►
From YouTube: OpenShift Commons Briefing: EDU SIG OpenShift Origin at Duke with Chris Collins (Duke)
Description
From OpenShift Commons .EDU SIG Meeting on July 13, 2018
OpenShift at Duke: State of the Cluster
Speakers: Christopher Collins Sr. Automation Engineer
Duke University Office of Information Technology
A
So
I
work
for
the
central
IT
department
at
Duke
University.
We
started
looking
at
open
shift
about
a
year
ago
and
talked
to
Stephen
and
his
his
team
just
down
the
street
from
us,
and
they
came
over
and
give
us
a
little
presentation
about
how
they
do
things
and
it
kind
of
made
it
clear
that
this
is
probably
the
way
we
wanted
to
go
with
container
hosting
we've
been
using
containers
for
a
long
time
at
first
just
kind
of
on
their
own
and
then
with
an
in-house
developed
application.
A
So
we've
been
setting
up
a
group
of
concepts
and
we
think
if
it's
going
to
work
for
us
a
little
better
than
like
docker
swarm
or
you
know,
Rock
kubernetes,
because
we,
like
the
rest
of
you,
probably
have
a
whole
bunch
of
customers
that
are
external
to
our
department,
as
opposed
to
some
folks.
That
are
just
doing
hosting
for
themselves
or
hosting
for
a
customers
or
something
so
kind
of
what
we
want
to
really
get
out
of.
A
This
is,
let
me
show
dissipated
winds
here,
the
two
big
things
we
really
want
to
get
like
off
the
bat,
I
think
or
automation
of
our
web
hosting
you
guys,
I
heard
we're
just
talking
about
WordPress.
We
have
a
whole
bunch
of
Drupal
websites
and
it's
cumbersome
to
try
to
manage
those.
We
also
want
to
try
to
use
the
overshift
and
kubernetes
for
research
computing
or
HPC.
A
So
I'll
talk
about
those
two
briefly
for
web
hosting
right
now
we
have
several
hundred
servers
and
getting
really
close
to
a
thousand
websites
that
our
Drupal
at
WordPress.
We
got
two
full-time
Linux
administrators
that
spend
probably
an
unimaginable
amount
of
time
and
holding
the
website
deployment
and
handling
upgrades
in
general,
like
developers
that
affection
is
kind
of
low,
they
want
to
be
able
to
do
their
deployments.
A
They
don't
you
know,
it's
wait
for
the
handle.
You
know:
creation
of
their
websites,
creation
of
VMs
and
Arion.
How
solution
for
a
container
hosting
is
was
our
first
attempt
at
trying
to
fix
that,
but
it's
it's
kind
of
opaque
right
there.
We
don't
have
the
same
visibility.
That
OpenShift
gives
a
developer
with
the
ability
to
look
at
the
containers
that
are
running,
see
what's
inside
them
kind
of
automate.
That
process-
and
you
know
the
CLI
tools
for
open
shifts.
A
A
We
also
want
to
start
using
open
ship
for
research
computing,
we're
specifically
looking
into
being
able
to
use
GPU
support,
that's
in
Coober
days
now
and
try
to
run
some
work
clothes
in
the
cloud.
If
this
is
a
it's
still,
a
very
early
thing
that
we're
looking
at,
but
that
we've
identified
that
as
being
something
that
would
provide
a
huge
benefit
and
honestly,
if
we
can
do
this
web
hosting
automation
and
the
research
computing
aspect
with
open
ship,
and
even
if
we
never
put
anything
else
in
it,
it
would
be
a
gigantic
win
for
us.
A
That
said,
we
also
have
a
couple
other
things
that
we're
intending
to
do.
We,
we
have
a
project,
that's
open
source
right
now
that
has
to
do
with
deploying
honeypot
containers
to
try
to
gather
threat,
intelligence
and
centralize.
Some
of
that
information
in
an
automated
fashion
and
I
think
open
ship
to
be
extremely
useful
for
that,
as
we
have
pods
that
can
be
spun
up
and
shut
down
kind
of
quickly
and
automatically.
A
We
can
move
these
things
around
things
that
look
like
Drupal
web
sites
or
WordPress
sites,
things
that
look
like
as
eh
demons
that
are
running
wide,
open
and
kind
of
entice
attackers
to
give
us
their
information,
but
at
the
same
time
also
make
it
unclear
to
them
whether
or
not
any
given
thing
they're
looking
at
is
is
a
real
website
or
for
a
you
know,
honeypot
website,
so
we
could
gather
we
can
make
it
harder
for
them
to.
You
know,
feel
like
they
want
to
attack
our
infrastructure.
Really.
A
We
also
want
to
provide
the
ability
for
the
Linux
and
Windows
admins
to
deploy
the
infrastructure
services
that
they
do
as
a
central
IT
department.
They
may
not
be
something
that
we
think
of
as
as
OpenShift
running
when
you
talk
about
things
out
and
out
of
the
world,
but
you
know
even
just
like
DNS
servers
or
NTP
servers.
Things
like
that
I
think
work
really
well
in
containers,
and
it
would
be
useful
for
us
to
be
able
to
use
OpenShift
to
deploy
those
kind
of
things
and
then
functions
of
the
service.
A
I
think
it's
probably
something
that
we're
going
to
make
much
more
use
of,
especially
with
the
honey
pots
and
some
of
the
other
developers
that
we
have
that
are
doing.
Work
with
kind
of
containerized
functions
as
a
service
at
Bethel.
I
think
will
give
us
a
lot
of
density
and
provide
a
lot
of
ability
for
the
developers
to
kind
of
more
flexibly
deploy
the
things
they
want
to
deploy
and
change
out
little
pieces
of
micro
services
rather
than
having
to
develop.
You
know
full-fledged
framework
applications
at
any
given
time
and
manage
those
all
together.
A
You
know,
and
then
kind
of
at
the
very
end
of
some
things
that
we're
starting
to
look
at
is
a
coop
Virts.
The
the
idea
of
being
able
to
use
the
kubernetes
virtualization
layer
on
top
of
KBM
and
start
managing
virtual
machines
through
openshift
is
extremely
appealing
and
the
idea
of
using
federated
clusters,
I,
there's
rumors
that
that's
going
to
be
available
relatively
soon
that
we'd
love
to
be
able
to
connect
clusters
running
here
in
our
data
center
to
clusters
and
other
data
centers
or
in
AWS
or
Google
compute
or
Microsoft
Azure.
A
So
that's
that's
kind
of
what
we're
hoping
to
get
out
of
OpenShift,
where
we
are
now
we're
in
a
very
small
alpha.
Internally,
we've
got
a
small
development
cluster
that
we've
stood
up.
Our
operations
team
is
trying
to
kind
of
figure
out
how
they're
going
to
manage
this,
where
we're
still
trying
to
build
community
as
well
get
people
involved.
A
We
started
an
open
shift
users
group
for
Duke
we're
gathering
some
interested
developers
from
other
teams
and
giving
them
kind
of
early
access
so
that
they
can
play
around
with
it.
And
you
know
everything
right
now
is:
we've
got
a
cluster,
it's
not
really
production
ready,
but
everybody's
learning
we've
decided
that
you
know.
We've
made
it
very
clear.
A
Some
of
our
technical
challenges-
I'm
sure
everybody-
runs
it
because
some
of
these
things,
but
most
recently,
is
the
Red
Hat
acquisition
of
core
OS.
We
are
running
our
development
cluster
right
now
on
top
of
relative
coasts.
At
this
point,
we're
there's
been
a
bunch
of
talk
about.
You
know
what
the
roadmaps
gonna
be
for
relative
ik,
but
I
think
the
system
ins
are
a
little
bit
uneasy
we're
trying
to
set
up
a
productive
service,
but
it's
not
we're
not
sure.
Actually
what
operating
system
is
going
to
be
running
on
in
six
months?
C
A
That'd,
be
great,
that'd,
be
great
and
I.
Think
I
think
it's
gonna
be
fine.
We
we
got
some
early
access
to
atomic
I
think
we're
plan
to
get
some
early
access
to
I
guess
what
container
linux
is
going
to
be
so
we'll
have
a
chance
to
play
with
it,
but
it's
it's
just
so
sort
of
hard
to
move
forward
without
knowing
what
the
challenges
are
going
to
be
because
it's
going
to
be
deployed
on
something
completely
different.
Yeah.
C
A
That
sounds
great.
Thank
you.
One
of
the
other
issues
is
that
everything
that
we're
doing
is
sort
of
a
moving
target.
Everything
is
is
changing.
Really
fast,
kubernetes
is,
being
you
know,
is
releasing
new
changes
quickly.
Open
ship
is
releasing
new
changes
quickly.
Open
ship
ansible
repository
is
changing
really
relatively
quickly,
so
we
have
a
little
bit
of
trouble,
keeping
up
with
where
we
are
in
any
given
situation,
especially
since
we're
still
learning
each
of
these
pieces.
We're
we're.
Definitely
not
experts
at
this.
We're
we're
just
kind
of
getting
and
started
to
it.
A
A
Given
time,
then
everything
needs
to
kind
of
work
exactly
the
same
for
the
end
user
and
if
we
have
to
pick
different
people,
storage
classes
or
change
the
way
that
they
interact
with
their
storage.
That's
going
to
be
I.
Think
that's
going
to
be
a
pain
point
for
us.
So
I
think
what
we're
really
interested
in
doing
is
evaluating
the
container
native
storage
that
Red
Hat
has
been
talking
about
at
the
Red
Hat
summit,
using
the
Gluster
storage
inside
of
containers.
A
So
we're
evaluating
that
we
just
don't
have
any
Gluster
or
suffix
experience
with
our
teams.
Yet
so
that's
a
learning
learning
thing
for
us
as
well.
We're
also
planning
on
using
s3
storage
on
premises
or
our
container
images,
and
we've
been
doing
that
for
a
while.
So
I
think
that,
should
that
shouldn't
actually
be
an
issue
we're
just
you
know,
we've
got
to
figure
out
how
it
is
that
this
stuff
works
and
kind
of
gather
the
expertise.
A
And
then
you
know
our
existing
infrastructure
was
developed
for
a
different
time,
and
a
different
type
of
workload
like
containers
are
completely
different.
Open
ship
is,
is
completely
different,
so
we
have
to
figure
out
how
our
infrastructure
works
within
open,
shipped
environment.
How
do
we
adapt
our
infrastructure
or
how
do
we
adapt
open
shift
to
work
with
our
infrastructure?
And
some
of
that
is
a
very
cultural
thing
which
kind
of
leads
into
the
next
topic.
A
We
have
some
non
technical
challenges
as
well
and
I
think
everybody
really
does
with
containers
in
general
and
open
ships,
specifically
there's
there's
kind
of
a
learning
curve,
especially
for
operations.
The
idea
of
running
a
full
sized
cluster
that
is
very
interdependent
and
has
a
lot
of
moving
pieces
and
a
lot
of
moving
technologies
is,
is
a
it's
difficult
to
kind
of
wrap.
Your
head
around
at
first
and
you
know
people
sort
of
fear,
change
right.
A
So
this
isn't
what
they've
done
for
the
last
ten
years
of
their
system,
administration,
career
or
the
developer
career,
and
so
we're
hoping
to
kind
of
ease
that
in
or
folks
so
that
we
can
show
them
the
benefits
and
help
help
bring
them
up
to
speed
in
a
way
that
it's
clear
what
the
value
is.
We
have
some
challenges
with
internal
team
structures
where
things
are
a
little
siloed.
A
A
We
also
have
issues
with
what
what
is
a
priority
for
us.
It's
clear
from
upper
management
that
open
shift
is
accepted.
It's
something
that
we're
going
to
doing
it's
the
way
that
we
want
to
go
with
containers
it's
the
way
we
want
to
go
with
development,
but
there's
a
layer
of
I,
guess,
there's
a
gap,
there's
a
gap
between
upper
management
and
the
the
actual
folks
doing
the
implementation,
where
the
focus
on
what
needs
to
be
done
now.
What
needs
to
be
prioritized?
A
Does
everybody
drop
everything
and
work
on
open
shift
time?
Do
we
work?
You
know
only
when
we've
got
free
time
like
how
do
we?
How
do
we
decide
that
this
is
if
this
is
something
we're
going
to
be
doing,
and
this
is
definitely
an
organizational
issue-
it's
not
a
technical
issue.
It's
not
even
an
open
shift
issue.
It's
just
the
way
that
technology
has
been
adopted.
It
has
been
a
little
bit
different.
It's
it's
not
all
right!
So
guess
I
was
on
the
existing
culture.
A
We
really
just
need
to
try
to
figure
out
how
this
stuff
is
gonna
fit
in.
With
with
what
we're
doing
you
know
using
the
cloud
federated,
you
know
computing
jobs,
couvert
Sdn,
you
know
OpenShift
coop
control.
All
that
stuff
is
that,
as
my
boss
put
it,
it's
like
training
for
Mars
right
and
we've
got.
We've
got
a
bunch
of
really
in
town,
really
intelligent,
really
talented
people
that
are
already
great
fighter
pilots,
and
now
we
just
need
to
bring
them
that
extra
level.
A
A
A
Like
everything
all
the
way
down,
you
know
got
his
coat
and
a
git
repo
got
it
cloned
into
the
PHP
builder
image
deployed
it
added
a
route
out
of
persistent
volumes,
and
so
I
was
I
was
a
little
bit
shocked
and
I
was
like.
How
long
did
this
take
you
like?
Who
are
you
working
with
and
he's
like?
Oh
I
just
was
playing
around
with
it.
A
I've
only
been
doing
it
for
about
15
minutes,
so
just
on
his
own,
without
any
system
intervention,
he
was
able
to
deploy
a
Drupal
website
using
the
dev
cluster
that
we
have
and
that
that
just
kind
of
made
my
day.
That's
that's
all
that
we
have
been
trying
to
do
for
years
and
and
OpenShift
made
it
that
easy
for
him.
A
So
with
that,
if
anybody
would
like
to
chat
if
you
would
like
a
copy
of
this
presentation
without
you
know
the
30,000
internet
connection,
problems
that
I've
had
I
can
be
reached
here
and
I'm.
Definitely
looking
forward
to
like
working
with
everybody
in
the
edu
sig.
You
know,
UNC
has
been
super
helpful
and
giving
us
some
advice
and
I'd
love
to
see
what
other
people
are
doing
as
well
and
we're
happy
to
share.
A
C
You
Chris
I
think
it's
really
it's
good
to
hear
from
people
at
all
of
different
stages
of
deployments
and
things,
and
while
you
were
offline
for
a
few
minutes,
there
was
a
little
bit
of
a
conversation
I'm
going
to
ask
even
to
repeat,
as
I
asked
him
the
question.
If
any
of
those
storage
problems
sounded
familiar
to
him
and
he
had
a
good
answer
to
that.
A
B
When
we
start
out,
our
intention
is
to
start
out
with
the
cluster
on
DNS.
We
do
have
some
in-house
cluster
knowledge
outside
of
openshift
entirely.
We
just
use
clusters
as
a
product,
but
in
openshift
I
haven't,
got
it
up
and
running
quite
yet.
It
appears
today
because
that
rapke
to
me,
is
it
Hetty,
an
API
layer
around
managing
the
storage
and
so
from
what
I
read.
It
should
become
automated
deployment.
B
You,
when
you
associate
to
a
volume,
claim
I,
think
and
I'm
hoping
and
the
thing
we
liked
about
it
is
that
you
don't
need
to
actually
interface
it
directly
to
any
particular
back-end
storage,
yeah
Cole.
Ok,
with
the
you
know,
the
node,
and
so
that
makes
it
cross-platform.
If
we
want
to
go
to
AWS
or
Prem
or
whatever,
which
is
really
attractive.
Is
it
one
way
to
access
it.
A
B
C
The
Gloucester
people
will
be
happy
to
hear
that
and
well
well,
we'll
definitely
try
and
get
some
of,
maybe
that's
the
topic
for
for
a
future
meeting
as
well.
So
you
can
get
someone
from
that
team
to
come
in.
We
figure
out
exactly
what
the
questions
are.
We
want
to
ask
the
other
thing
that
you
mentioned
at
the
beginning:
Chris
that
was
I.
Think
interesting.
Who
was
your
plans
or
your
hopes
for
the
research
and
HPC
bits
and
I?
C
Think
that's
common
across
lots
of
lots
of
edu
organizations
and
and
research
groups
as
well
there's
a
lot
of
even
in
the
data
science
departments,
and
that
there
a
use
of
Jupiter,
hub
and
hosting
guru.
Folks
and
we've
got
some
really
good
content
out
there
already
on
how
to
use
Jupiter
and
that
sort
of
stuff.
There's
some
good
contents
on
GPU.
That's
out
there
as
well
that
we
can
hook
you
up
with,
and
it's
been
a
lot
of
work
on
performance.
C
So
once
you
get
the
basics
done
of
getting
through
your
openshift
origin
deployment
up
and
running
I
think
there's
there's
a
lot
you
can
do
with
it.
It
will
be
well
we'll
be
happy
to
hear
more
every
Mindy
every
three
months
there,
six
months
get
an
update
from
you
on
on
where
you're
at
with
these
things,
yeah.