►
From YouTube: Kubernetes Kops Office Hours 20180316
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
Hello,
everyone:
it
is
Friday
March
16th.
This
is
cops
office
hours.
We
have
a
fairly
light
agenda
today.
Please
add
things
you
want
to
the
agenda.
Otherwise,
I
will
get.
I
will
jump
right
in
and
start
with
the
first
time
on
the
agenda,
which
is
I
guess
on
the
I
added,
which
is
the
serious
of
cops
one,
the
119
released
for
cops.
We
did
an
alpha
one.
A
Almost
two
weeks
ago
now
we
found
a
bunch
of
issues
relating
primarily
to
I,
guess,
I'd,
say
lifecycle
issues,
so
upgrades
and
repeatedly
running
cops
update,
and
so
we
are
working
through
those
I
think
we
have
basically
got
them
all
done,
except
there's
one,
maybe
issue
still
open
around
detection
or
validation,
of
instance,
groups
in
a
rolling
update.
So
but
yes,
probably
ready
to
do
the
next
alpha
gets
the
fixes
that
we
have
done
out
there
and
hopefully
get
more
feedback,
and
hopefully
that
way
the
last
round
of
problems.
A
Even
earlier
and
I've
been
adding
things
like
a
I
call
it
like
a
lifecycle
integration
test,
but
essentially
by
mocking
the
Animus
cloud
API
is.
We
are
now
able
to
run
effectively.
It
cops,
create
the
cops
update
and
it
cops
update,
for
example,
and
that
lets
us
see
the
bug
which
otherwise
to
do
props
great
pops
up.
That
comes
update,
but
it's
now
a
it's
a
it's
effectively
a
unit
test
like
it
doesn't
touch
any
any
rate
real
api's,
and
so
we
should
never
have
that.
A
Well,
once
it's
once
it's
fully
once
it
covers
all
our
bar
one
of
our
scenarios,
then
we
won't
have
that
problem
again
going
forwards.
It's
still
like
this
requires
mocking
the
hrs
API,
which
is
obviously
a
lot
of
work,
but
it
is
we're
getting
there
and
we
have
coverage
on
the
these
limits
where
we
had
bugs
in
one
I
know
alpha
one:
that's
actually
all
we
have
on
the
agenda.
I
know.
A
A
I
mean
definitely
I,
think
I
think
that
that
applies
to
cops.
It
also
applies
I,
think
there's
also.
We
don't
necessarily
have
we're
not
adding
new
resources,
particular
this
time,
but
I
think
another
one
that
might
be
is
like.
You
might
expect
an
alpha
to
delete
your
cluster
by
mistake.
In
other
words,
like
your
data
may
disappear,
where
Z
would
be
very
unhappy
if
a
beta
did
that
regardless.
B
A
Like
we
are
simply
because
we
in
alpha
one,
we
found
some
issues,
I,
don't
think
any
of
them
were
they
lost
type
issues.
We
certainly
found
some
issues
where
it
was
like
doing
unexpected.
We're
trying
to
do
unexpected
things
with
the
cloud
resources
which
could
have
escalated,
so
I'd
want
to
keep
myself
and
hopefully
fast,
follow
on
the
beta,
because,
hopefully
no
more
well.
Hopefully
there
will
be
no
bugs
if
they're
already
bugs
out
there
found.
D
A
A
E
I
think
me,
and
a
couple
of
guys
are
new
and
we
just
joined
today.
We
are
really
interested
in
trying
to
help
out
on
the
OpenStack
side
of
things,
but
we're
really
worried
about
testing
our
environment
specifically,
is
on
okada
and
like,
as
you
said,
you
mocked
up
like
the
AWS
api,
there's
a
lot
of
changes
that
are
going
on
in
OpenStack
and
we're
not
really
sure
how
we're
going
to
be
able
to
successfully
help
you
guys
with
OpenStack
without
it
being
kind
of
pigeon-holed
into
what
our
company
is
currently
running.
E
A
I
think
I
think
so.
I
was
regretting
on
having
done
the
AWS
mocking
earlier
and
having
like
that,
seems
very
high
return
on
investment.
It's
a
bit
of
a
pain,
but
it's
very
hard
to
invest
in
terms
of
like
getting
really
like
tested
really
test
things
without
doing
a
whole
lot
of
work
gives
you
like.
Well,
we
have
to
have
a
little
bit
protests.
Did
you
think
a
lot
of
upfront
work
like
to
mock
out
the
API
is
but
then
or
implement
the
interfaces,
but
then
it's
it's
pretty.
A
You
get
a
lot
of
value
at
it
or
we
found
a
lot
of
issues
with
it
that
you
know
it's
nice
to
have
continuously
tested.
The
other
nice
thing
about.
That
is
that
we
can
have
testing
that
doesn't
require
an
openstack
API
implementation,
so
it
can
run
in
the
normal
ete
like
continuously
without
you
having
to
get
into
like
federated
testing
type
scenarios.
Yeah.
E
That's
actually
what
I
was
crazy
worried
about
I
was
like
I'm,
just
gonna
spin
up
a
bunch
of
different
OpenStack
cloud.
In
my
server
rack
at
my
house,
yeah
I
was
like
I'll,
just
run
three
iOS
and
and
I'll
just
test
against
those,
but
now
I
guess
we
can
we'll
have
to
think
about
that
when
we're
running
our
stuff
I
think.
A
E
It
really
depends
there's
so
many
services
moving
all
at
the
same
time.
Heat
and
Magnum
are
probably
good
in
between
one
major
version,
but
after
two
major
versions
there
could
be
there
could
be
breaking
changes,
but
luckily
because
their
device
design
like
CloudFormation
they're
backwards
compatible,
usually
so
I
wasn't
so
much
worried
about
that.
As
I
was
like
just
a
vast
majority
of
different
networking
types
and
different
vendors
that
are
involved
in
OpenStack,
that's
where
it
gets
scary
and
that's
where
AWS
is
way
more
streamlined,
so
yeah
I.
A
Mean
it
would
be
wonderful
to
have
your
contribution
that
I
will.
However,
I
can
I
think
if,
if
you
want
to
like
don't
worry
about
targeting
other
people's
use,
cases
is
my
I
sort
of
advice.
I
think,
like
other
people
will
will
speak
up
on
their
use
cases
and
I
think
the
most
important
thing
is
to
get
one
use
case
working
for
OpenStack
and
then,
if
that's
our
use
case,
that's
great,
because
it's
really
hard
to
work
for
someone
else's
use
case,
and
once
we
have
one
then
the
second
will
be
much
easier.
So
yeah.
A
E
A
That's
a
related
project
to
effectively
bring
the
management
of
instances
under
AWS,
sorry
under
kubernetes
management,
and
what
that
will
mean
is
so,
for
example,
when
you
you'll
have
like
a
your
no-till,
we
manage
as
an
API
object
and,
if
you
wanted
to
add
another
note
which
might
not
be
is
read
well
I.
Guess
anyway,
you
want
to
add
another
node.
A
You
might
either
create
a
machine
object
which
is
from
like
a
pod
or
you
might
change
the
target
number
of
machines
in
a
machine
set
or
a
machine
deployment
type
concepts
and
it
would
go
and
create
the
pod
like
machine
object
and
a
controller
would
then
go
in
create
a
mission
so
we're
trying
to
get
there.
Everyone
is
sort
of
the
criminis
community's
trying
to
get
there
and
it
will
mean
that
specifically
for
cops
or
what
you're
working
on
it
means
that
running
the
masters
or
running.
A
E
E
F
A
F
A
The
changes
are
coming
back
on
the
dip
already
changes,
it's
like
a
diff.
Yes,
if
you
want
to
go
and
like
so
one
of
the
things
you
can
do
is
on
an
update.
You
can
effectively
go
through
the
changes
and
effectively
cross
them
off.
You
know
them
out
and
then
at
the
end
you
can
check
if
the
object
is
reflective
equals
to
the
empty
object,
in
other
words,
whether
you
have
actually
made
all
the
changes
that
have
been
made.
A
F
H
So
the
controller
itself
is
ready.
It's
just
the
cops
integrations
that
I'm
working
on
so
actually
one
question
or
two
questions
actually
so.
First
one
is
because
digitalocean
droplet,
don't
support
kind
of
the
iron
features
that
like
AWS
or
GPS,
so
we
can't
access
kind
of
the
cloud
features
without
having
a
key
inside
the
actual
machine.
So
that's
a
problem
for
node
up
and
pretty
much
pretty
important
for
cops
to
work.
A
H
A
F
A
There
is
also
something
else:
coming
or
as
I
have
I
have
a
machine
controller
like
concepts,
implementation
which
brings
it
up
on
the
em
where,
but
what
it
does
is
it
brings
up
a
machine
and
then
it
SSH
is
to
the
machine
and
drops
the
stuff
in
that
way.
So
it
doesn't
help
with
the
masters,
but
it
could
be
an
option
for
like
the
future,
but
yes,
I,
think
for
now
instance.
Data
is
probably
great
I
think
you
know
the
you
have
to
be
careful
about
security,
but
so.
H
F
Okay,
the
other
thing
I
was
gonna
mention
Andrews
that
just
and
I
was
talking
about
Machine
API
yeah,
that
was
nice
follow-up
question
was
I,
saw
the
proposal
merged,
and
so
it's
not
a
place
where
we
might
and
you
in
digitalocean
needs
it
right
and
you
guys
don't
have
instance,
groups
yeah.
A
I
I
mean
I,
have
a
proof
of
concept
for
adopting
a
version
from
like
a
month
ago,
which
makes
it
gets
into
the
bare
metal
environment,
and
so
what
it
does
is
it
I
think
we
shift
the
bundle
idea
where
you
know
we
take
the
files
that
aren't
as
three
and
the
way
it
works.
Is.
It
then
will
SCP
them
across
and
SSH
to
exec,
so
it
doesn't
even
need
user
data
or
like
a
Boonton,
ops,
scribbs.
A
It
just
needs
the
ability
to
come
up
and
recognize
a
public
key
and
I'm
and
to
find
the
IP
address.
But
those
are
the
two.
So
that's
like
a
much
more
generic
I
think
the
machines
API
is,
you
know,
kubernetes
type
project,
so
this
will
work
everywhere
in
in
like
not
just
cops
but
anywhere,
and
it's
good,
because
then
things
like
the
autoscaler
will
work
consistently,
whether
you
install
the
cops
or
she
can't
hear
you
KS
or
whatever
it
is
right.
A
Justin,
yes,
but
the
I
think
the
biggest
open
question
which
is
currently
under
discussion
and
I,
think
Jessica,
o
Chen
is
hardly
working
on.
It
is
the
bootstrapping
and
how
you
bring
up
effects
like
so.
If
you
magically
have
some
masters,
it
it's
a
great
model
and
simple,
and
it
seems
easy,
I
think.
But
it's
trapping
is
the
slightly
open
question
I.
A
A
Yeah
again,
if
there's
anything,
I
do
to
help
just
ping
me
on
slack
and
that
ie
my
psyche
is
a
bit
of
a
disaster
zone,
but
I
will
eventually
circle
back
to
them.
Okay,
thanks
yeah,
I!
Guess!
If
no
one
else
does
anyone
actually
does?
Is
there
anything
else
anyone
would
like
to
talk
about
or
bring
up.