►
From YouTube: Kubernetes Office Hours 20170906
Description
Come bring your questions for Kubernetes Developers!
Post your questions to slack: https://kubernetes.slack.com/messages/office-hours/
A
A
A
A
A
B
A
A
The
idea
is,
is
that
if
you're
in
kubernetes,
Hache
kubernetes
des
users
or
you're
being
stuck
trying
to
figure
out
kubernetes
that
you
can
come
to
an
office
hours
once
a
month
and
at
least
present
your
problem
to
the
community
and
then
see
how
we
can
help
you
out,
we've
got
a
panel
usually
of
developers
that
are
coming
to
help
help
users.
So
what
we'll
do
is
we
hang
out
in
hash
office
hours
on
slack
and
I've?
How
did
that
link
to
the
YouTube
description
of
the
livestream?
A
So
you
can
just
click
through
there
and
sign
up
for
slack
if
you're
not
already
signed
up
and
then
start
to
ask
your
questions
so
we're
in
hash,
offish,
ours
and
the
way
it
works.
Is
you
ask
a
question
and
then
we'll
present
it
to
our
panel
and
then
we'll
do
our
best
to
answer
it
for
you
so
some
caveats
here?
A
Not
all
questions
will
be
answerable.
So
if
you
have,
for
example,
say
a
really
complex
networking
question,
usually
our
experts
can't
ssh
to
your
machines
or
you
know,
have
access
to
all
the
stuff
that
you're
running
so
generally
speaking,
there's
only
a
certain
set
of
questions
that
we
can
answer.
So
what
we
try
to
do
is
make
those
as
generic
and
reusable
for
you
as
possible.
A
So
while
we
might
not
be
able
to
debug-
let's
say
a
worker
node
that
problem
that
you're
having
because
we
can't
actually
see
the
logs
or
whatever
we
should
be
able
to
kind
of
walk
you
through
the
steps
on
on
where
you
can
get
started.
How
you
can
you
know
where
to
find
the
logs
that
you
need
to
do
that
that
kind
of
debugging
that
sort
of
stuff-
and
they
kind
of
basically
gets
you
at
least
pointed
in
the
right
path
to
help
you
solve
your
problem
with
that.
A
We
do
ask
our
panel
that
when
they
do
explain
something
to
kind
of
give
a
little
history
or-
or
you
know
a
little
bit
of
tidbit
of
generic
information
about
that
part
of
kubernetes,
so
that
everybody
kind
of
understands.
Okay,
that
person
asks
a
question
about
the
a
node.
You
know.
Oh
okay.
This
is
why
that
is
sometimes
like.
Last
time
we
had
a
person
who,
like
helped
write
that
feature.
You
know
kind
of
explain
why
things
were
that
way,
and
things
like
that,
so
the
idea
being
is
that
you
watch
enough
of
these.
A
You
should
start
to
understand
how
the
different
components
of
kubernetes
fit
and
work
together,
so
without
further
ado
in
the
office
hours.
If
you
guys
just
want
to
go
channel,
you
guys
want
to
go
ahead
and
start
pasting
in
your
questions,
just
type
question
in
all-caps,
colon
and
then
your
question,
and
then
we
will
get
to
them
as
soon
as
we
start
receiving
them.
So
for
those
people
who
aren't
here
last
time.
B
A
B
A
So
we'll
just
give
us
a
minute
here:
there
is
a
perceived
delay
here
or
actual
delay
between
when
we
say
stuff
and
then
people
can
respond.
I
did
I,
did
change
YouTube
to
a
low
latency
configuration,
but
it
doesn't
support
my
resolution.
So
it's
it's
normal,
but
it
worked
decent
enough
for
last
time.
As
long
as
we're
not
expecting
to
have
two-way
conversations
with
people
in
the
channel
so
yeah
it
sounds
like
you
have
sound
No.
C
A
Awesome
so
we've
got,
we've
got
two
developers
here,
great
I've
gotten
some
responses
that
people
might
be
dropping
in
and
out
as
the
as
the
hour
goes
by
note,
no
worries.
We
have
enough
experts
here
to
to
make
this
happen.
So
now
we're
just
kind
of
waiting
for
the
questions
to
start
coming
in
from
slack
and
I
will
start
to
monitor
Twitter
as
well.
If
you
have
any
questions
that
you
want
to
send
them
in.
C
B
A
A
All
right.
You
ready.
We
got
a
first
question
from
a
blank:
oh
I'm,
creating
nginx
ingress
controller,
but
when
it's
starting
appears,
the
air
air-crane,
creating
client
configuration
open,
environs
secrets,
kubernetes
io
service,
blah
blah
blah.
Sorry
it's
in
the
channel,
so
you
guys
can
read
it
no
such
file
directory.
Why
exist
a
file,
token
CSV
format
and
Master
and
minions?
Please
help.
B
Gets
automatically
created
and
mounted
into
your
containers
in
your
pods
of
saying
that
it
exists
in
the
master
in
the
nodes
is
a
little
confusing.
You
would
have
to
go
look
inside
the
actual
container
at
that
path
and
see
if
it's
there
and
if
it's
not.
There
could
be
a
couple
reasons.
We've
had
issues
in
the
past
in
older
versions
of
kubernetes,
where,
for
whatever
reason
there
were
fine
mounting
issues
or
permissions
issues
or
selinux
issues.
So
there
are
a
variety
of
reasons
that
you
might
be
running
into.
B
That
I
would
maybe
start
with
like
a
simpler
pod
like
trying
to
do
cute
controller,
run
busy
box
and
see
if
that
particular
path
and
file
exists
and
a
busybox
pod,
and
if
it
does
then,
or
if
it
doesn't,
then
it's
more
likely
a
generic
problem.
It's
gonna
affect
everything
and
you'll
need
to
try
and
diagnose
from
there.
The
first
log
file
that
I
would
go
look
at
would
be
the
log
file
for
the
cubelet,
because
that
is
what's
ultimately
responsible
for
placing
that
file
at
that
path.
B
B
C
B
So
there's
also,
if
you
were
doing
a
custom,
install
the
admission,
there's
an
admission
controller
that
you
have
to
enable
in
the
API
server
to
actually
get
the
service
account
token,
as
an
automatically
mounted
volume
at
all
of
your
pods.
So
I'm,
assuming
you
probably
have
that
by
default.
But
that's
one
other
thing
to
check
as
well.
C
Use
of
your
building
index
controller
or
creating
one.
Is
it
one
that
you're
writing
from
scratch,
or
is
that
one
of
the
examples
remember
from
one
of
the
community
readers
so
yeh
sort
of
it
sounds
like
it's
more
generic
issue
with
with
these
the
secret
man
the
default
date.
This
thing
is
called
default
service
account
secret.
That's
general
problem
here,
that's
something
you
have
probably
and
ya
wonder
it
probably
isn't
related
to
to
whether
you're
writing
your
own
controller
or
not.
A
And
it
looks
like
he's
typing
a
response
and
slack
so
we'll
go
ahead
and
wait
and
come
around
for
him.
We
could
definitely
use
more
questions.
People
in
the
channel
if
you
want
to
go
and
post
some
and
while
that's
happening,
I've
got
a
quick
question
for
the
panel
of
a
question
that
I
was
thinking
of
this
week.
As
far
as
workloads
go
on
a
node
when
choosing
like
constraints
as
far
as
memory
and
CPU.
Are
there
any
tricks
as
far
as
Japan
off
the
stream
again.
C
A
Yep
all
right,
it's
working
good
for
one
person.
If
we
can
get
a
confirmation,
we'll
just
continue
we'll
try
to
power
through
here.
B
A
B
B
B
B
Not
gonna
be
able
to
reach
into
each
other's
so
I.
Don't
know
this
answers
the
question,
but
my
advice
is
basically
know.
What's
in
your
docker
files
of
the
images
that
are
running
and
if
you
want
to
change
things,
you
can
override
command
and
the
arguments
that
are
running
you
can
I
believe
override
user.
So
that
should
and
I
believe
that's
pretty
much.
A
Okay,
and
if
you
have
a
follow-up,
please
go
ahead
and
post
it
while
we're
waiting
for
more
questions,
just
encourage
everyone,
who's
started
to
join
us.
Please
keep
posting
your
questions
in
hash
office
hours
and
we'll
get
to
them.
The
stream
was
giving
us
some
problems
when
I
asked
this
question,
so
I'm
just
gonna
react
it
and
ask
you
guys
to
just
go
ahead
and
restate
your
answers.
A
So
the
question
was:
I
I
have
a
friend
who's,
doing
a
lot
of
Java
workloads
on
kubernetes
and
he
had
a
question
on
how
to
figure
out
exactly
how
to
determine
the
limits
for
his
CPU
and
memory
based
on
you
know.
Java
does
like
garbage
collection
and
things
like
that,
and
are
there
any
gotchas
or
any
recommendations
as
far
as
figuring
out
how
to
allocate
resources
based
on
the
workload
that's
running.
B
D
D
A
D
A
Yeah
so
I
feel
like
one
of
the
first
mistakes
I
made
when
I
was
learning
is
not
setting
limits
on
anything
and
then
things
would
get
killed
when
things
got
busy
and
I
would
never
understand
why
that
is
so.
I'm
really
glad
that
work
is
going
into
that.
A
Okay,
anybody
else
have
anything
to
add
to
the
Georges
Java
question
before
we
move
on
okay,
Nexxus
from
ceará
I
have
a
cluster
setup
with
cube
admin
using
net
weave
as
CNI
now
I
have
some
pods
running
on
the
host
IP
like
the
cube,
API,
server,
proxy
and
so
on,
and
they
work
fine.
My
own
pods
are
running
with
an
IP
of
10.30
to
0.0,
slash
the
default
of
weave
and
those
containers.
Don't
have
internet
access.
Any
suggestions
on
how
to
solve
that.
C
See
you
are
basically
saying
that
your
pods
are
not
able
to
access
it.
I,
wonder
whether
you've
tried
to
enable
yeah
I
mean
this
could
happend,
if
you,
if
you
create
network
policy
of
any
sort.
So
if
you
we're
playing
around
with
natural
policy
and
then
maybe
stopped
in
so
that
could
have
triggered
these
canopy
hey
there.
A
B
D
A
D
Think
committee
M
supports
all
the
major
distros
now
so
it
it's
a
dealer's
choice
of
what
features
you're
looking
for
and
what
what
you
want
in
an
operating
system
right
some
people
like
the
immutability
of
cloud-based
operating
system,
we're
operating
in
the
cloud,
and
you
don't
want
anything
to
change,
that's
where
core
OS
or
one
of
the
Atomics
works.
Well,
if
you're
looking
for
a
traditional
because
you're,
you
know,
you're
a
blue
to
shop
or
whatever
it
supports
the
blue
to
I.
Think
it's
all
based
upon
the
requirements
of
your
environment,
really.
B
I
would
echo
that
using
cue
video
is
the
Avenue
I
would
recommend
so
whatever
it
currently
supports,
as
Tim
mentioned,
I
personally
have
been
using
Fedora
CentOS
Ravel
for
years,
just
in
history,
so
I'm
very
familiar
with
that,
and
and
it's
like
a
you
know,
personal
comfort,
but
you
know
Debian.
It
certainly
works
fine.
Well,
basically,
whatever
cube
ADM
supports.
A
Yeah
I
was
just
gonna
add
you
know
my
previous
employment
had
me
on
the
abouttwo
side
of
the
house,
so
I've
actually
been
stepping
out
of
my
comfort
space
to
try
different
os's
and
any
OS
that
supports
cube.
Admin
is
pretty
much
almost
exactly
the
same
I've
not
quite
yet
decided
whether
it's
good
or
bad.
For
me,
if
moving
over
to
something
like
an
atomic
or
a
core,
OS
is
right
for
me,
but
there's
definitely
options
there.
A
I
I
think
the
way
I'm
gonna
answer
the
OS
question
from
now
on
is
basically
all
the
major
ones
are
supported
by
cube
admins,
so
use
like
if
you're
doing
this
at
work
use
the
same
OS
that
you're
going
to
be
already
using.
You
know,
whatever
part
of
your
platform
is
I,
don't
I,
don't
think,
there's
a
reason
for
you
to
change
something
that
you're
not
familiar
with.
A
You
know,
even
though,
once
you
install
kubernetes
you're,
not
really
interacting
with
the
host
that
much
but
I
don't
know,
that's
just
me:
I'm,
definitely
having
a
lot
of
fun
playing
with
different
operating
systems.
So
so,
let's
see
still
waiting
for
more
questions.
Let
me
just
check
to
see
if
I
missed
any.
A
A
A
C
Based
upon
logic,
you
play
project,
which
is
that
a
rathole
style
replacement
for
four
come
on
guys
climb
type
stuff
for
SAT.
If
you
kind
of
bits
out,
if
you
could
all
you
can
try
to
play,
it's
very
experimental,
it's
personal
project
if
you
lose
Ruby,
as
as
they
can
syntax,
but
like
other
than
that,
this
is
fairly
straightforward
thing
to
the
town
to
use.
If
you
don't
wanna
put
on
yes
pods,
that's
that's
all
namespaces
or
not
even
a
in
you
play
you
just
right
cause
you
get!
C
A
Please
please
toss
that
URL
in
the
slack
channel,
so
we
can
check
it
out.
I'm,
always
looking
for
alternate
means
of
poking
at
things.
Alright,
so
Slavic
responds
back
with
zookeeper
Java
has
a
bug.
Client
has
a
bug
and
the
zookeeper
server
IP
changes
client
has
to
be
restarted.
It's
unlikely
that
bug
would
be
resolved
anytime
soon.
Other
than
forking
zookeeper
or
changing
designed
to
remove
zookeeper
was
thinking
to
temporarily
work
around
the
issue
by
statically
assigning
the
IP
zookeeper
cluster
members
managed
as
a
stateful
setting
kubernetes
there.
D
A
A
Would
be
great
just
toss,
those
in
the
channel
and
Slavic?
Let
us
know
how
that
works
out
for
you.
That'd
be
great.
A
Mulholland
follows
up
with
question
when
using
the
cube
admin
cluster
running
authorization
mode
equals
always
allow,
is
it
possible
to
add
additional
nodes
using
cube
admin
join
in
the
past.
I've
tried
this,
but
I
always
hangs
during
the
join
stage.
Sorry,
if
this
quest
is
vague,.
B
D
A
A
B
A
D
We're
just
making
some
updates
to
which
is
useful
for
post
cluster
stand-up
validation.
So
if
you're
doing
an
upgrade
or
if
you're
doing
a
initial
cluster
stand-up,
so
don't
lose
a
tool
we
can
use
or
you
can
use
for
validating
your
clusters.
All
good,
there's,
there's
a
web
page
for
that.
I
can
also
make
that
that
you
cannot
but
we're
making
updates
to
it
to
basically
have
a
better
format
as
well.
As
you
know,
getting
ready
for
the
one
eight
release
cycle
too
and
we're
doing
validation
there.
Mm-Hmm.
B
A
D
D
A
B
A
C
A
A
It
feels
like
the
internet
gods
are
with
us,
because
these
stream
health
has
definitely
been
rock
solid,
the
past
20
minutes
or
so
because
our
view,
our
viewer
rate,
keeps
going
up.
So
that's
good,
yet
no
worries,
thanks
to
you
thanks
for
coming
sorry,
we
couldn't
answer
your
question,
but
we'll
get
to
it
as
soon
as
we
we
get
a
dashboard
person,
probably
next
time,
so
keep
an
eye
out
for
that.
I
guess.
A
And
the
cube
admin
upgrade
is
now
behind
a
feature
flag,
so
I
can
write
right.
So
for
those
of
you
don't
know,
you
can
now
do
an
upgrade
with
cube
admin.
So
I
believe
it's
cube.
Admin,
Alf
space
alpha
and
then
type
help.
I
think
it's
alpha
upgrade,
but
I
don't
want
to
I.
Don't
want
to
go
on
the
record
to
say
that,
but
the
the
flags,
the
command
line,
flags
and
everything
for
cube
admin
are
complete,
so
I
always
just
look
them
up
on
the
CLI.
C
D
A
Really
interesting,
I've
done
the
whole
server
in
the
back
room,
one
node
of
kubernetes
via
cube
admin,
to
kind
of
learn
how
to
throw
things
on
it
and
then
every
week
or
so
I
tear
it
down
and
I
and
I
rebuild
it
just
so
I
can,
you
know,
feel
comfortable
rebuilding
it,
so
we're
starting
to
run
out
of
time.
So
maybe
one
or
two
more
questions
there.
If.
A
And
then
we
will
be
done
and
then
anyone
have
any
ideas
for
next
month's
office
hours.
We
could
do
one
at
the
end
of
this
month
or
the
beginning
of
October
I'm,
starting
to
lean
towards
doing
a
u.s.
West
Coast
friendly
one
so
that
we
can
snag
some
of
the
West
Coast
volunteers
that
would
like
to
help
during
office
hours.
B
A
A
A
A
A
All
right-
and
it
looks
like
we're-
not
gonna-
have
any
more
questions
for
for
the
rest
of
the
session.
So
thanks
to
our
panel,
Ilya,
Andy
and
Tim
for
stopping
by
I
will
go
ahead
and
post
the
next
date
next
month.
That's
at
some
point
the
more
we
stream,
the
more
data
we
get
about,
how
reliable
the
stream
is.
So
sorry
about
the
issue
there
15
minutes
in,
but
my
YouTube
analytics
are
not
filled
with
a
whole
bunch
of
information.
A
I
need
to
dig
through
so
the
more
information
you
give
us
about
the
quality
of
the
audio,
the
quality
of
the
video,
the
better
decisions
we
can
make
as
far
as
encoding
and
whatnot.
So
we
really
appreciate
that
so
with
that.
Thank
you
very
much.
We'll
go
ahead
and
stop
the
stream
now
and
we'll
see
everybody
in
next
month.
Thanks.