►
From YouTube: 2021-04-08 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
Hey,
I
guess
I
can
get
it
started,
because
I
have
the
only
demo
item
here
and
it
just
quickly
came
up
because
we
just
finished
this
a
few
minutes
ago,
thanks
to
graham's
work
and
then
we're
also
fixing
a
little
detail
again.
To
give
you
background
on
what
we
are
working
on
for
the
registry
work
and
pre
that
we
need
to
do
for
migrating
metadata
over
to
a
real
database
instead
of
having
it
in
a
blob.
A
We
also
wanted
to
run
qa
tests
and
pre,
but
we
didn't
have
runners
there,
and
so
we
needed
to
set
up
runners
for
the
pre-environment
and
graham
worked
on
that,
and
he
did
this
by
adding
a
new
helm
release
which
is
setting
up
a
runner
in
kubernetes,
just
by
setting
runners
enabled
in
a
value
for
this
environment,
and
we
just
made
this
work
today
and
I
just
wanted
to
show
you
that
a
little
bit
I
will
share
my
screen.
A
So
this
is
the
r
that
I
linked
to
and
if
you
look
at
this,
the
main
thing
is
that
you
need
to
set
this
okay,
parana
and
able
to,
and
here
what
graham
did
he
added
a
new
release
called
gitlab,
which
is
also
running
in
its
own
namespace
by
default,
so
that
you
don't
mix
it
up
with
the
github
namespace.
A
But
you
can
change
this
if
you
want
to,
and
it's
just
adding
the
runner
chart
and
setting
up
a
secret
and
that's
basically
it
to
enable
it-
and
we
did
this-
and
I
can
just
show
you
the
quick
result,
which
is,
I
think
here
going
to
pre
after
this
was
getting
deployed.
We
needed
to
do
nothing
else
and
already
see
the
runner
being
added
here
and
being
available.
A
A
I
didn't
test
it
yet
because
it
just
you
know,
got
working
a
few
minutes
ago,
so
we
should
see
if
you
can
run
something
on
it.
Some
special
things
to
mention
about
this
is
that
this
runner
is
running
in
privileged
mode
because
at
least
on
three,
we
want
to
be
able
to
run
docker
and
docker,
but
we
also
set
up
a
row
based
access
control
and
we
disabled
using
privilege,
as
let's
say,
the
privilege
to
run
set
user
id
binaries.
A
So
if
that
blocks
us,
we
need
to
enable
this,
maybe,
but
let's
try
if
we
can
run
with
that
with
provided,
let's
just
set
as
low
as
we
can
on
first
test
of
how
far
we
can
get
with
that
and
yeah.
I
think
this
is
nice
and
cool.
B
I
have
a
question
which
is
we
added
to
new
runners.
Sorry
give
me
one.
Second,
my
cat
is
trying
to
get
into
a
box
hey.
B
I
don't
want
him
there,
so
we
added
a
new
chart
to
our.
A
Official
runner
chart
right,
but.
A
I
think
from
what
I
understand
is
that
you
just
need
to
go
to
the
environment
that
you
want
to
enable
runners
in
and
set
the
github
runner
enable
true
setting,
and
then
I
need
to
make
sure
I
think
the
what
we.
What
graham
did
is
that
he
is
getting
the
secret
needed
to
access
the
runner
out
of
our
chef
fold,
so
that
should
generically
work
now
for
each
environment,
where
we
have
the
setup.
That's
how
I
understand
it.
A
Because
I
think
that's
should
give
us
maybe
a
possibility
to
do
at
numbers.
I
really
need
to
look
into
the
details.
This.
B
The
same
instance,
but
they
need
isolation
when
it
comes
to
scaling
up
and
down
configuration
and
so
on,
so
the
manager
can
still
be
the
same
because
a
single
runner
manager
can
have
multiple
runners
defined
naming
in
gitlab.
Runner
is
really
terrible.
I
know,
but
the
I
don't
remember.
B
If
our
charts
support
the
ability
to
say,
provide
a
yaml
configuration,
let's
say,
basically,
you
say
just
just
run
the
spot,
but
instead
of
generating
the
default
configuration,
which
is
some,
I
think
what
we
are
running
here
so
like
please
run
this
executor
configuration
in
that
case.
In
that
case,
you
should
be
able
to
define
more
than
one
with
different
settings,
and
so
you
can
have
the
share
it
or
yeah,
basically
duplicate
what
we
have
today.
C
B
C
B
The
point
I
was
I
was
making
with
this
sorry:
I
shouldn't
have
used
shared
rather
manager.
I
I
was
wanting
to
say
something
like
shared
run
as
manager,
meaning
meaning
we
have
four
of
those.
I
think
four
different
runner
managers
that
are
attached
to
different
types
of
project
on
multiple
instances.
B
So,
in
order
to
get
out
of
that
business
of
having
to
beg
someone
to
add
new
managers,
it
would
be
interesting
to
see
whether
we
can
just
create
a
new
manager
that
will
run
these
workloads
and
attach
it
to
the
same
instance.
So,
for
example,
for
staging
we
would
be
able
to
do.
You
know
like
create
or
add
the
runners
to
the
bombardier
tests
or
qa
tests
or
right,
like
all
of
these
different
projects
that
have
their
own
requirements
and
they
now
all
of
them
use
one
shared
runner
manager
right.
B
So
it
would
be
ideal
if
we
can
create
additional
one
of
these
using
the
same
cluster
but
isolation
on
that
side.
B
So
I
want
to
point
out
something
which
I
think
is
important
here.
That
is
that
it's
we
cannot
expect
to
just
replace
runners.
It's
a
general
available
runners
for
customer
with
kubernetes
one
because
they
don't
behave
exactly
the
same,
so
it
would
be
just
a
hell
for
support
because
they
will
just
be
inundated
by
tickets,
that
their
ci
bills
are
no
longer
working.
B
I'm
not
even
suggesting
that
this
is
nothing
to
do
with
customer
facing
this
is
all
to
do
with
internal
facing
testing
that
we
relatively
often
get
the
request
for.
Can
I
get
in?
Can
I
add
the
runner
for
this?
Can
I
can
I
you
know
test
somehow
this
staging
so,
ideally,
it
would
be
as
easy
as
adding
another
configuration
option
or
something
to
these
charts
and
then
just
running
it
off
and
not
having
to
build
completely
new
infrastructure
every
time
someone
wants
to
test
something.
B
But
why
do
you
want
to
have
more
than
one
in
that
case,
because
this
can
scale
up
to
yeah
how
many
resources
you
have
in
the
cluster?
Basically,
so
why
do
you
want
to
have
more
than
one
because
looking
at
how
things
actually
get
clogged
up
on
current
managers?
So,
for
example,
someone
fires
off
tens,
hundreds
of
qa
tests.
B
It
backs
up
everything
else
on
that
on
that
box.
For
for,
for.
C
B
C
Also
with
like
with
qa
with
qa,
specifically,
I
think
we
switched
to
using
shared
runner
manager
for
qa
on
ops,
because
that
spins
up
you
know
every
single
job
gets
its
own
virtual
machine
and
we
didn't
have
the
same
type
of
congestion
issues.
C
We
we
like,
we
flirted
with
the
idea
of
using
the
kubernetes
runner
a
while
back
and
we
ended
up
kind
of
abandoning
it
because
it
just
wasn't
as
reliable
as
using
the
shared
printer
manager.
Maybe
maybe
we
should
look
at
it
again.
I
don't
know,
but
we
currently
use
the
shared
runner
managers
on
ops,
staging
and
production.
B
Right
and
I'm
I'm
not
even
saying
that
we
need
to
replace
that.
What
I'm
also
saying
is
that
you
also
have
those
synthetic
loads
tests
that
you
have
in
your
projects.
Right
and
all
of
those
are
using
shared
runner
managers
and
they
might
not
need
to.
C
Yeah
I
see
yeah
I
mean
like
I've
been.
I
think
we've
been
a
lot
happier
since
going
to
the
shared
runner
manager.
Just
because,
like
you
know,
it
scales
a
lot
better
than
anything
else
like
definitely
scales
better
than
using
a
single
vm
like
we
do
for,
like
the
release,
runner
and
the
chat,
ops
runner.
C
We
also-
I
don't
know-
maybe
maybe
it's
worth
taking
a
look
at
kubernetes
again
to
see
if
it
can
meet
any
of
our
needs,
because
it
would
be
nice
for
even
for
like
release
tools
and
shut
ops,
it
would
be
nice
to
have
more
dynamic
scaling
right
now.
We
just
have
these
beefy
vms,
with
like
lots
and
lots
of
cores
mostly
so
that's
idle,
probably.
B
And
I
know
for
a
fact
that
the
kubernete,
the
the
well
actually
that's
another
question.
This
is
deployed
through
the
helm
chart
into
kubernetes,
but
does
this
also
mean
that
we
configure
the
kubernetes
executor?
C
C
B
Is
is
is
very
different
and
the
the
real
I
mean
the
isolation
is
a
problem
for
us
because
we
host
the
thing
for
others,
but
the
other.
The
other
real
problem
is
that
the
service
pattern,
so
how
your
ci
job
interacts
with
the
service
that
you
define
to
github
ciamo
is
completely
different
because
they
are,
they
are
all
running
on
localhost.
B
My
point
here
is:
we
just
want
to
get
out
of
the
business
of
how
do
I
add
a
runner
for
my
testing
and
the
runner
on
staging
is
overloaded,
or
how
do
I
make
sure
that
I
get
my
tests
on
pre
running
and
so
on,
like
if
it's
easy
to
get
this
done
through
this
setup,
then
it
theoretically
gives
us
much
more
flexibility
for
non-produ
environments
for
people
who
want
to
do
testing,
including
ourselves,
right,
like
we
have
a
bunch
of
different
migrations
in
our
queue
and
for
if
we
for
all
of
those
need
to
like
rely
on
just
one
runner
or
set
up
every
single
time.
B
Investment
and
yeah
andrew
thanks
thanks
for
that
link.
Actually
that's
that's
also
a
good
point
right,
like
he
andrew
posted
about
runner
for
tamland.
D
Yes,
because
sure
I
mean
it,
it
sort
of
ties
in
with
this,
and
I
didn't
create
it,
knowing
that
this
was
happening,
but
it
kind
of
matches
up
a
little
bit,
which
is
that
you
know
timeline,
sort
of
started
off
as
a
little
bit
of
a
toy
and
now
we're
using
it
everywhere
and
it's
becoming
part
of
processes.
And
it's
actually
running
on
the
ci
runners.
And
it
can
run
for
like
two
hours.
D
Sorry
on
the
chat,
tops
runners
and
that's
not
actually
a
very
good
thing
and
we
need
its
own
runner
and
it
actually
needs
to
be
a
bigger
machine
because
it
does
a
lot
of
number
crunching.
And
if
we
gave
it
a
faster
machine,
it
would
run
a
lot
quicker.
And
so
it's
kind
of
like
a
specialized
runner.
And
so
it
might
be
a
good
place
to
have
a
specific
runner.
And
if
you
could
run
it
in
kubernetes,
all
the
better.
D
B
So
we
had,
we
had
a
customer,
a
couple
of
customers,
I
think
ticketmasters
and
others
a
lot
of
they
were
using
a
lot,
the
the
kubernetes
here
and
they
were
contributing
a
lot
a
couple
of
years
ago
and
they
were
providing
a
lot
of
kind
of
hooks
where
you
can
define
environment
variables
in
your
job
and
this
kind
of
interacted
with
the
process
of
scheduling
the
job.
I
do
remember
something
for
user
access
for
limiting
the
the
account
the
I
don't
remember,
the
name
in
kubernetes.
Basically,
there's
a
lot.
B
There
was
a
lot
of
things
for
scheduling
and
things
like
that.
So
maybe
there
is
also
something
that
allows
us
to
put
labels
or
something
like
that
and
then
we'll.
Maybe
we
schedule
where
we
want,
but
I'm
not
sure.
D
B
Cool
thanks
thanks
for
entertaining
my
thought
here,
henry
it's
generated
a
good
discussion.
I
think
so,
thanks
for
doing
this.
B
Cool,
I
mean,
don't
I
don't
think
we
have
anything
else
given
that
we
had
a
very
short
week
so
and
skype
is
not
around
as
well
so
yeah.
A
I
think
it
was
the
last
state
I
think,
for
the
api
migration
is
that
we
have
an
mr
at
least
for
canary
now,
but
I
think
we
need
to
still
review
it
and
then
see
how
we
get
forward
with
that
one.
But
I
can't
speak
too
much
to
it.
I
just
briefly
looked
at
it
and
for
people
that
have
skype
would
be
able
to
give
insight
into
that
data.
Maybe.