►
Description
OpenShift Commons Gathering at Kubecon/EU 2021
May 4th, 2021
From OKD to OpenShift in 3 Years
Speaker: Josef Meier (Rohde and Schwarz)
https://commons.openshift.org/gatherings/OpenShift_Commons_Gathering_at_Kubecon_EU.html
For more information visit: https://commons.openshift.org
Additional Resources:
https://okd.io
https://groups.google.com/forum/#!forum/okd-wg
A
A
So
we
created,
we
had
to
create
a
cluster
on-premises
for
our
developers,
so
they
can
access
the
source
code
and
do
builds
or
their
artifacts,
and
the
second
cluster
should
be
in
the
public
cloud.
So
our
customers
can
access
them
because
we
don't
serve
our
software
from
our
on-premises
cluster
to
the
internet.
We
have
separate
classes
for
that.
A
We
had
a
few
requirements
for
that,
at
least.
There
were
three
very
important
ones.
The
first
one
was
don't
pay
any
license:
fees
for
the
kubernetes
distribution
because
yeah
we
started
with
our
digital
business
and
we
didn't
want
to
have
a
burden
of
the
license
fees
on
them
and
their
motto
was
let
their
business
grow.
First
there
was
this
the
most
important
requirement
in
the
beginning.
For
us,
the
second
one
was:
the
system
must
be
stable,
that's
obvious
yeah,
but
we
learned
that
it's
not
so
easy
to
achieve.
A
We
must
take,
and
the
distribution
should
take
care
about
everything
that
you
don't
want
to
mess
around.
Normally,
with
just
networking
with
storage
and
a
few
more
things
yeah,
we
learned
a
lot
about
that.
It's
a
hard
way
that
it's
not
easy
to
maintain
these
things.
If
you
have
to
so
yeah,
and
also
if
you
look
back
it's
it's
one
of
the
biggest
and
most
important
requirements,
you
should
take
into
account
if
you
choose
kubernetes
distribution.
A
Also,
the
third
requirement
was
that
we
would
like
to
have
the
same
stack
on-premises
and
in
the
public
cloud
and
the
same
user
experience
so
that
our
developers
don't
have
to
switch
around
in
their
minds
with
the
usage
of
the
tooling
independence
of,
if
say,
use
the
on-premise
cluster
or
the
public
cloud
cluster.
We
wanted
to
have
a
look
and
feel
that's
the
same
everywhere.
A
Then
we
went
into
an
evolution.
Phase
five
months
is
very
tough,
so
we
rushed
through
that
very
fast.
First,
we
tried
the
obvious.
We
used
vanilla,
kubernetes
to
create
our
first
clusters
and
had
take
care
about
everything
on
our
own
storage.
Networking
usability
was
was
disastrous
in
the
beginning,
and
so
we
gave
up
very
soon.
That
was
not
the
way
we
wanted
to
work
and,
together
with
our
company,
though,
we
were
searching
for
something
better.
A
So
we
tried
out
several
community
driven
kubernetes
distributions.
I
don't
want
to
name
them,
but
we
had
mixed
experiences.
We
had
problems
with
stability.
I
I
remember
one
tool
that
had
an
automatic
installer
for
clusters
and
every
second
installation
failed
because
of
bugs
user
experience
was
not
so
good
on
the
others.
So
we
were
yeah.
We
had
no
good
feeling
that
we
are
on
the
right
track.
A
During
this
evaluation
phase
it
was.
It
was
a
pure
coincidence
that
we
attended
a
sales
presentation
for
openshift,
because
openshift
violated
our
most
important
requirement
that
we
don't
want
it
to
spend
money
for
our
kubernetes
cluster.
You
remember,
we
did
not
want
to
have
the
burden
of
license
fees
on
our
digital
business,
but
yeah
it.
It
sounded
very
good
what
we
heard
here.
The
salesman
did
a
very
good
job
in
this
presentation
and
yeah.
He
told
us
about
that.
A
A
Regarding
the
features
it
took,
care
about
storage
network
had
a
nice
ui
at
that
time
and
and
great
dev
tools
to
care
about
builds.
Everything
was
integrated,
very
good
in
the
web
ui.
It
was
great
for
our
developers.
We
also
got
very,
very
good
feedback
from
them
and
the
third
one
was
set:
ok,
d3
we
could.
We
could
install
it
everywhere,
on-premise
on
our
vsphere
clusters
and
in
the
public
cloud
in
azure.
It
was
very
easy
to
get
clusters
running.
We
had
lots
of
configuration
options.
A
A
A
A
year
later,
in
2019
we
delivered
even
more
cloud
products
yeah
and
we
were
the
heroes
because
we
enabled
all
of
them
with
yeah
it's
a
great
distribution
in
2019.
Everything
was
everything
was
cool.
This
was
okay.
We
were
very
happy.
We
didn't
regret
that
we
chose
it
also
in
2019
we
improved
and
automated
our
cloud
ecosystem
because
for
the
mvp
we
have
taken
lots
of
shortcuts
and
workarounds
because
we
were
not
so
experienced
with
kubernetes,
and
the
next
goal
was
to
automate
everything.
A
So
we
found
lots
of
tools
that
helped
us
a
lot
in
this
phase.
Ansible.
We
had
experience
with
that
before
I
found
terraform,
that's
absolutely
great
tool
for
creating
infrastructure
on
yeah
different,
look
yeah
with
different
providers,
it's
available
for
vsphere
azure
aws.
For
for
everything
you
can
imagine,
so
we
use
terraform
to
create
the
infrastructure
and
ansible
to
install
and
configure
okd.
A
Then
we
created
cicd
pipelines.
I
liked
a
lot
that
openshift
had
great
support
for
jenkins
or
has
great
support
for
jenkins.
Everything
is
tightly
integrated
in
the
web,
ui,
that's
nice,
and
also
we
created
our
first
service
self-service
portal.
That's
a
tool
running
on
our
cluster
that
provides
our
developers
simple,
wizards
in
a
web
user
interface,
where
you
fill
out
a
few
fields
and
get
tasks
done
on
the
cluster
like
setting
up
a
ci
cd
environment
with
jenkins,
with
the
proper
secrets
everything
completely
automatically
set
up.
A
In
the
beginning
months,
we
learned
that
the
last
release
of
ok
3
occurred
on
in
autumn
2018
and
no
new
version
came
came
out
at
that
time.
Openshift
4
in
the
beginning
of
2019,
I
think
openshift
4
was
released,
but
no
ok,
d4
was
available
and
all
over
there
was
a
completely
year.
No
ok
4
was
inside,
and
this
was
a
problem
for
us,
because
more
and
more
tools
did
not
work
on
ok,
d3,
because
uber
needs
version.
A
I
think
it
was
1.11
got
a
2o
for
lots
of
tools
and
we
had
to
wisely
choose
which
tools
we
use.
This
was
manageable,
but
yeah.
We
were
waiting
for
something
new
for
ok
for
and
it
did
not
come
so
we
started
to
yeah
learn.
What's
blocking
the
release
of
ocd4,
I
myself
was:
I
tried
okay
for
alpha
in
november
2019.
A
A
I
was
so
happy
the
I
remember
that
this
web
ui
was
so
much
advanced
over
that
we
already
laughed
with
okay,
three,
it
was
so
so
much
better,
and
but
it
was
not
easy
to
get
there
yeah.
I
had
to
do
lots
of
mental
steps
hacking
around
in
the
os
in
the
linux
console
to
find
problems.
Why
why
the
installation
failed-
and
it
wasn't
alpha,
it
was
okay
and
yes
and
it
worked
on
vsphere-
well,
very
good
it
if
it
ran.
A
A
That
is
used
in
openshift
several
few
problems
with
that
not
no
big
ones,
but
this
was
my
first
attempt
to
create
pull
requests
to
the
ok,
d4
community
github
reports
and
my
first
pr
was
so
big
because
I
also
patched
a
terraform
code
and
it
was
far
too
far
too
big
and
vadim
rudkowski.
One
of
the
main
supporters
of
okd
was
refusing
it.
A
A
A
I
did
a
lot
lots
of
testing.
I
also
organized
vsphere
license.
There
is
a
try.
Now
it's
not
a
trial.
It's
it's
called
vmware
user
group.
I
don't
remember
exactly
the
product
name
it
it's
available
for
150
euros,
it's
very
affordable,
and
it
did
also
that,
because
I
wanted
to
get
ok
for
life
and
I
reported
lots
of
bucks
fixed
several
of
them,
not
all
bugs
are
so
complicated
to
solve.
A
I
found
out-
and
yes
so
this
was
my
was
a
time
where
also
our
team
learned
much
about
the
insights
of
ok
d4
and
that
we
can
use
the
mechanics
to
almost
solve
any
any
task
we
wanted
to
achieve
yeah.
It
was.
It's
it's
a
great
great
thing.
I
also
did
something
that
may
sound
a
bit
a
little
bit
crazy,
but
I
created
a
t-shirt
for
the
working
group,
video
meetings.
A
I
always
attended
them
regularly
and
the
idea
was
to
increase
the
release
pressure
and
if
everyone
always
sees
this
okd
4ga
on
my
shirt,
it's
what's
not.
So
it
was
more
a
funny
idea
and
I
promised
to
not
change
the
shirt
before
the
release
has
been
made,
but
it
took
a
few
months
yeah.
I
have
to
admit
that
I
changed
the
shirt
in
between.
I
never
told
that
to
anybody.
A
A
A
This
is
also
great,
so
you
have
everything
in
get
no
no
scripts
running
once
and
developers
are
changing
configuration
and
nobody
knows
afterwards
who
has
changed
what,
because
everything
is
in
the
cluster
git
is
the
single
source
of
truth.
That's
nice
with
argo,
and
especially
in
combination
with
okd,
for
we
changed
ourselves
service
portal
to
use
git
ops,
both
of
that
and
also
we
migrated
all
on-premises
apps
from
ok
d3
to
okd4.
We
had
to
change
the
routes
and
other
few
things.
A
Of
course,
a
dns
name
for
ok
d4
contains,
I
think,
a
part
that
is
called
apps
in
the
in
the
url.
That's
a
little
bit
annoying
but
yeah.
We
had
to
change
that
for
all
our
apps
and
in
the
end
it
worked
since
july.
2020
we
upgraded
okd
for
on
premises.
Very
often,
it
almost
always
worked
great
between
openshift
4.6
and
4.7.
There
were
a
few
hiccups,
but
we
could
always
fix
it
or
find
workarounds
together
with
the
community
around
them.
A
Yes,
since
2018,
we
attracted
many
of
our
developers
to
start
a
kubernetes
journey
on
create
a
digital
business
on
our
kubernetes
platform.
That's
great.
I
counted
last
week
that
we
had
onboarded
more
than
50
projects,
not
only
playgrounds
but
real
projects
on
our
okd
clusters
and
it's
available
for
more
than
2000
developers.
You
know
my
company,
it's
running,
running,
very
stable
and
but
we
are
using.
We
are
moving
more
and
more
business
critical
applications
to
our
okd
classes.
A
We
have
lots
of
interest
in
my
company,
we
have
business
critical
applications
and
we
always
say
that
this
should
be
the
time
to
invest
in
commercial
support,
and
we
did
that.
A
few
weeks
ago
we
started
creating
an
arrow
cluster,
that's
the
abbreviation
for
azure
red
hat
openshift
on
azure
for
our
public
cloud
cluster.
A
A
In
fact
you
have
support,
but
not
for
everything,
and
we
are
not
using
all
the
features
of
openshift
at
the
moment
for
all
our
environments,
and
because
of
that
we
chose
okay
for
some
clusters
and
openshift
is
than
the
full-fledged
version
for
the
services.
We
need
full
support
and
yeah
for
the
moment.
We
are
very
happy
with
this
decision
and
to
conclude
what
I
told
you
in
this
presentation,
I
am
absolutely
thankful
and
to
yeah
to
have
okd
during
our
journey.
It
helped
us
tremendously
to
launch
our
digital
business.
A
A
A
They
always
were
very
helpful
and
yeah
vadim,
especially
radim,
is
seems
to
be
online
24
7
on
slack,
and
without
this
guys
we
would
not
have
managed
the
first
steps
with
okd4,
and
thank
you
all.
This
is
our
yeah.
This
was
our
journey.
It
took
us
three
years
now
we
are
absolutely
experienced
in
kubernetes.
I
can
compile
some
modules
on
my
own.
We
know
the
insights
of
of
okd
and
kubernetes
very
good.