration
for
service
discovery,
sure
somewhere.
J
So
can
I
exactly
this
problem
is
intermittent
right.
Is
it
on
one
server
per
system.
B
H
K
Sanity
check
we're
sure
that
it's
not
the
connection
to
the
replicas.
It's
the
dns
lookup,
that's
failing
and
we're
certain
it's
not
the
network
policy
problem
that
we've
seen
before
yeah.
B
So
we
just
validated
doing
this,
that
we
are
we're
not
blocking
ourselves
via
network
policy.
What
was
your
first
question.
K
The
first
question
was
related
to
ports.
This
doesn't
show
you
the
ports,
but
the
replicas
have
different
and.
K
B
Ports
now
this
is
where
we're
sending
the
request
to
oh.
B
K
J
G
I
think
that
I
know
how
to
see
the
configuration:
let's
do
it.
Maybe
we
should
stop
recording
as
well,
because
I
think
it
will
print
out
the
maybe
can
you
do
the.