►
From YouTube: Kubernetes Kops Office Hours 20181026
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
October
26th.
This
is
cops
office
hours.
We
have
one
or
two
items
on
the
agenda,
but
otherwise
a
fairly
light
agenda.
So
this
may
be
a
very
short
meeting,
but
do
add
any
other
items
the
agenda
so
that
we
can
be
sure
to
get
to
that.
Otherwise
we
can
just
dig
right
into
it.
A
The
thing
which
I'm
rehab,
that
is
there
is
a
company
called
platform.
9
proposed
a
similar
project
called
Etsy
da
DM,
which
is
more
of
a
CLI
driven
approach,
and
the
approach
that
we're
proposing
is
to
basically
merge
those
two
projects
so
that
you'll
have
a
sort
of
CLI
option
for
like
when
things
go
wrong
or
if
you
want
to
do
some
power
user
stuff,
whatever
that
might
be
an
ED
CD
manager
will
effectively
drive
the
CLI
for
you,
so
will
be
an
automation
there.
A
On
top
of
that,
so
you're
able
be
able
to
like
go
in,
and
you
know
introspect
it
and
in
theory
a
CD
manager
can
be
more
willing
to
give
up
because
they're
like
when
things
are
unsafe,
because
there
is
something
you
as
administrator
could
do.
Hopefully
we
won't
actually
need
that,
but
particularly
on
bare
metal,
it's
certainly
something
that
it
can
be
much
harder
to
achieve
safety
so
that
that
cap
is
proceeding
and
I
think
it's
on
track
and
we
should
be
able
to
get
at
CD
manager
into
a
community
sync
project
and
then
Rodrigo.
A
The
next
item
looks
like:
when
are
we
planting
111,
Peter
and
112?
Alpha
I?
Think
I
think
once
we
get
one
eleven
beta
out,
we
should
do
one
12
alpha
pretty
pretty
soon
after,
if
not
at
the
same
time,
I
think
I
can't
remember
exactly
what
we
said
in
terms
of
when
we
default
at
CD
manager
on
or
off
so
that
might
be
the
one
blocker
for
a
112
alpha
to
make
sure
that
we
are
comfortable
with
that.
But
it
is
only
an
alpha
in
terms
of
111.
A
Beta
I
haven't
seen
any
major
issues,
yet
the
only
thing
that
I'm
aware
of
is
I
think
there
was
something
to
do
with
s3
permissions,
where
I
made
a
I
made
a
change
risky
to
optimize
it,
and
it
looks
like
we
might
actually
want
to
backward
that
optimization
because
it
also
fixes
some
security
requirements
like,
in
other
words
the
permissions
needed.
So
but
that's
not
that's
not
that
doesn't
really
impact.
A
A
B
B
A
Really
one
point:
thank
you
so
much
Elena
for
like
driving
it
that
you
were
really
the
person
that
kept
the
pressure
on
and
I
think
we
I
think
there
was
like
I
can't
remove
a
person's
name
that
actually
changed
the
PR,
but
he
or
she
did
a
lot
of
revisions
on
it
to
make
it
sort
of
more
merchant,
and
so
they
they
did
a
great
job
as
well
there.
So
thank
you
so.
A
A
A
A
B
C
A
A
I,
don't
know
if
that
person
has
a
name,
but
do
you
wanna?
Do
you
want
to
do
me
a
like
an
ED
CD
note,
or
do
you
mean
like
a
so
that's
it
Lister?
If
you
lose
a
humanities,
note
like
a
non
master,
it
doesn't
matter
currently
all
the
Masters
in
cops
run
at
CD.
If
you
have
one
master
and
you
lose
it
then
you're
down.
If
you
have
three
masters
and
a
3-node
at
CD
cluster,
if
you
lose
one
note
at
CD,
we'll
continue
running
the
it'll
be
fine,
it
still
has
quorum.
A
A
If,
yes,
if
two
nodes
of
us,
the
cluster
does
go
down,
it's
quite
specific,
though,
because
what
we
have
in
the
way
we
run
it
in
cops,
and
this
is
sort
of
SUV
manager
will
will
continue
to
support.
This
is
if
you,
if
a
this,
is
only
if
the
instance
is
sort
of
still
running,
but
if
you
actually
shut
down
or
terminate
cert.
If
you
terminate
the
instance
they're
running
in
auto-scaling
groups,
a
new
instance
will
spin
up,
it
will
remount
the
EBS
volume
and
will
rejoin
the
cluster
automatically.
A
A
But
if
you
lose
one,
if
you
only
lose
one,
so
if
you
lose
one
zone
or
if
you
lose,
if
you
lose
an
entire
zone,
you
can
still
keep
running
and
yet
the
tricky
the
tricky
cases
are
when
the
like,
for
example,
like
suppose
you
suppose
you
manage
to
like,
corrupts
the
disc
on
one
of
the
one
of
the
SUV
nodes,
that's
the
tricky
scenario
where,
like
xev,
would
still
be
running
the
API
server
on
that
sucker
note
wouldn't
be
able
to
write
to
it.
I
don't
know
that
we
handle
that
brilliant
Nate.
C
A
Think
yeah,
I
think
so,
I
think
down
the
the
the
you
have
to
be
very
careful
about
the
finding
whether
you
mean
that
a
note
like
is
sort
of
like
what
do
you
mean
by
a
note
being
down
and
what
I
mean
like
by
no
coming
back
like
if
you
just
shut
down
and
terminate
all
your
notes.
That'll
come
back
no
problem,
it's
the
the
challenging
scenario
is
when
you
have
a
note
that
gets
the
volume
back,
but
for
whatever
reason,
Etsy
still
can't
start.
A
Okay
yeah
for
the
simple
situation
where
you
just
terminate
a
note,
it
should
work
absolutely
great.
You
can
actually
terminate
if
you
have
a
test
cluster
or
worry,
or
a
cluster
you're
feeling
brave
on.
You
should
shut
shut
down
everything
and
it
will
all
just
come
back.
You
should
omit
if
you
just
shut
down
all
your
masters.
What
happens
is
all
the
nodes?
A
A
Anything
that
talks
to
API
server
will
not
be
able
to
talk
to
API
server
during
that
time,
but
the
sort
of
traditional
kubernetes
design
is
that
components
keep
running
even
if
they
can't
reach
it
the
iced
over,
so
that
the
cubelet,
for
example,
like
won't,
shut
down
your
pods.
If
it
can't
talk
to
api
server,
it
will
just
like
leave
them
running
as
they
are,
so
it
works.
A
Fine
for
you
know,
hours
or
maybe
days
not
much
longer
than
that,
because
everything
sort
of
like
gradually
keeps
going
with
a
sort
of
state
of
the
world
as
it
were
as
it
was.
But
you
know,
as
things
crash,
you
don't
wanna,
do
it
for
weeks
because,
as
things
crash
like
they
won't
be
recovered
correctly
and
yes,
you
can.
You
can
lose
all
three
canoes
three
of
three
yes
and
that's
a
deal
automatic
recover,
because
the
data
is
still
on
EBS
volumes
and
we
remount
the
EBS
volumes
automatically.
C
A
Not
yet,
but
it's
talking
about
it,
which
is
why
I
asked
yeah,
it's
not
it's
not
in
there.
Yet
all
you
have
to
do.
Is
you
have
to
drop
a
file
into
s3
to
tell
it
to
do
it,
but
I
have
not
built
it
in.
But
yes,
that's
only
I
want.
It
was
part
of
the
unification
with
the
edge
CDC.
That's
the
ADM
CLI
tools,
so
at
CDC
SEO
ADM
will
be
the
command-line
tool
that
your
your
use
and
will
try
to
build
in
features
to
restore
from
backup
lists.