►
Description
April 28, 2020
Status Updates
steve
finalized backup progress PR
fixed issue with unresolvable includes/excludes
reviews
[stephanie] - update on where we are currently with the velero 1.4 release
https://github.com/vmware-tanzu/velero/milestone/23
https://github.com/vmware-tanzu/velero-plugin-for-csi/milestone/2
[carlisia]
Adding tests to the new BSL controller (https://github.com/vmware-tanzu/velero/issues/1967)
Recorded a tutoria to show the build/run/test workflow for Velero: https://www.youtube.com/watch?v=Aiz9M4aVFdA
On community support this week
[nrb]
Was out sick late last week, not a ton of updates from me. Spent yesterday replying to things I’d missed
Need to get 2373 reviewed and merged this week so we can get CRD restore fixes started
Will also make a test of using the v1beta1 client approach
Will be on Brett Fisher’s YouTube channel this Thursday at 1pm ET - https://youtu.be/pRgap975b5E
[ashish]
Community support this past week
Reviewing some PRs and getting CSI closer to the beta ship
Discussion Topics
[steve] demo of backup progress
[stephanie] - can we document how to execute backup and recovery in an airgapped environment?
[nrb] - this is partially documented at https://velero.io/docs/v1.3.2/on-premises/#on-premises-environments, though it’s incomplete since there’s a lot of possible options in airgapped/on-prem environments
[from timh] Perhaps we can reuse some of the docs over at Sonobuoy
[carlisia] Question: should we (RE)consider not crashlooping for when there’s no BSL? This issue was posed here: https://github.com/vmware-tanzu/velero/issues/1967#issuecomment-620208525
[marccampbell] Custom Restore Hooks proposal: https://github.com/vmware-tanzu/velero/pull/2465
Contributor Shoutouts
@raghavendrabhat - clarify wording in velero restore describe for included namespaces (https://github.com/vmware-tanzu/velero/pull/2449)
@mynktl - skip VolumeSnapshotLocation validation for backups when --snapshot-volumes=false (https://github.com/vmware-tanzu/velero/pull/2450)
A
Good
morning,
good
afternoon,
good
evening,
everyone
and
welcome
to
today's
episode
of
the
valera
community
meeting,
slash
open
discussion.
The
the
date
is
April
28th,
we
are
almost
done
with
April.
April
went
way
way
faster
than
March
I.
Think
most
of
you
will
agree
there.
So
we're
almost
done,
but
we
do
have
one
more
community
meeting
here
for
bolero
for
this
month.
So
I'm
gonna
share
my
screen
and
we'll
go
through
the
status
updates,
and
then
we
have
a
few
discussion
topics
as
well
and
start
here.
A
B
Quick
update
here
so
I
finalized
the
code
for
the
backup
progress
PR,
which
I've
been
talking
about
for
the
last
couple
of
meetings
and
I'm
gonna
do
a
demo
of
that
in
a
few
minutes,
and
that
is
ready
for
review
from
from
reviewers
yesterday,
I
fixed
an
issue
that
we
have
bug
that
we
found.
We've
got
a
couple
of
reports
of
actually
around
included
and
excluded
resources
that
are
unresolvable
through
the
discovery
API.
B
There's
a
bug
where,
potentially
you
end
up
with
your
plugin
getting
run
for
every
resource
type,
which
is
is
definitely
not
desired.
Behavior
so
anyway,
I
found
a
way
to
fix
that
and
got
that
merged.
Yesterday,
it
came
up
with
our
CSI
testing,
so
that
was
kind
of
why
I
popped
up
and
then
beyond
that,
just
a
bunch
of
reviews,
again
CSI
stuff
and
looking
at
some
of
the
CRD
issues
which
I'm
sure
no
and
I'll
mention
a
little
bit
and
yeah.
That's
about
all
for
me.
B
I'll
paste
it
in
the
chat
here
to
the
milestone
in
github,
so
there
are
about
15
open
issues
that
are
still
in
there.
We've
got
pr's
open
for
I'd,
say
at
least
half
of
them,
so
some
are
waiting
on
review
or
testing,
but
we
we
do
still
have
several
CSI
related
issues
open.
We
have
a
couple
of
CRD
related
issues,
so
you
know
feels
to
me
like
we
may
be
ready
to
ship
a
beta
sometime
next
week,
but
I
I,
don't
think
we're
quite
ready.
This
week
we
still
got
some
work
to
get
through
that.
B
E
B
I
mean
if
we
think
we
can't
fix
them.
41.4.
Okay,
have
that
discussion,
and
you
know
the
obviously
the
CSI
plug-in
can
evolve
independently
of
Valero
itself,
so
we
could
always
continue
to
make
fixes
and
changes
and
ship
a
new
version
of
the
CSI
plug-in
without
having
to
ship
a
full
version
of
bolero.
But
I
do
think
you
know
we
should
at
least
either
look
at
these
and
say
that
we're
not
going
to
fix
them
for
some
reason
or
we
should
try
and
fix
them.
Yeah.
E
B
And
then
you
know
just
to
cover
our
usual
process:
I
mean
we
usually
ship
at
least
one
beta,
so
we'll
get
it
out
there
once
we're
kind
of
feature
complete,
give
everyone,
including
ourselves,
a
chance
to
test
unknown
version
and
then,
depending
on
kinda,
how
the
feedback
goes
there.
If
there
any
major
issues,
we
need
to
remediate,
we
may
ship
a
second
beta
or
if
there
are
no
major
issues,
its
it's
just
minor
things
and
then
we'll
typically
go
straight
to
GA,
so
that
would
that
would
kind
of
be
the
plan.
C
F
F
A
D
D
D
Yeah-
let's
see
here
yeah
this
this
one,
so
there
was
we
the
last
two
weeks,
I
think
we
talked
about
like
the
backup
version,
backup
format,
version
number
being
changed
to
semantic
versioning
and
Raphael
found
that
actually
we
use
straight
integer
in
arnhem
in
our
data,
so
that
wouldn't
work.
So
we
need
to
introduce
new
field.
D
So
hopefully
we
can
get
this.
This
particular
PR
wrapped
up
and
merge
this
week,
please
sooner
rather
than
later
this
week,
I
suspect,
but
don't
know
that
this
will
be
the
basis
for
a
fix
for
our
CRD,
the
one
beta
at
1
and
V
1
conflict
issues,
but
in
the
meantime,
I'm
also
going
to
start
working
on
a
an
alternate
path
to
see
just
to
compare
those
two.
Those
two
approaches
and
I
know
this-
that
this
this
will
have
merged
conflicts
with
other
PR.
D
So
I'd
like
to
get
this
in
there's
a
bunch
of
other
reviews
that
I
have
going
out
there,
but
now
or
call
them
at
this
time
and
a
just
another
reminder
just
this:
Thursday
I
will
be
on
the
Bret
Fischer
YouTube
show
this
will
probably
be
a
high-level
introduction
to
Valero,
but
that
will
be
at
1:00
p.m.
Eastern.
D
This
Thursday
and
the
link
is
there
for
when
it
goes,
live
so
yeah
I
think
that's
it.
For
me,
oh
and
I
I
see,
mark
Campbell
is
on
the
call
and
I
know
we
didn't
get
to
his
topic
two
weeks
ago
and
he
wasn't
available
last
week.
So
I
don't
know.
If
Mark
you
wanted
to
add
your
customer
store
hooks
item
to
the
agenda
this
week.
Yeah.
A
Awesome
yeah,
we'll
we'll
definitely
send
the
link
out
here
on
slack
as
well
so
remind
everyone
and,
of
course
out
on
Twitter,
and
so
everyone
can
watch
and
yeah
follow
along
as
Nolan
is
introducing
Valero
to
the
world
again.
So
yeah,
that's
going
to
be
super
super
fun
on
Thursday,
they're,
awesome,
Thank,
You,
Nolan.
Any
questions,
comments
for
Nolan.
E
I,
don't
have
very
many
updates.
I
was
in
community
support
last
week,
getting
through
a
bunch
of
PRS
for
getting
closer
to
ship
CSI,
stuff
I
think
we
have.
As
Steve
mentioned,
we
have
ik
four
five
ish
issues
open
that
needs
to
be
addressed,
so
I
will
be
working
on
those
yeah.
That's
pretty
much
for
time
out
going
on.
I,
don't
have
anything
else.
That's
yeah.
A
G
I
just
added
in
there
so
great
awesome.
You
know
we
had
opened
up.
You
know
this
is
like
early
still
a
couple
of
months
ago.
We
opened
up
an
issue
just
to
see
if
the
Valera
team
would
be
open
to
some
concepts
or
under
custom,
restore
hooks
to
mirror
the
custom
back
up
hooks
and
there
seemed
to
be
a
little
bit
of
interest
in
the
in
the
issue.
A
couple
other
people
were
interested
in
it,
and
so
we
going
gone
ahead
and
submitted
a
in
early
initial
design
proposal
for
how
we
could
do
that.
G
B
D
Same
I
think
this
is
a
gap.
That's
really
important
because,
like
you
said
markets
we
have.
We
have
the
the
backup
hooks,
but
we
are
missing,
like
I'm,
seeing
here,
there's
PG
dump
for
post
press,
but
there's
no
equivalent
for
PG
restore.
So
that's
that's
a
pretty
big
missing
piece
of
the
puzzle,
so
at
a
high
level,
I'm
I'm
pretty
I'm
pretty
on
board
with
this,
though,
a
little.
G
Kind
of
context
and
background
will
we've
implemented.
We
have
a
product
here,
replicators
called
cots
that,
like,
let's
you
know,
people
ship
their
party,
software,
kubernetes
and
we've
just
finished
a
really
deep
integration
to
Valero.
So
love
to
show
you
guys
that
sometime
and
like
like
a
lot
of
UI
around
it,
a
lot
of
like
like
total
automation
of
Valero.
It's
like
the
preferred
backup
method
and
so
like
this
just
helps
the
restore
process
a
little
bit
better
in
our
products.
G
A
B
Okay,
so
yeah
so
I've
talked
about
this
feature
for
the
last
couple
of
weeks.
But
you
know
the
main
issue
that
we're
trying
to
solve
is
that,
while
a
backup
is
running,
there's
not
a
lot
of
visibility
to
the
user.
In
terms
of
you
know
how
many
items
are
going
to
be
backed
up,
how
many
items
have
already
been
backed
up,
and
so
we
wanted
to
to
make
some
changes
to
make
it
easier
for
a
user
to
get
a
sense
of
what's
going
on
within
the
backup.
B
So
I'm
just
gonna
kick
off
a
backup
here
of
my
my
cluster.
Nothing
special
about
the
cluster
other
than
I
have
development
image
in
here,
so
I'm,
just
gonna
save
a
little
backup,
great
demo
and
then
I'll
back
up
the
full
cluster
and
actually
before
I,
kick
that
off
up
here
on
the
top
I'm
going
to
just
put
a
watch
on,
let's
see
demo
on
the
describer,
so
we
don't
have
it
yet
so
I'm
gonna
kick
off
the
backup.
Now.
B
B
So
what
we
first
do
is
look
at
how
many
we
we
scrape
items
from
the
API
server
that
matched
the
backup
specs
so
that
matched
the
resource
filters
and
the
namespace
filters,
and
we
count
how
many
of
those
there
are
and
then
as
we're
backing
them
up
we're
actually
incrementing
an
additional
field
which
has
already
disappeared
because
the
backups
completed
that
shows
how
many
items
we've
backed
up
so
far
and
so
that'll
that'll
increment.
As
the
backups
running
once
the
backups
done,
we
simply
show
how
many
items
were
actually
written
to
the
backup
tarball.
B
B
So
in
the
status
section
we
just
have
a
new
field
called
progress,
and
these
names
are
certainly
still
up
for
debate.
I'm
trying
to
describe
things
as
accurately
as
I
can
without
making
them
too
long,
but
so
the
first
field
we
have
is
the
items
matching
spec,
which
is
how
many
things
were
scraped
from
the
API
server.
B
This
number
may
be
different
from
these
other
two
numbers
for
a
variety
of
reasons,
which
include
things
like
backup
item
action,
plugins
that
include
additional
items
so
that
we're
actually
writing
more
items
to
the
tarball.
Then
initial
eMac
match
to
backup
spec
could
be
because
there
are
excluded
annotations
on
items
such
that
we're
not
actually
backing
them
up
and
a
couple
of
other
reasons
but
anyway,
hopefully
that
that
kind
of
gives
us
answer
what's
going
on
here
and
yeah.
That's
that's
pretty
much.
It
so
happy
to
take
any
questions.
B
So
it's
it's
in
this
case,
what's
happening,
is
that
I
also
have
the
CSI
snapshot
code
deployed,
so
what's
happening
is
I
said,
create
a
full
backup
of
the
cluster.
We
scraped
everything
from
the
API
server
and
there
were
598
of
those
we
went
through
and
processed
all
all
598
of
those.
However,
as
part
of
backing
up
the
PVC,
in
this
case,
we
actually
created
a
CSI
volume,
snapshot,
custom
resource
and
a
volume
snapshot,
content,
custom
resource
and
both
of
those
got
written
into
the
tarball.
B
F
Am
I'm
wondering
I,
don't
my
head
is
not
in
that
right
now,
but
I'm
wondering
if
we
can
reword
any
of
these
labels
to
make
it
clear
that
those
numbers
are
not
expected.
So
much
I
was
I
was
gonna,
say
that
and
then
John
urges
reefa
sized,
because
I
think
we're
going
to
always
get
questions
like
this.
B
E
E
D
Also
say
we
can
the
the
naming
of
these
fields
may
be
hard
harder,
but
we
can,
if
we
use
Valero
back
up,
describe,
we
can
get
more
verbose
like
the
these
fields
on
the
data.
Type
might
be
tricky.
Excuse
me,
but
envelope
I've
described
granted.
We
can't
necessarily
you'll
watch
on
that
Italy,
but
if
we
we
do
the
output
there
for
all
of
them,
we
can
get
a
little
bit
more
robust
and
accurate,
but
I
agree.
It's
it's
tricky
to
document
yeah.
B
You
will
you
will
notice
that
while
the
backup
was
running,
the
describer
was
showing
you
no
items
matching
the
spec
in
items
processed
and
at
the
end
of
the
backup.
Those
two
numbers
are
expected
to
match,
but
once
the
backup
is
completed,
I'm
just
showing
items
backed
up
so
that
there's
not
really
room
for
seeing
those
two
different
numbers
and
wondering
why
they
differ
like
once
the
backups
completed.
This
is
really
the
number
that
cares
about.
B
E
B
H
Quick
comment:
I
think
the
idea
of
the
number
changing
you
know
you
see
the
598
while
you're
watching
and
then
at
the
end
it
suddenly
discours
in
this
600
shows
I.
Think
that
actually
might
add
to
the
confusion,
I
think
I.
Think
if
you
leave
the
you
know
somehow
find
a
way
to
show
both
numbers
with
some
text
similar
to
what
you
have
there
explaining
why
they're
different
might
be
clearer
than
just
because
cuz
you
look
at
the
wrong
file
there.
B
That's
right,
you
know,
if
you're
not
doing
a
full
cluster
backup
if
you're,
if
you're
just
backing
up
a
namespace,
for
example,
then
yeah
the
PD
would
be.
You
know,
wouldn't
be
initial,
wouldn't
be
identified
in
the
initial
scrape
of
the
API
server,
but
would
end
up
being
backed
up
as
an
additional
item.
So
yeah
I.
E
B
Yeah
we
can
I
mean
yeah
I'm
happy
to
consider
all
these
ideas.
It's
like
it's
definitely
a
little
bit
tricky
to
to
make
perfect,
and
you
know
if
you're,
if
you're,
constantly
incrementing
the
total
number,
then
that
has
its
own.
You
know
its
own
confusion,
so
we
didn't
discuss
some
more
on
the
PR
but
I'm
happy
to
to
modify
things
to
whatever
we
think
is
the
clearest
yeah.
H
I
think
up
into
total
memory
that
seems
consistent
with
other
systems.
I've
seen
I've,
just
I
can't
I
can't
think
an
example,
but
you
know
you're
looking
at
progress
of
download,
for
example,
and
you
find
out
in
the
middle
of
your
download
that
it's
actually
bigger
than
you
thought
it
was
no.
You
can
see
that
updated
in
real
time
and
so
you'd
see
you
know
3
of
204
of
204
of
201.
You
know
I,
think
I
think
that's
fairly
clear.
Going
on.
Oh
ok,
we
had
more
than
we
thought
we
did.
That's
fine.
B
D
Yeah,
like
I've,
compared
this
to
like
the
old
windows
volume,
current
windows,
progress
bars
like
as
you're
scanning
a
directory
to
do
a
file
copyright
as
you
recurse
down.
You
may
not
have
the
full
count
of
things
so
that
could
change
and
your
your
speed
for
a
file
copy
may
change,
so
there's
there's
definitely
comparisons
that
can
be
made
to
other
systems,
so
we're
certainly
not
the
first
to
run
into
this
kind
of
problem.
B
C
A
D
It's
a
very
generic
document,
because
each
time
this
has
come
up
before
we've
kind
of
struggled
because
there's
a
lot
of
different
ways
you
could
set
up
in
here
today.
I
forgot
environment
and
we
can't
document
necessarily
every
single
object
store
so
like
the
one
we
picked
was
Mineo,
because
we
have
examples
for
that.
But
we
can
also
can't
document
how
to
set
up
a
production
video,
because
we
we
can
only
extend
ourselves
so
far.
D
D
I
know
Michael
had
talked
to
a
cloudian.
I
think
was
one
of
the
vendors
that
had
set
up
commercial
support,
there's
also
different
on-prem
plugins
for
volume.
Snap
shop
support
so
be
sweet
being
one
of
them
now,
so
the
the
struggle
has
always
been.
How
do
we?
How
do
we
cap
in
such
a
way?
That's
generic
enough
that
we
recover
enough
scenarios
but
specific
enough
that
it's
useful
and
I
don't
know
that
we've
ever
captured
that
balance?
D
Well,
the
document
I
link
to
is
what
we
have:
no
I,
don't
I,
don't
didn't
link
it
to
say
that
is
it's
good,
miss
Cheerilee,
but
it's
what
we
have
so
the
the
two,
the
two
things
that
are
really
necessary.
Are
you
have
an
object
store
with
again
that
is
compatible?
Oh,
it's
something
an
SP
compat
thing,
because
those
just
happen
to
be
my
internet
connections
in
stable.
Those
just
happen
to
be
what
what
are
replicated.
D
C
B
The
other
requirement
for
an
air-gapped
environment
would
be
an
image
registry
locally,
which
I
don't
think
we
cover,
because
it's
not
strictly
a
requirement
for
you
know.
Potentially
you
could
you
know
you
could
use
one
on
the
internet
even
if
you're
running
on
Prem,
but
but
that's
another
part
of
it.
Yeah.
D
A
A
I
C
B
A
F
J
Yeah
I
can
describe
the
scenario
so
for
one
of
the
use
cases
we
have
is
you
know
when
you
deploy
Valero?
We
actually
don't
create
that
default
certification,
and
this
is
because,
when
you're
running
backups,
you
actually
specify
you
know
we
create
a
backup
so
in
location
and
then
for
every
backup.
We
specify
the
location
of
the
BSL
that
we
want
to
run
the
backup
against.
So
in
terms
of
like
this,
the
BSL
liveness
readiness
probe.
So
since
we
deploy
bolero
without
creating
that
default
strategy
location.
J
You
know
in
this
case.
If,
if,
if
the
Valero
quad
will
crash
loop,
if
it
finds
there
is
no
storage
location,
then
you
know
it
will
be
hard,
for
example,
external
systems,
to
know
what
is
the
cause
of
that
crash
loop?
Is
it
because
of
there's
no
BSL
or
could
the
could
could
a
crash
to
be
due
to
some
error?
Maybe
the
plug-in,
maybe
there's
a
plugin
error
when
installing
bolero
or
just
some,
some
installation
error
that
occurred
during
the
deployment
of
valera.
J
If
you
could
have
the
letter
not
go
into
a
crash
loop
so
that
you
know
systems
that
Chuck
the
stairs
of
the
variable
part,
if
a
crash
the
polkas,
then
the
system
will
know
for
sure
this.
This
some
issue
with
Valero
and
that's
basically
to
just
to
know
for
sure
that
the
crash
group
is
not
due
to
the
BSL,
but
some
other
error
condition.
I.
J
Yeah,
just
word
and
I
think
you
know
if,
as
a
cubed
mean,
it
could
be
easy,
as
Nolan
mentioned
in
the
inter
and
the
issue
that
you
know,
cubed
min
could
just
look
at
the
logs
and
if
they
see
that
the
crash
strip
is
because
there's
no
BS,
then
you
know
it
could
be.
The
user
could
easily
know
that
okay
I
just
need
to
create
a
BSL
and
then
with
a
few
a
few
seconds
and
Valera
will
be
back
up
and
running.
J
B
I
But
so
I
was
thinking
while
we
were
describing
we
could.
Instead
of
having
to
look
at
logs
or
whatnot,
we
could
set
the
status
field,
so
a
user
could
just
interrogate
his
status
and
know
for
sure
what
yeah
we
actually
liked,
heuristic
that
in
Antony
described,
but
then
I
thought
he
said
something
I
thought
you
said
something
about
status
at
the
end.
F
I
Forgive
me
for
not
knowing
the
answer
to
this
question,
but
is
not
having
a
default
set,
truly
and
not
ready
state,
because
they
have
to
describe
his
use
case,
which
is
always
specifying
the
backup
storage
location
along
with
backup,
which
means
there's
a
way
to
use
the
application
without
a
default
BSL.
So
I
guess
my
question:
is
this
crash?
Look
even
does
that
even
make
sense
in
this
case,
since
there
are
ways
to
use
it.
J
That
is
a
valid
question:
yeah
right
and
I,
just
like
add
more
in
the
use
case.
So
for
us
you
know,
as
I
mentioned,
we
deploy
every
letter
without
a
default
location,
but
before
we
run
any
protection
we
actually
go.
First,
we
first
create
a
you
know,
valid
back
every
location
and
then
now
we
use
that
with
the
backup.
J
D
The
other,
the
other
thing
we
were
trying
to
do
here
was
unify
some
of
the
behavior.
So
right
now,
velaro
halts
the
the
program
halts
when
it
doesn't
have
a
default
on
startup,
and
if
you
let's
say
you
have
a
default
on
startup,
it
starts
up,
and
then
you
delete
all
your
your
backup
storage
locations.
D
Valera
will
keep
running.
So
we
wanted
to
make
sure
that
in
both
those
cases,
whether
they're,
whether
it
was
already
running
and
you
deleted
all
your
get,
your
backup
storage
locations
or
you
had
no
storage
location,
backup,
storage
locations
and
you
just
started
a
wouldn't
make
sure
it
behaved
the
same
whatever.
Whatever
that
decision,
1
is
1
make
sure
it's
behaved,
the
same
the
the
question
that
you
had
Tim
about.
Well,
there's
valid
way
to
use
it
without
that's.
D
D
If
there
is
a
single
backup,
storage
location,
then
we
assume
that's
the
default
I
think
in
Anthony's
case
it's
when
there's
zero
locations
at
all
it
right,
Anthony
like
if
I.
If
I
remember
your
system,
it's
at
very
first
install,
there's
no
locations
and
then
at
first
backup
a
location
will
be
created
and
then
the
backup
will
happen
right,
right,
I
think,
ok,.
H
E
D
Also
wanted
to
answer
Anthony's
question
on
the
the
issue:
Anthony
was
asking
about
the
liveness
check,
whether
that
would
be
exposed
through
a
service
or
anything
liveness
checks
in
my
interpretation.
I
may
be
incorrect
here,
but
liveness
checks
from
my
understanding
are
meant
for
the
cubelet,
so
they're
only
exposed
on
the
node
and
it's
for
the
cubelet
to
ask
the
program:
hey.
Are
you
available
and
able
to
do
stuff?
D
D
I,
don't
I,
don't
know
if
that's
a
pattern
that
expects
that
others
that
other
programs
expose
it
into
the
cluster
I'm,
just
not
sure
there
but
I.
My
intention
with
it
was
just
to
use
it
with
the
cubelet,
but
we
could
also
expose
it
through
if
we
plumbed
it
through
to
a
service
and
said:
hey
external
services
hit
this
endpoint
to
say
to
ask
Valero.
Are
you
ready
to
do
stuff
that
might
be
a
solution.
I
D
F
Yeah,
my
original
thinking,
big
but
I
I,
am
winning
to
the
first
people
who
know
the
patterns
of
kubernetes
applications.
You
know
I'm
a
proficient
at
that
than
I
am
so.
My
original
thinking
was
put
up
for
aro
then
start
creating
things
I
had
to
plug
in
at
the
backup
location,
and
then
you
can
create
backups.
So
in
that
case
the
server
would
be
up,
but
but
like
it
seemed
just
touched
on
it.
You
know
maybe
well
the
the
pattern
should
be
just
make
it
very
clear.
F
E
D
That
was
kind
of
my
intention
with
proposing
the
crash
loop
stuff,
like
it's
very
clear.
If
your
class
crash
looping,
it's
very
clear,
it's
not
it's
not
ready
to
do
stuff,
but,
as
Anthony
brings
up
crash,
looping
isn't
necessarily
without
looking
at
kubernetes
events.
It's
not
necessarily
clear
what
the
cause
of
crash
looking
is
so
you'd
have
to
look
at
cube.
Ctl
describe
to
say:
is
it?
Is
it
because
it's
an
image,
pull
back
off?
Is
it
because
we
don't
have
a
plugin?
Is
it
because
we
don't
have
BSL?
D
E
F
J
C
J
Crash
looping
because
there
is
no
BSL,
but
what?
If
there's
some
other
issue?
Because
for
us
we
also
install
plugin
for
the
object,
store,
plugin
and
the
way
we
you
know
we,
you
know
all
this
is
in
automatically
deployed
by
the
system
and
so
the
first
one.
Once
after
we
deploy,
we
check
that
the
villa
report
status
is
up
and
running,
so
I
cannot
that
everything
was
installed
and
the
plugin
works
yeah
and.
D
I
was
thinking
if
we
sent
a
kubernetes
event
its.
These
aren't
super
long-lived,
but
if
we
sent
a
custom
event
that
said,
Valero
no
BSL
or
something
like
that
to
me
indicate
for
classroom
we're
crash
sleeping
because
there
are
no
event.
There
are
no
vs
else.
That
would
be
I,
think
that's
kind
of
a
kin
to
what
Tim
was
saying,
with
put
a
status,
to
give
a
clear
indication
that
there
are
no
BS
else.
This
is
why
we're
actually
paying
so
it
so
you're,
not
relying
on
just
looking
at
a
heuristic.
D
E
D
I
just
want
to
say
like
if
you
look
at
if
you
use
the
heuristic
to
say
oh
you're
crash
looping
and
we
see
there's
zero
ESL's,
that's
not
great,
because
just
because
there's
zero
be
yourself
doesn't
mean.
That's
the
same
reason:
your
crash
looping,
it's
a
guess
but
I'm
starting
to
lean
towards.
Maybe
we
don't
crash
loop
forward,
no
BS
else.
D
I
I
I
Problem,
it's
a
backup
problem,
so
life
cycle
is
different
for
this
error.
So
what?
If
the
status
of
the
Valero
pod
and
going
back
to
status,
that's
used
pretty
heavily
and
contoured
I
think
to
good
effect.
Anyway,
if
you
can
interrogate
the
Valero
pas
or
pods
for
to
basically
enumerate
the
list
of
backups
that
couldn't
start
up
because
they're
missing
something
because
it's
not
the
Valero
is
broken,
it's
that
these
backups
individually
or
these
scheduled
backups
or
whatever
don't
have
all
of
their
prerequisites.
It's
really
stretching
the
metaphor.
E
Yeah
I'm,
starting
to
question
the
dirt
crash.
Looking
is
that
I
think
to
do
here
because,
usually,
like
the
controllers
are,
are
far
like
long-lived
objects
in
communities
where
you
have
like
a
con
fake
and
you
have
the
controller
working
to
or
you
have
a
spec.
And
then
you
have
the
controller
trying
to
reconcile
respect
to
status.
But
in
the
case
of
backup
storage
locations
it
is,
it
doesn't
have
to
be
long
and
it
can
be
like
you
can
have
like
short-lived
backup
storage
locations
that
are
valid
just
for
this
backup.
E
So
maybe
maybe
we
can
just
at
the
beginning
of
the
backup
we
can
verify
whether
do
I
have
a
backup,
storage
location
where
and
store
this
backup
thing.
And
if,
if
you,
if
there
is
no
backup
storage
location
that
you
can
use
for
this
backup,
you
update
the
status
of
the
backups
backup
object
and
just
takes
it
and
fill
the
backup.
E
If
you,
if
you
think
about
like
I
know,
we
don't
have
multi-tenancy
right
now,
but
in
the
future,
when
we
decide
to
have
multi-tenancy,
it
may
be
possible
that
some
user
in
a
multi
some
tenant
to
the
market
Annette
world
and
they
have
a
invalid
backup,
storage,
location
or
a
backup
storage
location
that
they
can't
use.
At
that
point
crash
the
whole
little
installation
or
David
I.
I
J
Yeah
yeah
I
think
yeah.
That's
what
meter
emits
because
you
know
once
we
run
the
backup,
we
check
the
we
describe
the
backup.
So
if
there's
some
error
with
the
backup,
we
should
be
able
to
pick
it
up
from
the
the
back
upstairs
it's
just
when
we
deploy
bolero,
we
don't
configure
that
default
BSL,
and
so,
if
you
crash
looping,
it
makes
it
hard
to
really
know
what
the
crash
strip
is.
If
there's
some
other
header
condition.
D
D
Totally
get
what
if
she
was
saying
like
in
Fusion
timber,
saying
like
it's,
the
actual
concern
is
at
backup
time
the
the
reason
the
decision
was
made
was
because
is
we
wanted
to
say,
hey
person
installing
glare.
Oh,
maybe
you
want
to
install
a
backup,
storage
location
now,
while
you're
doing
the
whole
setup
versus
hey
I,
set
a
schedule
and
walked
away,
and
then
you
come
back.
Who
knows
when
and
just
your
backups
weren't
running,
but
in
Antony's
case
it's
different.
F
So
this
is
my
understanding.
I
think
we
are
in
the
consensus
here.
We
are
not
going
to
have
a
controller
for
BSL.
We
are
going
to
change
the
behavior
of
a
little
to
not
check
for
BSL
we're
not
going
to
crash
loop.
Anything
we
are
going
to
just
add
a
validation
to
the
backup
process,
to
check
I
think
we
may
already
have
that.
I,
don't
know
to
check
if
there
is
a
corresponding
base
or
not
in.
There
is
no
such
thing.
F
B
B
You
know
periodically
check
that
it
can
be
connected
to
and
have
some
indication
of
status,
but
then
we'd
want
to
make
sure
that
when
we
run
them
back
up,
we're
probably
retrying
that
check.
So
you
know
I,
don't
know
exactly
what
it
would
look
like,
but
I
I
wouldn't
necessarily
discard
the
controller.
Just
because
we're
gonna
do
the
validation
at
backup
time.
Okay,.
F
A
B
And
I
just
grabbed
from
valera
core
and
the
plug-in
repo,
so
these
were
both
from
Valero
core,
so
Ragavendra
who
got
a
PRN
to
help
clarify
some
of
the
wording
in
the
Valero
restore
describe
command
around
what
included
namespaces
mean.
So
this
is
because,
on
a
restore
you
know
if
you're,
including
all
namespaces,
it
really
means
all
namespaces
that
are
in
the
backup.
So
so
we've
got
some
clarification
there.
So
that's
that's
really
helpful
for
users
who
are
looking
at
that.
B
Thank
you
and
then
my
uncle
so
got
a
PR
and
to
basically
skip
any
validation
of
Vol
snapshot.
Locations
for
backups
when
snapshots
are
explicitly
disabled.
So
this
would
be
a
use
case
where
you're,
maybe
just
don't,
want
snapshots
or
you're
using
rustic
to
backup
your
volumes,
and
so
we
just
want
to
ignore
the
state
of
the
volume
snapshot
occasions.
So
it's
that's.
Definitely
another
usability
enhancement.
That's
that's!
Gonna,
make
things
a
little
bit
easier.
So
thanks
a
lot
monk.