►
From YouTube: Kubernetes Community Office Hours 20160630 Vishnu Kannan
Description
Office Hours
-----------------
Kubernetes developers hold office hours over VC where we discuss questions raised on slack or by attendees about how Kubernetes works.
In this video of Office Hours Vishnu Kannan answers --
suraj -- where do i get started?
From googlecontainers - Upon creating a cluster it creates a bunch of default pods under the "kube-system" namespace, (fluentd-cloud-logging-gke, kube-dns, etc) which seem to automatically re-spawn when deleted. Is there a way to limit the CPU / Memory resources for these?
sarahnovotny - What is the sig node team working on right now?
mahipal -- 1.3 questions - can the kubernetes dashboard be pushed into dockerhub? if not dockerhub, what is the preferred process?
vladimirvivien - What is the best way to get beyond git rebase content conflicts for *generated.go files?
colemickens - Is there any guidance around on how to choose a log level for certain types of logging messages?
loren - how are capabilities specified in a pod manifest? i need to add NET_ADMIN to a container. --
keithballdotnet - How can I ping from one pod to another using the pod name of the other pod? Is it possible? -- not possible
Mike Brown - What do you need to modify in the makefile in order to support passing the debug parameters into the go compiler? Is the makefile the right place?
nalum - Hey all, what is the difference between the readiness and liveness probes?
akapap - Is there a way to mount only one file? Sth like volumeMounts and only put one file in container space?
shivania - I want to inspect the kubernetes code since I am planning to implement what if any tools that can help me with inspecting the code.
danbopes - I'm going to be creating a service that's going to make outgoing connections to external endpoints (Basically a large array). What's the best way to handle which item in the cluster has to connect to which service?
stonith - when using daemonsets, how do I configure other pods/services to connect to the daemonset running on the local node?
JustinSB - Can a DaemonSet run on the master? This comes up with e.g. NewRelic running as a DaemonSet
A
A
C
Yeah
I
started
with
contributing
to
documentation
that
I've
known
works
like
outdated
needs.
Some
changes
also
I'm
looking
for
delivering
to
code,
but
I
have
not
seen
anything
that
I
mean
this.
Most
of
the
things
are
not
easy
to
get
started
with,
so
I
am
looking
for
something
that
I
can
start
so.
Yeah
I
have
seen
Vishnu
sending
a
lot
of
tias
a
huge
PS,
a--,
specifically
specific,
and
I
would
like
to
have
it's
like
a
direction
or
something
so.
B
C
I've
been
looking
into
networking
and
node
sake,
and
not
much
but
yeah
I
think
that
leaves
like
you,
you,
you
feel
like
it's.
It's
something
good
fodder
at
UV
that
could
be
good.
I
work,
I
work
at
red
hat,
but
contributing
to
Copernicus
is
not
part
of
my
day
job,
but
I
I
would
like
to
do
that.
It's
open
source
so.
B
That's
a
good
thing
yeah,
so,
like
my
jungle,
advice
would
be
that
we
mark
a
lot
of
issues
with
an
help
want
attack.
I
had
pointed
label
to
be
precise,
so
those
are
the
ones
to
start
with,
but
I'll
be
frank
that
not
all
the
issues
there
have
all
the
relevant
information
for
a
new
person
or
new
contributor
to
get
started.
So
I'd
say
like
take
taking
of
those
help
on
with
use
commands
that
you
think
you
can.
B
You
can
sort
of
understand
what
is
required
and,
if
not
like
being
the
author
or
like
just
just
like
bring
it
up
on
our
slack
channel
and
ask
us
like
what
should
be
done
for
this
issue.
Like
tell
me
what
needs
to
be
done
and
if
that's
not
clear
and
one
of
us
can
help,
you
figure
out
what
needs
to
be
done.
You.
B
A
Siraj,
there's
also,
I
think
you
said
cig
network,
which
is
sort
of
a
magical
magical
word
for
me.
Getting
involved
with
the
special
interest
groups
for
your
interests
in
relating
to
larger,
larger
chunks
of
the
project
is
another
great
way
to
see.
What's
going
on
and
ask
for
help,
that's
more
focused
and
more
specific.
Are
you
already
a
go
coder.
A
A
A
A
Google
containers
has
been
sort
of
both,
but
the
idea
longer
term
is
for
that
to
be
more
gke
focus
in
cooper,
Nettie's
users
to
be
more
agnostic,
run
it
on
metal,
run
it
on
awss
questions
there,
but
as
far
as
the
mailing
lists
go,
those
are
those
are
the
two
that
I
know.
Oh
how
about
Cooper
Nettie's
novices,
although
that
isn't
necessarily
dev
focused,
but
if
we
don't
have,
we
don't
have
people
here,
asking
dev
questions.
We
can
also
go
to
committees,
novice
and
ask
questions
or
yeah.
B
B
B
I
created
a
cluster
which
is
default
default
pods
in
the
queue
system
namespace,
it
seems
to
automatically
respawn
and
deleted.
Is
there
a
way
to
limit
the
cpu
and
memory
resources
for
these?
Well,
when
you
set
up
a
cluster,
a
communist
cluster
has
requires
a
bunch
of
components
and
we
try
to
run
most
of
the
cameras
components
inside
covers
itself,
so
we
have
heaps,
which
is
our
metrics
by
climate,
ransom
pod.
We
have
our
dns
service
running
as
a
pod.
B
Let
me
have
a
long
serviced
and
running
as
a
pod,
and
all
these
entities
are
like
pre-configured
and
depending
on
how
you
set
up
the
cluster,
it
might
be
easier
difficult
to
update
their
configurations
on
so
for
this
specific
question,
it's
not
straight
for
right.
Now,
it's
not
super
straightforward.
To
change
these
configurations.
It
is
possible
to
update
configuration
of
a
running
pod,
but
that's
not
going
to.
Let's
not
go
to
be
persisted.
What
happens
is
like
on
the
cover
is
master.
We
one
of
the
deployment
process.
B
We
move
a
bunch
of
configuration
files
onto
the
master,
and
that
is
the
process
running
on
the
master
that
ensures
that
required.
Kunis
cluster
components
in
the
cube
system
namespace
are
running
under
there.
Besides
state
matches
the
state,
that's
that
in
these
files.
Well,
unless
you
can
go
and
update
the
files
inside
of
the
master,
if
it's
not
possible
to
persist
your
changes.
B
Think
I
think
in
the
future.
We
want
to
make
it
easy
for
people
to
change
these
configurations.
We
don't
we
don't
want
like
deal
with
oil
salt
and
if
you
want
everything
to
be
deterministic
not
like
people
know
what
exactly
is
happening
inside
the
system
and
then
they
can
go
and
update
it
right
away.
But
we
want
that
yet
and
by
extending
this
on
on
clusters,
services
like
GK,
since
the
whole
system
is
managed
for
the
end
user.
We
don't
expect
the
end
users
to
change
and
figure
it
Lester
components.
B
B
B
You
rather
want
to
keep
your
pods
alive,
so
we
have
you're
adding
a
bunch
of
features
which
will
like
killbots,
like
your
priority
pods
if
high
priority
pods
are
taking
it
more
discreet
sources
or
like
go
and
remove
existing
images.
If
you
have
pulled
in
a
lot
of
images
on
there
causing
this
pressure,
so
it's
like
making
sure
that
the
node
stays
alive
as
much
as
possible
without
causing
too
much
use.
The
destruction
applies
to
memory
and
whole
bunch
of
other
resources,
and
we
also
trying
to
improve
security.
B
So
if
you
are
making
a
lot
of
effort
there
to
make
sure
that,
like
Chris
can
scale
to
alternate,
run
things
and
it's
super
easy
to
build
features
and
like
you
don't
block
on
each
other
and
there's
like
innovation,
happen
everywhere
and
a
lot
of
other
things
happening.
It
was
like
a
lot
of
project
maintenance
work
happening
to
like
improving
testing
hello,
yeah
I
mean
I
can
keep
going
on
about
that.
Well,.
D
A
A
B
D
B
The
reason
why
we
haven't
gone
with
darker
has
is
that,
like
your
work
for
a
lot
of
users,
that
performance
at
dalka
hub
has
been
very
poor,
so
just
to
make
sure
that
it
can
give
the
best
performance
possible
and
you've
chosen
to
use
GC
r
dot
io,
which
is
a
good
registry.
So
from
a
user
perspective
it
shouldn't
really
matter,
except
that
you're
typing
a
few
more
characters
are
mentioned
by
mentioning,
but
otherwise
that's
actually
be
the
same.
B
So
on
I
can
I
can
definitely
bring
this
up
with
a
release
thing
or
maybe
even
you
can
go
and
file
ahead
file.
An
issue
against
our
release.
Kunis,
release,
rapoo
and
ask
like,
is
this
possible,
but
my
understanding
is
that
performance
has
been
historically
very
bad
and
that's
the
reason
why
we
don't
push
it
or
her
by
default.
D
B
D
It
seems
the
intention
is
to
do
it
for
private
thing,
like
its
interval
container
management
system
like
light
cycle
management.
So
that's
what
we
are
trying
to
do.
I,
most
probably
we
want
to
hire
some
summer
tabs
and
things
like
that.
Instead
of
pods,
we
want
to
make
some
changes
with
tabs
and
excited
that's
what
we
are
trying.
Okay,.
B
So
for
you
to
iterate,
there
are
two
ways
right.
Like
I
said
you
can
build
your
own
dashboard
image
that
you
can
do.
We
like
the
dockerfile
remains
the
same.
What
name
you
give
to?
It
is
up
to
you.
Our
release
process
uses
the
GCR
dot
io
registry,
and
it
gives
a
name
that
signifies
that
the
image
is
stored,
energy
at
I/o,
but
you
can
like
totally
replace
that
with
water,
a
name
you
want
and
most
likely
it's
not
going
to
push
the
images
for
you
which
is
going
to
build
them.
B
So
you
can
also
do
like
a
darker
tag
and
you
can
change
the
image
name,
that
the
release
process
generates
into
an
image
name
that
that
would
suit
your
purpose.
Like,
for
example,
you
could
change
from
GCR
on
io
/
google
underscore
containers
/
do
doc,
your
your
docker
hub
account
/
dashboard
on,
and
that
should
be
possible
through
darker,
and
then
you
can
do
an
awkward
bush.
D
A
We
had
someone
else
join
us.
This
is
so
exciting.
We
have
mhb
who
joined
us.
Who
may
have
a
question
so
I'm,
hoping
that
that
person
will
come
off,
mute
and
tell
us
if
there
are
questions
or
maybe
they're
just
lurking
all
right.
Well,
I
have
one
more
question
that
I
pulled
out
well,
one
more
right
now
and
then
I'll
keep
looking
while
you
answer
it.
The
question
is
hello
there.
What
is
the
best
way
to
get
beyond
git
rebase
content
conflicts
for
the
dot
the
star
generated,
go
files.
B
Whenever
you
build
the
build
systems,
are
will
be
set
up
in
such
a
way
that
all
the
generated
code
would
would
just
be
automatically
done
in
the
background
for
you
and
testing
just
works,
and
whenever
you
have
to
do
a
comet,
your
generated
files
are
not
picked
them
into
your
color,
so
we
are
working
on
tooling
you're,
just
trying
to
find
the
right
kind
of
people.
People
who
know
make
to
make
sure
they
can
validate
what
Tim
has
done,
and
it's
actually
hard
find.
B
So
I
apologize
for
the
fact
that
you
have
to
deal
with
this
right
now,
but
as
far
as
I
can
think
of
other
than
rebase
eyes
is
no
other
go
on.
It
should
have
really
mattered
to
you
as
long
as
you're,
not
like
renaming
a
field
or
deleting
a
fee
in
API
more,
it's
just
a
simple
rebates.
When
you
try
to
do
a
rename
or
delete
it
gets
more
difficult.
A
A
A
Alright,
so
I
have
another
one
while
you're
looking
for
that
link,
I
will
go
pull
up.
Another
question,
which
is:
is
there
any
guidance
around
how
to
choose
a
log
level
for
certain
types
of
logging
messages?
So
I'm
presuming
this
is
when
people
are
developing
cooper,
Nettie's
and
choosing
what
log
level
to
put
messages
in
within
the
cooper
Nettie's
code.
Do
we
have
boundaries
and
guidelines
for
things
no.
B
Not
really
I
think
there's
like
some
sort
of
an
answer
guideline
that
is
sort
of
there
and
most
of
the
developers
heads.
But
it's
a
sure.
As
you
know,
it's
not
yet
been
standardized,
usually
like
lot.
Logging
level
changes
based
on
how
important
it
is
and
what
purposes
itself
on
if,
if
a
log
is
only
going
to
be
for
debugging,
and
it's
going
to
like
increase
your
logging
space,
a
lot
we
usually
like,
tend
to
push
it
to
a
very
high
number
like
we
even
set
it
up
to.
B
Like
logging
level
stem,
we
have
numeric
levels
between
0
and
10.
I
like
0
means
you
always
log,
and
then
all
other
logging
levels
are
like
lower
priority,
and
then
you
can.
You
can
choose
a
logging
level
for
each
binary
that
you
run
on
so
so
typically
going
from
on
a
priority.
If
you
want
a
log
to
always
be
there,
because
it
helps
you
debug
most
of
the
issues,
then
you
just
set
it
to
the
default
0
logging
level
or
if
you
think
this,
this
log
is
useful
but
doesn't
have
to
be
all
the
time.
B
B
People
go
all
the
way
up
to
10,
just
because
it's
like
we
don't
want
people
to
accidentally
turn
on
these
logs,
because
they
will
occupy
a
lot
of
disk
space,
so
we
sent
to
10
so
that
even
if
they
keep
going
higher,
they
never
reach
that
10-point
arm,
but
oftentimes
it's
off.
If
you
set
the
logging
level,
2
10
in
many
of
her
components,
you're
actually
stepping
through
the
code,
you
actually
know
what
the
binary
is
exactly
doing,
because
you
have
so
much
blogs.
Oh
so
I'd
say
like
it's
an
odd
it's!
B
A
Cool,
so
that's
the
sort
of
the
broad
guidelines,
but
we
don't
have
them
written
down.
Sadly,
we'll
get
there,
it's
a
young,
it's
a
young
project
will
get
there.
Okay,
so
I
have
another
one
for
you.
How
our
capabilities
specified
in
a
pod
manifest
I
need
to
add
net
admin
to
a
container
I'm,
trying
to
reverse
engineer
how
to
do
this
from
the
code.
But
I
haven't
figured
out
the
magic
incantation.
A
A
A
A
B
B
You
can
also
specify
I
just
get
to
a
security
complex,
so
you
can
also
specify
capabilities
here
so
spacek
your
security
context.
You'll,
have
you
will
have
a
type
inside
a
container
coaches
security
context
and
in
there
you
can
specify
capabilities
and
let's
look
at
what
capabilities
are
so
capabilities
are
basically,
you
can
add
and
remove
capabilities,
and
they
basically
match
o
6
+
tax
for
capabilities.
B
Here,
alright
you're
specifying
your
containers
as
part
of
a
bawd.
So
it's
actually
a
replication
controller
I'm.
You
can't
specify
careful
duties
here,
so
you
have
a
key
for
really
section
under
security
context
and
then
you're
saying
I
want
to
add
IPC
lock
so
yeah,
you've,
carefully
system
and
in.
B
A
E
E
I'm
putting
you
know
talking
about
the
developer
experience
how
you
can
I
can
do
set
up
and
then
because
really
the
docs
are
a
little
messed
up
for
you
know:
developer
md,
just
a
heads
up
there
and
I'd
starting
to
work
on
that
I'll
play
push
a
bunch
of
pr's
and
to
work
on
the
developer.
Docs,
like
weddings,
where
we're
looking
at
yesterday,
was
getting
debug
working
and
need
to
modify
the
make
file
so
that
it
supports
you
know
passing
parameters
into
to
the
goat
compiler.
I'm
just
making
sure
that
make
file
is
the.
E
B
E
E
E
E
B
E
E
F
A
A
B
So
I'm
just
going
to
pretend
I
didn't
see
that
yeah,
so
basically
about
players
like
some
readiness
is
for
figuring
out
whether
your
application
is
as
logically
come
up
on.
So
actually
sorry
that
I've
it
on
readiness.
This
is
to
convey
the
fact
that
your
application
is
ready
to
serve
traffic
and
liveness
is
just
to
check
check
that
your
application
is
logically
up
and
running.
B
So
you,
you
would
probably
want
to
restart
your
application
whenever
it's
failing,
liveness
probes,
you
might
be
very
stringent
about
liveness,
because
your
application
gets
stuck
for
any
reason
we
already
started,
but
for
readiness
you
might
actually
give
it
more
time,
because
maybe
it's
trying
to
talk
to
a
back-end
database
or
talking
to
its
other
peer
is
slow.
There,
like
two
different
semantics
and
generally,
whenever
a
pod
passes
all
its
readiness
strobes.
That's
when
the
power
is
not
as
ready
and
that's
when
it
can
start
serving
traffic.
So
by
default.
A
A
B
B
Don't
think
we
restrict
I,
don't
think
we
restrict
whether
the
path
has
to
be
a
file
or
directory.
So
based
on
a
comment
here,
it
represents
either
ghost
file
or
directory.
So
I
don't
think
the
system
tries
to
prevent
people
from
mounting
and
just
files
and
subdirectories.
So
if,
if
this
person
was
this
question
is
another
of
host
Pat
volumes,
then
they
should
probably
be
informed
about
that
we
can
use
in
hospital.
You,
man,
Montana
file
or
directory
yeah.
A
So
I
will
ask
you
that
one
of
the
same
questions
that
I
that
I
put
a
to
Phillip
on
Tuesday,
because
I
think
it's
a
good
one
to
get
from
everybody
if
I
want
to
inspect
the
kubernetes
code
since
I'm
planning
to
do
something
and
he
wants
to
know
what
are
good
tools
to
help
with
that.
So
what
is
what
is
your
tool
chain
and
how
does
how
does
that
fit,
and
what
would
you
recommend
for
people
so.
B
B
So
I'm
good,
so
I'm
going
to
show
you
two
directories
which
you
should
be
aware
of.
So
we
have.
We
have
a
top
level
directory
called
come
on,
and
here
we
have
all
the
rappers
that
build
binaries.
So
usually
we
try
to
split
code
and
such
a
way
that
we
have
reusable
libraries
and
then
we
compose
these
libraries
to
binaries.
So
everything
inside
of
command
is
where
we
be
with
flags
and
like
all
the
aspects
of
building
a
binary
on
and
libraries
a
store
inside
package.
B
So
everything
in
here
is
the
library
and
like
most
of
these
correspond
to
one
or
more
components.
So,
let's
just
go
over
like
one
or
two
of
them
to
give
an
idea
on
how
you
navigate.
Since
we
already
mentioned
volumes,
volumes
is
very
straightforward.
So
if
you
want
to
say
hack
in
a
storage
space,
you
go
in
two
volumes
and
most
of
the
volume
or
muscle
storage
work
is
done
inside
of
the
wall
gets
directory.
So
here
you
have
a
whole
bunch
of
different
types
of
volumes
and
all
these
volumes
are
implementing
certain
interfaces.
B
You
usually
go
into
a
package
and
package
name
that
go.
You
have
the
public
interfaces
that
belong
to
that
to
that
package,
and
so
here
you
have
a
few
volume
methods
and
all
of
all
the
sub
valium
pipes
that
little
bit
inside
the
style
tree
are
going
to
implement
those
public
methods.
So
we
have
two
main
public
methods
for,
or
public
interfaces
for
volumes
which
is
and
volumes.
B
So,
looking
at
these
two
files
with
value
what
sort
of
interfaces
you
should
go
for
and
like,
if
you
just
keep
tracking,
if
you
just
do
a
search
for
the
interfaces
here,
then
you
will
know
like
where
these
volume
plugins
are
being
initialized
and
and
then
you
can
like
sort
of
draw
draw
a
line
between
how
volumes
interacts
with
the
rest
of
the
writers,
I
think,
there's
also
a
persistent
volume
on
part
to
this.
It
should
also
be
a
part
of
volumes
on
that
deals
with
the
cluster
level
orchestras
in
four
volumes.
B
So
we
can
usually
start
with
API
server,
don't
go
and
that's
usually
your
starting
point
like
package
and
pocket
/
package
name
dot
go
is
usually
the
best
starting
point
and
but
then
API
server
also
includes
a
whole
bunch
of
dependencies
like
API
is
a
PA
machinery,
but
you
should
like
find
the
file
a
fine,
the
first
method
that
that
the
binary
in
works
for
you
to
like,
or
you
like,
draw
an
overall
architectural
diagrams.
So
going
back
I
said
initially
that
you
will
start
with
a
command
folder.
B
So
let's,
let's
look
at
the
API
server
here,
so
we
have
QV
case
over
here
and
anyhow
eps
ever
not
go
I
mean
here
we
have
a
mean
so
I
would
say,
like
you
would
start
with
this
and
then
you'll
start
like
going
through
each
and
every
method.
So
we
have
appt
daughter
on
here
which
in
turn
gets
into
app
and
then
I
think
we
have
a
service
or
go
here,
and
then
you
have
a
run
in
here.
B
B
B
No
I
think
a
good
space
so
yeah,
so
we
have
like
primitive
tools
that
will
help
you
navigate
code.
But,
like
you
want
to
do
tags
referencing,
then
you
need
to
use
like
eclipse
or
like
Oracle
one
because
it
what
that,
whether
you
have
gotten
to
a
point
that
all
the
ecosystem
tools,
our
own
goal
line,
will
work
at
the
Communist
Jacko
but
like
each
person
has
to
poison
to
have
tonight
cool.
A
Well,
that
is
that
is
helpful,
and
it
was
great
to
see
the
how
to
think
about
the
repo
as
well.
So
I've
got
another
question.
If
nobody
who
is
lurking
with
us
has
further
questions,
nobody
ok
feel
to
jump
in.
If
you
do
have
questions
dan
bobe's
asks
I'm
going
to
be
creating
a
service,
that's
going
to
be
making
only
outcome,
outbound
connections
to
external
endpoints,
basically,
a
large
array.
What's
the
best
way
to
handle
which
item
in
the
cluster
has
to
connect
to
which
service,
which.
B
B
B
Looks
like
we
have
a
network
policy,
but
isolation
here.
It
says
so
here
it's
like
trying
to
set
up
isolation
between
different
services
and
pots.
So
let
me
see
if
I
can
glean
what's
in
here,
which
groups
of
frauds
and
define
how
the
spots
should
be
allowed
to
communicate
with
each
other
mentation
of
the
policy.
B
That's
not
included
increased
side
base
wallasey,
which
is
actively
so
I.
Guess,
is
probably
what
the
question
was
started
hitting
on
it's
not
yet
being
solved,
but
I'd
say
like
my
cooking
sig
is
a
place
to
go
and
ask
this
question
specifically
to
answer
the
first
part
of
the
question.
It
is
possible
to
set
up
a
commuter
service
but
being
backed
by
applications
that
are
running
outside
covers,
so
we
have.
We
have
a
concept
called
as
an
external
service
on
and
excellent
services
can
be
backed
by
entities
that
are
not
inside
the
cluster.
B
B
So
without
wasting
a
lot
of
time
trying
mind
this
shot,
answer
is
like
you
can
create
any
cell
as
you
want,
and
you
need
not
specific
a
label
selector
for
that
service.
So
whenever
you
create
a
service,
you
specified
label
selector
for
one
commuters
will
try
to
not
find
end
points
for
a
given
service
and
then
what
you
can
do
is
like
each
services
will
be
back
by
an
endpoint
output,
so
you
can
create
your
own
endpoints
right.
B
Show
your
endpoints
object
here,
so
you
can
create
your
own
end
points
so
and
instead
of
those
endpoints,
you
can
put
your
own
IP
addresses
and
ports
for
the
applications
that
are
running
all-time
record
is
faster
and
then
what
colors
will
do
is
like.
Whenever
you
try
to
reach
service
that
you
created
XL
servers
you
created,
it
will
use
the
IP
addresses
that
you
had
provider
rope
traffic
to
those
entities.
B
And
I
think
that's
the
new
feature
that,
like
that
they're
coming
outside
federated
services,
is
going
to
make
this
easier,
but
I
think
that's
a
long
topic
by
itself,
because
you
need
to
first
miscers
watch.
Related
services
are
before
analysis
questions.
So
maybe
you
not
do
it
here
in
the
next
30
minutes
we
have
30.
A
A
A
A
B
So,
typically,
applications
which
require
load,
balancing
wouldn't
be
running
at
the
demon
set
on
the
demon
sets
are
usually
meant
for
class
or
administrators
run
apps
like
running
a
cron
task.
Are
you
running
a
monitoring
or
logging
agent?
Are
you
writing
some
code
career
services
on
it's
meant
for
that?
So
it's
not
meant
for
slowing
traffic,
although
you
could
use
it
like.
B
If
you
want
look,
if
you
wanna,
have
your
database
and
you
wanted
to
run
only
on
these
specific
nodes
that
it's
used
even
set
to
achieve
that,
but
I
don't
think
we
have
I,
don't
think
we
have
any
plans
for
adding
more
support
for
like
latency
based
routing
inside
our
local
proxy,
so
shocked,
I.
Think
Sean
answer
is
like
no
for
now
and
for
the
foreseeable
future.
Mm-Hmm.
A
B
That's
interesting,
I'm,
trying
to
think
of
like
other
way
achieve
this.
If
we
like
one
way
would
be
like
if,
if
there's
no
local
demons,
a
demon
sets
allowed
to
expose
themselves
on
on
the
North
Port
nor
I
p.m.
port,
then
maybe
we
could
inject
the
node
IP
into
those
pods
and
they
can
try
to
access
through
that.
Oh.
B
Yeah,
I
don't,
I
don't
think
this
feature
has
been
explored,
so
maybe
the
best
place
to
start
would
be
the
final
issue
and
have
a
lot
of
Plus
Ones
against
it.
All
right
like
release
and
github
terminology
and
might
be
a
good
place
to
start
because
a
lot
of
people
want.
But
then
maybe
we
can
start
looking
investigate.
A
Yeah,
it's
the
I
want
to
do
this.
How
can
it
be
done
in
this
case
yeah
interesting,
all
right,
I'm,
running
low
on
issue
well,
issues,
I,
have
plenty
of
issues:
I'm,
printing
on
questions
out
of
the
chat
channels,
Mike
or
Doug
or
Alex
brand
is
a
new
person.
Do
any
of
you
have
any
questions
further
for
fish
or
Justin.
We've
got
more
people
in
here.
B
I,
don't
think
from
a
scheduling
perspective
you
restrict
like
fundamentally,
the
system
does
not
restrict
bots
from
being
placed
on
a
master,
be
from
a
demon
said
or
not
from
some
other
controller,
but
typically
the
way
we
set
up
clashes.
We
make
me
mark
the
Masters
unscalable,
so
generally
they
don't
land
on
the
master.
But
that
said,
if
the,
if
your
label,
selector
policy
for
a
demon
said
all
saying,
was
a
nostril,
we
shouldn't
be
restricting
at
0.
B
B
Cluster
doesn't
seem
to
be
responding,
but
I
was
just
going
to
try
to
show
that
you
can
do
and
edit
on
the
master
config
object.
So
you
can
do
like
a
cute
cuddle,
edit
notes,
/
Chris
hyphen
master,
and
then
you
will
have
an
unscalable
variable
in
there
and
then
you
can
just
it
will
be
marked
as
true.
So
you
can
you
set
it
to
false
and
said
save
I
found
that
point
onwards.
Your
master
should
be
scalable
or
your
parts
can
be
scheduled
on
the
master.
B
B
A
Yes,
thank
you.
It
is
my
birthday.
Thank
you.
Everybody
keeps
saying
thank
you
happy
birthday
in
the
chest
all
right.
Well,
let's
call
this
successful
and
any
one
of
you
who
have
been
lurking
and
wants
to
be
one
of
these
hot
seat
office
hours,
questions
on
the
fly
people
please
tell
me
because
I
want
to
keep
doing
these.
They
seem
to
have
been
fun
and
from
what
I
can
tell
several
of
you
have
gotten
interesting
nuggets
out
of
this.
So
fantastic
thanks.
A
It'll
be
fun
and
we'll
put
it
up
on
YouTube
and
then
we'll
do
them
again
and
we'll
we'll
get
more
of
them.
And
then
you
know
they'll
just
be
a
fun
thing.
That's
sort
of
I,
don't
know
video
podcast
e
with
questions
it
is
sort
of
like
a
radio
call-in
show
is
what
it
feels
like
to
me
her
colon
and
ask
our
experts
alright,
have
a
great
day
guys.
I
will
talk
to
you
all
another
day
on
another
video
chat,
I'm
sure.