►
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
and
welcome
to
the
valero
community
meeting
today
is
april.
20Th
2021
and
we're
gonna
go
through
some
status
updates
and
then
dive
into
a
few
discussion
topics.
First
up
we
have
alicia
with
a
fantastic
table
of
things.
C
Hi
everyone,
so
I
don't
have
like
a
huge
mind
to
report
on
items
working
towards
the
roadmap.
I
was
just
doing
some
internal
follow-up
stuff
following
the
1.6
release,
but
for
the
roadmap
roadmap
items
I
have
started
some
design
work
on
plug-in
versioning,
so
I've
been
having
some
discussions
with
dave.
Aren't
that
I'm
hoping
to
try
and
get
ideas
and
everything
kind
of
solidified
into
a
document
and
hopefully
have
something
to
share
with
the
community
soon.
D
Yeah,
I
am
returning
from
break
catching
up
on
stuff.
I
I,
admittedly
have
not
looked
at
github
yet
this
week,
but
I
I
did
want
to
say
for
coming
up.
I'm
picking
up
picking
up
some
design
work
that
I
had
put
out
before.
There's
an
older
pr
for
the
valero
object
manifest
and
I'm
going
to
go
over
that
in
the
discussion,
topics
and
kind
of
maybe
give
people
more
context
on.
Why
it's
there
and
then
discuss
an
issue
that
I
see
with
it
as
it
stands.
A
Sounds
good,
I
also
have
a
question
for
you
nolan,
so
we
we
talked
about
going
over
road
maps.
I
think
we
said
the
first
week
of
every
month
should
we
schedule
that
for
the
beginning
of
may
as
well,
the
1.7
roadmap.
D
E
Well,
what
do
we
want
to
go
over?
I
mean
we,
don't
we've
got
the
one,
seven
mode
maps
out
there
we
may
or
may
not
have
any
adjustments
to
it
by
then.
D
E
A
Yeah,
it
was
just
that
we
we
we
set
that
up
to
have
a
the
first
meeting
of
every
month
that
we
go
over
the
roadmap,
no
matter
if
it's
a
new
roadmap
or
a
roadmap
that
we've
been
working
on.
So
we
would
do
the
same
thing
for
june
and
july
and
august
and
so
on.
F
I'm
fine
with
whatever
I
I
haven't
been
around
enough
to
have
an
opinion
on
this.
I
think
that
the
roadmap,
I'm
hoping
in
a
month
or
two
we'll
be
able
to
have
it
really
priorities
based
on
the
vision,
but
I
don't
think
we'll
have
that
by
may
1st,
apparently
once
we
prior,
if
we
do
decide
to
reprioritize
based
on
the
vision
for
sure
we
should
have
a
meeting,
then
for
that.
E
All
right,
yeah,
mainly
working
with
eleanor
on
putting
together
the
product
definition,
vision,
statements
and
figuring
out
what
we're
doing
next.
A
Yeah
I'm
kidding,
I'm
kidding,
sounds
good.
Any
questions
comments
for
dave.
G
For
me,
I
have
been
trying
to
scope
out
scope
out
the
work
items
that
we
want
to
take
up
for
the
one
seven
with
respect
to
csi.
That's
pretty
much
I
will.
I
will
have
more
to
share
once
I
have
it.
I
have
more
complete
ideas
of
what
involved.
B
Alrighty
so
just
wanted
to
let
everybody
know
we
have
a
folder
on
the
github
repo
with
logo
assets.
This
is
for
yeah.
B
Looking
at
plenty
for
1.7,
just
getting
making
sure
the
my
the
issues
are
assigned
to
me
and
broken
down.
I
just
I
started
doing
that
work.
I
have
not
finished.
B
I
am
also
learning
how
carvel
work,
mainly
the
ytt
and
cap
tooling,
out
of
that
too
sweet
and
tldr
it's
it's.
A
combination
of
ytt
is
a
combination
of
hell
and
customize.
It's
pretty
cool,
it's
very,
very
neat,
so
we're
going
to
be
up
for
version
1.7
we're
going
to
have
a
villa
artifacts
that
can
be
used
with
that.
Just
like
we
can
use
help,
and
so
some
prs
one
thing
of
interest
is
that,
as
I
was
making
sure
my
things
for
1.7
were
on
the
board.
B
I
will
fix
it
and
then
I
had
thought
that
I
would
add
it
myself
like,
but
it's
all
all
of
that
pr
excludes
that
that
action
excludes
prs
from
going
to
the
board.
So
that
means
like
there
is
a
bunch
of
things
there
and
it's
really
hard
to
tell
what's
issue.
What's
pr
like
so
if
people
people
like
maintainers,
please
just
make
sure
your
stuff
is
in
there,
that's
what
I'm
doing
with
my
issues
so
I'll
look
at
it
I'll
fix
it
this
week.
So
at
some
point
and
that's
it
for
me.
A
So,
first
of
all,
thank
you
for
for
adding
those
assets
we'll
link
to
the
the
vmware
logo
guidelines
as
well
from
the
general
guidelines,
but
yeah.
This
is.
This
is
great.
Thank
you
sure.
B
A
A
I'll
I'll
send
you
the
link
later
sure
thing.
This
is
awesome.
Thank
you.
All
right.
We've
got
two
discussion
topics,
three
discussion
topics
to
go
through
today.
First
up
we
have
rahul.
G
G
So
this
is
the
scenario
which
I
was
trying.
I
created
a
namespace
of
around
150
pvs
and
around
150
pods.
It
had
around
2400
resources
and
I
performed
around
30
odd
backups
and
tried
to
delete
one
of
the
backups
and
the
backup
remained
in
the
deleting
phase
forever,
and
I
must
mention
this
happened
without
valero
any
of
the
valero
parts
getting
restarted
and
thereafter,
when
I
submitted
additional
delete
request,
those
were
not
picked
up,
so
the
status
attribute
was
not
getting
updated
for,
depending
for
the
additional
delete
request
crs,
which
I
could
see.
G
So
I
thought
perhaps
the
delete,
controller
or
thread
is
getting
stuck
up
and
I
restarted
valero.
The
observation
after
that
is
again
the
original
backup
which
was
stuck
up
in
deleting
phase
it
remained
there,
but
the
newer
requests
were
served
and
thereafter
it
could
delete
one
backup
but
again
for
the
next
backup.
Similar
thing
occurred,
it
remained
in
deleting
phase
and
now
it's
in
the
same
state
it
doesn't
accept
additional
requests.
G
Log
files
may
be
okay,
I
will
get
the
log
files,
you
need
the
log
files
for
valero
as
well
as
backup
driver
as
well.
E
Yes,
please
I
mean
there's.
G
A
E
I'm
going
to
put
this
under
support.
I
don't
think
it's
really
a.
I
don't
want
to
get
too
much
of
our
discussion
wrapped
into
debugging
support
requests
unless
we've
got
lots
of
time
for
other
things.
So
please
take
this
for
support.
We
probably
need
to
involve
the
vsphere
team
because
it's
probably
a
vsphere
issue.
G
Sure
so,
while
I
get
the
logs
as
well
by
any
chance,
similar
issue
was
observed
earlier
and
by
any
chance
this
might
be
fixed
in
the
latest.
Belero
version
that
is
156
154
is
what
I'm,
using
by
the
way.
G
A
Thank
you
role
all
right.
Next
up,
we
have
nolan.
D
Yeah,
so
I
want
to
go
over
this
and
I
think
it
might
be
useful
if
I
whiteboard
some
stuff,
but
to
give
a
high
level.
Do
you
want
me
to
stop
sharing,
and
you
can
sure
sure.
D
So
right
now,
valero
backs
up
items
one
at
a
time
right,
so
you
get
object,
one
get
out
to
two
and
so
forth
it
doesn't.
It
doesn't
take
into
account
any
sort
of
relationship
between
them.
D
What
this
proposal
is
doing
is
it's
going
to
try
to
be
a
little
bit
more
intelligent
about
the
order
in
which
it
gets
things,
so
the
kind
of
main
thrust
of
it
is
to
let's
say,
take
I'm
going
to
take
cluster
api
objects
as
an
example,
because
this
is
really
what
what
spurred
this
on
the
problem
with
doing
these
one
at
a
time
is
when
we
restore
we
could
restore
in
alphabetical
order,
but
that
alphabetical
order
doesn't
get
dependencies
right.
D
So
in
cluster
api
there
is
a
cluster,
a
resource
set
binding
that
is
owned
by
a
cluster
and
what
valero
was
doing
was
it
was
trying
to
restore
the
crsb
first
and
the
cluster
api
controller
would
be
hey,
wait,
wait!
Wait!
You
can't
do
that
because
it's
owned
by
a
cluster
that
I
don't
see
you
don't
have
a
cluster
made
yet
because
of
how
the
naming
resolved
the
cluster
came
after
the
cluster
role
set
binding
and
that's
incorrect
from
the
controller's
point
of
view.
D
So
what
I'm
proposing
is
essentially
to
have
valero
follow
these
ownership
links
so
that
on
restore
the
cluster
would
be
restored
first
and
then
anything
that
depends
on
it.
D
In
this
scenario,
the
ownership
is
it's
a
fairly
safe
bet.
If
we
see
a
field
named
owner,
it's
a
fairly
safe
bet
that
it
is
an
object
reference.
D
However,
there
are
other
objects
that
may
use
a
different
field,
so
I
was
looking
at
some
of
the
the
work
that
folks
are
doing
internally
on
a
ci
type
system
and
they're
using
a
field
called
source
and
that's
actually
an
object
reference,
but
you
wouldn't
know
that
just
by
reading
yml.
D
So
if
I
stop
sharing
the
the
thing
I'm
I'm
looking
into
is
how
do
we
make
sure
that
the
owner
should
the
this
link
is
known
to
the
valero,
regardless
of
what
the
field
name
is
named
dave,
and
I
talked
about
this
a
little
bit
yesterday
and
one
thing
we
thought
about
was
doing
on
on
the
resource
definition,
putting
an
annotation
that
says
these
fields
are
actually
references,
but
I
was
thinking
last
night.
D
For
crds,
I
don't
know
about
built-in
types,
but
crds
we
have
an
open
api
specification.
It
says
this
field
does
this,
this
field
does
this
this
field
does
this,
and
if
we
can
fetch
that
and
say
give
me
all
the
fields
that
are
references
and
use
that,
then
I
think
that
makes
this
a
lot
easier,
but
I
need
to
do
some
more
research.
H
Let's
start
get
on
crds,
then
that's
permission
set,
which
I
guess
we
already
have.
So
maybe
it's
not
a
problem.
D
Yeah
yeah,
I
mean
we
have
to
do
that
to
be
able
to
back
it
up
at
all,
but
if
we
get
the
crd
and
look
through
the
schema,
I
think
that
can
tell
us
what
stuff
is
a
link
and
then
use
that
on
instances.
D
D
I
might
be
able
to-
and
this
is
why
I
was
bringing
it
up-
is.
D
H
That
might
be
a
safer
bet
to
go
look
at,
since
that
is
what
everything
ends
up.
Getting
translated
to
right,
so
crds
and
internal
types
get
served
by
a
discovery
api.
Does
that
aggregated
api
servers
so
like
right?
The
other
thing
you
have
to
worry
about
is
like
an
aggregated
api,
so
discovery
api
would
probably
be
your
best
bet.
I
think.
D
H
D
Well,
I
we
can
explore
it,
so
I
was
I
I
can
look
into
that
some
more.
H
The
the
question
I
was
going
to
ask
is
yeah:
are:
are
you
trying
to
trying
to
think
you're
trying
to
go
and
look
got
it
references
the
like
canonical
owner
reference
like
kubernetes
owner
reference,
you're,
saying
owner
here,
as
in
like
a
reference
inside
of
the
spec
to
a
given
other
resource
on
the
cluster.
D
Yes,
okay,
so
so
that,
instead
of
just
having
a
list
of
things,
we
have
a
graph
and
we
kind
of
know
what
the
dependencies
are,
so
that
when
we
restore
we
restore,
we,
we
don't
restore
sequentially
like
in
a
list.
We
restore
like
a
graph.
H
D
We
may
get
multiple
graphs.
Okay,
the
root
would
be
cluster.
I
imagine
the
root
is
cluster,
which
would
then
have
like
cluster
scoped
objects
and
then
with
like
name
spaces,
and
things
like
that.
But
carlesia
had
a
question.
B
It's
a
comment
that
I
have
so
I
was
watching
some
videos,
video
pres,
like
tk
the
gi
in
another
video
about
by
tt
and
k,
end
cap-
and
I
think
it's
ytt,
but
I'm
not
clear
one
of
the
two
has
a
command
and
that
command
has
a
flag
called
3-3
that
present.
Instead
of
presenting
a
list
of
the
objects
that
will
be
deployed,
I
think
it's
cap,
because
this
has
to
do
with
the
deploy.
B
It
will
present
a
graph
of
the
top
objects
and
all
of
this
dependency
as
as
leaves
on
a
tree,
it
might
be
worthwhile
looking
how
they
are
doing
it.
I'm
pulling
up
the
the
help
here
on
for
the
cli
and
I'm
not
finding,
but
I
have
to
look
for
it.
C
Yeah
I
tried
to
do
something
similar
to
this,
for
a
plug-in
for
sauna
boy
and
yeah.
I
remember
looking
it
wasn't
in
the
project
that
carries
you
mentioned,
but
there
was
there
was
another
one
called
like
there
was
like
a
plug-in
for
cube
ctl
at
the
time
cube
ctl
tree,
and
I
remember
looking-
and
I
think
maybe
octant
does
something
similar
as
well.
Given
that
it
can,
you
can
dynamically
look
at
objects
within
the
cluster
and
see
the
relationships
between
them.
C
So
I'm
wondering
if
there
are
some
like
kind
of
libraries
or
something
out
there
that
can
help
make
this
a
bit
easier,
because
you
know
the
thing
that
I
struggled
with
when
trying
to
do
it,
for
something
boy
is
that
the
owner
references
start
at
the
bottom
of
the
tree
and
work
their
way
up,
so
you
can't
start
with
the
node.
D
C
Yeah,
but
what
I
mean
is
like
in
terms
of
like
building
the
tree,
would
be
really
nice
if
you
could
start
somewhere
at
the
higher
level
and
have
the
kubernetes
apis
help.
You
form
that
tree,
but
a
bit
more
diy.
E
D
So
we'd
probably
start
like
this
because,
like
you
say,
dave,
there's
no
cluster
object.
Cluster
scoped
objects
would
have
to
get
restored
just
directly
into
the
cluster
you're
on
or
you
know,.
D
No,
not
necessarily
I
mean
it
could
be,
but
on
let's
say
I'm
on
an
existing
cluster.
I
think
you'd
be
right
in
that
cluster-scoped
resources
could
be
unordered.
E
D
Right
so
that
the
cluster
resource
has
to
exist
first
before
what,
before
the
cluster
resource
set
binding,
which
is
a
crd
that
says
these
nodes
like
vms
and
metal,
nodes
or
whatever
are
belong
to
this
cluster,
and
we
kind
of
included
around
it
by
adding
those
two
definitions
to
valero's
default
restore
priorities,
but
I
think
that's
a
temporary
workaround
and
we
can't
chase
down
all
of
these
and
maintain
that
list
perfectly
accurate.
D
E
Yeah
so
so
ordering-wise
so
like,
for
example,
if
the
cluster
has
vms
that
cluster,
not
kubernetes
plus
but
the
capital
cluster
object,
so
it
has
vms
we'd,
restore
the
vms
first
right,
then
the
cluster
object.
Then
the
crsv.
G
I
I
think
that
won't
work
right
so
for
the
nodes
to
be
there
they
need
to
be,
they
need
to
belong
to
a
cluster.
You
would
that's
the
exact
problem
we
are
trying
to
solve.
We
first
we
first
create
the
cluster,
and
then
we
create
the
resources
that
belong
to
the
cluster.
So,
for
example,
if
we
have
to
create
a
load,
balancer
object,
I
don't
think
that
proposal
has
been
implemented
yet,
but
the
load
balancers
will
also
be
represented
as
a
crd.
G
G
Load,
the
the
load
balancer
object
is
owned
by
the
cluster,
so
the
cluster
has
to
be
created
and
the
owner
has
to
be
created.
First
before
you
create.
H
H
G
What
it
is
it
is
set
up
as
owner
reference,
actually
I'm
not
sure
about
specifically
about
the
load
balancer
proposal
top
of
my
head,
but
there
are
certain
fields
that
are
on
the
references
and
certain
fields
that
are
not
done,
but
yeah
did
that
clarify
your
question.
G
G
So
let
let
me
let
me
bring
something
out
that
I
can
talk.
G
Okay,
so
the
the
way
the
capi
clusters
are
set
up
is
there's
a
cluster
object,
which
is
which
represents
the
definition
of
a
kubernetes
cluster
in
terms
of
what
cluster
version,
what
kubernetes
version
and
what
cluster
infrastructure
belongs
to
the
cluster
with
me
so
far
yeah.
G
So
when
we
are
backing
up,
we
back
up
everything,
but
when
we're
restoring,
we
first
need
to
there's
a
ordering
requirement.
G
G
E
G
Yeah,
what
is
paused
there
is
when
you
create
machine
object
like
when
you
create
machine
deployments
or
anything,
you
don't
want
the
capi
controllers
to
start
creating
machines
for
you,
because
the
machines
may
already
exist
right.
So
that
is
the
scenario
where
you
say:
hey
pause,
reconciliation
for
this
cluster.
I
am
going
to
be
creating
objects
and
then
you
will
see
everything
mapped
together.
E
It
does
because
the
order
that
you're,
restoring
things
in
here
is
you're
talking
about
a
top-down
restoration
and
the
other
and
for
like
other
operators,
we're
doing
bottom
up
so
that
the
reconciliation
loop
doesn't
kick
in
early,
so
pausing
the
reconciliation
loop
on
cappy
lets.
You
do
a
top-down
restoration,
because
the
reconciler
is
not
running
the
controller's,
not
creating
resources,
because
you
pause
views
that
don't
create
like
vms
for
this
cluster.
H
H
H
D
D
H
H
H
D
Cube
should
reconcile
this.
I
have
a
bug
filed
with
cluster
api
team
for
their
behavior.
However,
we
can't
we
as
a
backup
tool
should
not
be
enforcing
api
contracts.
We
should
be
restoring
stuff,
as
is
not
necessarily
strictly,
as
is
my
goal
here,
is
to
get
get
valero
so
that
it's
more
intelligent
for
designers
of
applications,
and
they
don't
have
to
worry
as
much
now
should
they
should
most
of
them
be
doing
stuff
that
can
reconcile
afterwards.
Yes,
should
lots
of
crds
not
need
status
to
be
be
restored?
D
Yes,
but
there's
people
that
want
us
to
restore
status,
so
I
think
you're
you're
you're
right
in
a
sense
sean,
but
I
don't
think
it's
going
to
be
on
us
to
enforce
these
api
contracts.
G
Yeah
and
we've
seen
that
multiple
times,
especially
with
the
way
crds,
are
implemented,
with
states
being
stored
in
status
and
not
the
controllers,
reconciling
state
and
like
updating
the
status
as
based
on
what
they
see
and
it
becomes
hard
for
a
tool
like
valero
to
have
strong
opinions
about
how
controllers
and
crds
should
be
implemented.
E
D
In
that
case,
it's
not
an
owner
ref,
it's
just
a,
and
that
is
literally
the
type,
is
literally
object.
Reference
owner
refs
are
a
special
cased
version
of
an
object
reference
where
if
the
owner
doesn't
exist,
then
you'll,
then
the
object
will
get
garbage
collected,
which
is
another
reason
why
the
owner
should
exist.
First.
E
D
H
I
think
we
can't.
I
think
that
goes
back
to
my
earlier
point,
which
is
like,
if
you're
restoring
objects
in
kubernetes,
should
you
not
expect
kubernetes
objects
to
behave
like
kubernetes
objects.
E
E
H
Well,
yes,
and
why
would
you
restore
the
so
if,
if
we
know
that
that's
what
is
going
to
happen
because
it's
owned
by
that
resource,
why
are
you
restoring
the
owned
resource
if
the
owner
is
going
to
recreate
it?
It
says.
B
H
D
So
the
owner,
so
in
the
owner
reference
case,
let's
say
you
want
to
use
rustic
with
valero.
We
have
to
restore
the
pod,
not
the
deployment
because
of
of
the
rustic
logic.
That's
that's
kind
of
our
our
burden
to
bear
with
things
like
cluster
api.
D
If
you
want
to
reclaim
a
running
vm
that
is
not
accounted
for,
then
you
wouldn't
want
the
controller
to
reconcile,
because
if
it
did
it's
going
to
create
a
brand
new
vm,
which
is
why
you
would
be
doing
all
this
pausing
to
say:
yes,
I'm
restoring,
but
there's
a
vm
that
already
exists
that
I
want
you
to
tie
to.
I
don't
want
you
to
make
a
brand
new
one.
H
D
D
So
this
proposal,
right
now
as
stands,
is
to
just
write
this
data,
not
use
it
for
a
restore
because
you're
bringing
up
some
good
points,
and
I
think
the
first
step
is
to
have
this
data
and
then
work
on.
What's
the
proper
restore
case.
E
I
would
say
why
don't
we
take
a
look
at
like
some
80
20
stuff
so,
like
your
proposal
on
owner
graphs
probably
covers
80,
and
then
we
have
like
an
additional
like
some
way
for
either
for
both
the
application
and
the
user
to
specify
a
strict
ordering.
So
they
may,
for
you
know
so,
some
way
you
can
write
a
crd
that
says
order
these
resources
like
this
and
I'd
like
to
make
this
something
that's
available
to
the
user,
because
you
know
like
sean
your
point
on
making
a
plugin
isn't
a
bad
one.
B
E
Then
they're
doing
a
backup
restore
like
it
doesn't
work
and
the
actual
maintainer
of
the
application
may
not
be
interested
in
creating
a
plug-in
for
it,
and
the
plug-in
right
now
is
pretty
heavy
weight
in
terms
of
the
amount
of
work
to
build
one
in
valero.
So
I'm
thinking
that
we
could
have
like
a
generic
case,
a
thing.
That's
that
lets
you
override
something
that
will
probably
cover
the
other.
You
know
nine,
you
know
29
and
then,
or
you
know,
then
the
last
five
percent
is
like
okay.
E
D
I
want
to
reiterate
this
is
not
going
to
be
a
complete
solution
because
for
things
that
don't
just
reconcile
we
may
have
to,
they
may
have
to
write
plug-in
as
sean
pointed
out.
Sometimes
the
correct
thing
we'll
to
do
will
be
to
manifest
new
resources,
really
just
let
the
controllers
do
their
thing,
but
as
the
state
of
the
world
is
right
now
with
a
lot
of
apps.
D
Yes,
they
should
be
designed
for
a
certain
way
for
kubernetes
they're,
not
that's
kind
of
the
reality
of
it.
Even
valero
is
not
a
super
good,
a
good
example
of
completely
declarative
stuff,
and
for
it
to
be
a
useful
tool.
I
don't
think
we
can
expect
everyone
to
to
maintain
perfect
api
contracts.
G
I
I
think
it
would
be
better
if
we
solve
this
just
looking
at
it
from
looking
at
things
as
communities
objects
and
not
bringing
what
the
controllers
may
or
may
not
do,
because
that
we
are
being
asked
to
back
up
kubernetes
objects.
We
will
backup
and
restore
communities,
objects
whatever
the
controllers
does,
then
it's
up
to
the
controllers
and
we
have
no
controller
control
over
the
controllers.
H
But
I
I
would
make
an
argument
that
I'm
the
world
that
I'm
presenting
is
actually
more
in
line
with
the
kubernetes
api,
but
yeah
like
if
you
have
an
owner
reference
right.
It
literally
means
that
that
thing
is
the
owner
of
it
yeah
it
owns
it.
Why?
Why
is
some
other
process
creating
that
resource
right.
H
No,
I
don't
think
I
am,
I
don't
think
I
am,
I
think,
I'm
making
a
case
against
backing
up
owned
resources.
Well,
we
do
that
now
we
do
that.
D
K
For
that
I
mean
for
our
plug-ins
on
our
side.
You
know,
with
the
our
migration
we're
actually
overriding
that
functionality
and
not
backing
not
restoring
pods
that
have
owner
references,
but
the
default.
You
know
when
you
don't
have
that
plug-in.
Is
that
their
blur
installs
the
restores
it.
K
I
noticed
at
some
point
it
was
added
and
I
I
was
before
I
was
working
on
this.
I
was
just
looking
at
history
at
one
point.
The
original
version
did
not
restore
pods
that
had
owners
and
then
at
some
point
it
was
added.
Perhaps
someone
wanted
it,
you
know,
I
don't
know.
Maybe
that
should
be
configurable
top
level.
That's
a
different
discussion.
K
E
D
Remember
what
you're
talking
about
scott?
I
don't
remember
the
bug
that
we
ran
into,
but
yeah
it
was
that's
how
valero
works
like
oh
throwing
a
reference.
Just
let
it
do
the
thing,
and
I
don't
know
if
it
was
because
of
the
rustic
support
that
we
decided
to
do
pods
that
way
or
not.
We.
K
I
think
it
was
with
diamond
sets
but
and
something
else
where,
if
we
restored
the
pod
that
had
the
under
reference
on
restore,
the
cluster
was
creating
another
pod
and
we
ended
up
having
duplicated
pods
because
they
weren't
the
only
references
themselves,
weren't
getting
preserved,
and
so
I
just
wrote
a
plug
again
to
nuke
those
and
say
you
know.
If
there's
another
reference
for
a
pod,
don't
restore
it.
D
Yeah
and
stateful
sets
are
another
huge
edge
case
that
we
don't
handle
well,
because
those
assume
those
assume
ordering
and
stuff
and
those
those
are
a.
H
D
If
you
have
it
create,
if
you
have
the
storage,
I'm
sorry
staple
set
recreate
all
of
its
pods.
It
has
no
information
on
what
volumes
they
were
linked
to.
D
It's
you're
right
that
this
is
all
weird
and
and
valero
is
doing
these
weird
inversions
of
how
kubernetes
is
supposed
to
work.
It's
that's.
The
challenge,
I
think,
is
getting
it
correct
and
part
of
the
thing
we
need
to
do
is
like
evangelism,
with
people
writing,
controllers
and
apps.
To
say:
hey
maybe
be
aware
that
you're,
not
the
only
thing
creating
these
objects,
I
mean
a
person,
could
just
ctl
apply
a
thing
and
that
maybe
you
know
that
may
not
be
the
right
thing
to
do.
D
H
E
D
I'm
going
to
do
a
time
check
here.
There's
there's
an
issue
out
that
I
would
like
people
to
reply
on,
I'm
looking
on
at
least
getting
the
proposal
to
a
state
that
people
can
agree
on
in
one
seven,
I
specifically
say
in
the
proposal.
D
I
I
don't
want
to
address
the
restore
of
the
the
data
right
now,
I'm
just
trying
to
get
to
a
point
where
we
can
have
this
data,
but
I
guess
I
wanted
to
introduce
it
and
see
and
show
some
of
the
the
problem
cases
I
was
having
and
sean.
I
would
actually
really
appreciate
if
you
you
commented
there,
so
we
have
a
documentation
of
these
issues
raised,
but
I'd
also
like
to
give
time
back,
because
I
see
we
have
two
other
topics:
yeah.
D
D
I
I
your
point
is
well
taken
at
this
point.
Yes,
it's
it
is
tough
and
we're.
We
are
getting
ourselves
into
a
something
that
maybe
a
no-win
situation,
but
if
we
can
give
tools
to
controllers
to
participate
in
their
restore,
I
think
that's,
probably
the
the
better
long-term
play.
A
All
right
great
discussion,
so
we've
got
two
more
topics:
ayush,
it
looks
like
there's
a
support
issue.
J
Yes,
so
actually
I'm
facing
one
issue
while
restoring
a
namespace
which
has
couple
of
pvs,
so
I
took
a
backup
of
a
namespace
with
few
pvs
did
not
update
it.
No
changes.
After
that
I
tried
to
restore
the
same
without
a
restore
pv
option.
I
I'm
trying
to
override
this
namespace
and
it
seems
like
the
download.
The
snapshot
operation
is
failing
and
the
retrial
logic
is
going
in
indefinite
as
an
infinite
loop
yeah.
So.
E
Again,
I
mean
I'm
going
to
stop
you
here,
because
this
is
a
support
issue.
We'd,
ask
you
to
open
something
on
github
and
logs,
please
logs
yeah,
I
mean
we
can't.
We
can't
do
live
debug,
it's
not
going
to
work.
This
is
also
a
good
topic
to
bring
for
office
hours
on
thursdays.
J
Yeah
I
have
logs
I'll
yeah.
L
Create
one
so,
okay,
this
is
number
I
think,
so
we
were
facing
these
challenges
and
we
thought
we'll
bring
it
up
to
your
attention.
Of
course,
we
will
provide
all
the
logs
and
whatnot
and
we
can
talk
about
this
on
friday,
but
one
thing
which
we
have
seen
is
repeated
this.
L
This
thing
that
something
was
working
in
the
previous
version
of
the
plugin.
Now
it
is
not
working.
You
can
correct
me
right.
I
think
the
way
it
is
is
in
the
previous
version.
This
functionality
was
not
a
problem,
but
in
the
latest
version
it
seems
to
be
a
broken
thing.
So
I'm
just
trying
to
understand
what
type
of
test
cases
we
have
when
we
are
moving
from
a
different
version
to
a
different
version.
Are
these
things
covered,
because
that
is
something
that
we
like
to
see.
E
Yeah
can
think,
especially
with
the
v-sphere
plug-in
you
know
it
went.
It
underwent
basically
a
huge
re-architecture
as
a
result
of
project
pacific,
so
we
do
have
a
set
of
tests,
it's
something
we
can
bring
up
with
xing
and
the
vsphere
test
team
and
see
if
these
cases
are
being
covered.
E
L
A
Thank
you
amber
all
right.
We
got
one
more
for
eleanor.
F
Here
double
muted
right-
I
just
I'm-
maybe
the
newest
one-
to
this
team.
I
think
that's
probably
fair
and
yeah.
I
mean
I've
still
been
here
a
month
now,
but
nonetheless,
that's
still
pretty
new
and,
in
short,
we're
seeing
different
faces
every
time
at
the
community
meetings,
and
I
would
just
love
to
know
sometimes
they
turn
out
to
be
from
vmware.
Sometimes
you
all
turn
out
to
not
be
from
vmware.
F
So
I
was
wondering
if
we
could
kind
of
in
some
way
identify
these
community
meetings
who
folks
are
and
what
their
affiliation
is.
So
I
don't
have
to
do
sleuthing
kind
of
googling
behind
the
scenes
so.
H
F
A
Yeah,
so
the
I
think
the
easiest
way
to
do
this
is
to
ask
everyone
in
the
beginning
of
the
meeting
to
just
add
in
their
their
name
and
affiliation
in
the
meeting
notes,
because
then
people
can
just
continue
to
add
in
there
and
we
will
see
who
is
there?
It's
easy
to
see
when
we
look
back
at
meetings
as
well,
who
was
in
those
meetings
and
and
who
listen
in
on
the
discussions
instead
of
having
a
physical
roll
call
on
video
and
audio.
F
That
works
perfectly
for
my
purposes.
Thank
you
very
much,
and
does
anyone
have
any
objection
to
this?
Of
course
you
don't
have
to
put
your
affiliation
if
you
don't
want
to,
but
it
would
just
be
helpful
for
all
of
us
to
understand
the
perspective
that
you're
speaking
from,
but
does
anyone
have
any
objections.
G
No
objection
to
that
so
the
way
I've
seen
this
work
in
some
of
the
other
sig
meetings
is
like
there's
an
attendee's
section
in
the
meeting
notes
and
everybody
just
adds
their
name
to
that
list.
I
added
an
attendee
section
to
today's
meeting.
So
if
you
all
want
to
go
up
to
add
yourself
to
that
list,
we
can
get
that
process
going
from
today.
B
I
missed
your
your
audition
ashish.
I
added
that
section
for
next
week's
meeting.
B
Yeah
every
I
think
it's
a
great
idea,
so
we
add
name
and
company.
A
Awesome,
thank
you.
You
can
stop
sharing
as
well
as
you
should
think.
B
A
As
well
as
collegiate
sorry
awesome,
and
then
we
have
one
contributor
shadow
that
we
want
to
do
as
well.
So
I'm
gonna
turn
my
screen
quickly
here,
so
you
can
all
see
it
there
we
go
so
we've
got
one
contemporary
shout
out
here
from
for
to
bs.
D
Yeah,
so
this
this
is
a
an
enhancement
to
the
helm
chart
that
adds
or
exposes
the
disabled
controllers
argument
to
the
helm
chart.
This
could
be
useful
if
you're
doing
valero
in
a
restore
only
mode
and
don't
want
any
of
the
controllers
for
backup
to
start
at
all
or
if
you
want
to
do
some
other
custom
stuff.
So
thank
you,
tobias
to
bs
for
that
change.
A
Yeah
might
be
tobias,
I'm
going
with
the
the
swedish
pronunciation.
Could.
A
You
never
know
awesome,
thank
you
and
with
that
that
is
the
last
item
of
today
of
the
valero.