►
Description
April 14,2020
Status Updates
[carlisia]
reviewing + testing of Helm PRs
Broke down new install CLI work into issues, see epic: https://github.com/vmware-tanzu/velero/issues/2419
wip: bsl/vsl/velero readiness: https://github.com/vmware-tanzu/velero/issues/1967
presenting/demo’ing Velero at a CNCF (online) meetup: https://www.meetup.com/Cloud-Native-Talks-Brazil/events/269858787/
[steve]
community support this week
lotta PR reviews
did some testing with WIP CSI plugins, reported a couple bugs
bunch of cleanup on the https://github.com/vmware-tanzu/velero-plugin-example repo, hoping to apply similar changes to the AWS/Azure/GCP ones too
continued on backup progress idea
[nrb]
CSI upload PR should be done & ready for review now
Tackling some CRD issues this week, mainly https://github.com/vmware-tanzu/velero/issues/2383 & https://github.com/vmware-tanzu/velero/issues/2370
Started working on https://github.com/vmware-tanzu/velero/issues/1970, but ran into https://github.com/vmware-tanzu/velero/issues/2429
Discussion Topics
[nrb] When do we increment the backup format version? Refer to this comment
PR that added backup version: https://github.com/vmware-tanzu/velero/pull/1117
AI - Nolan to document when we increment backup format versions (most likely following semantic versioning)
Current PR changes are backwards compatible, so they’re a patch version in semantic versioning parlance
[dymurray/sseago] Race condition with AdditionalItmes https://github.com/vmware-tanzu/velero/issues/1350
We haven’t seen others asking for this feature, but willing to let the RH team lead with a design doc on this
[nrb] I see this as a generically useful tool, +1 to RH taking the lead here
[dymurray/sseago] Restoring PVs that already exist (related https://github.com/vmware-tanzu/velero/issues/2170)
[nrb] Is this a specific case of https://github.com/vmware-tanzu/velero/issues/1977?
data-only restores https://github.com/vmware-tanzu/velero/issues/504
Could the PV rename mechanism work here? I was thinking of the StorageClass remapping
[carlisia] should “liveness” mean “ready”, and should we use a liveness check to check for readiness: https://github.com/vmware-tanzu/velero/pull/2382
Kubernetes definitions of readiness/liveness - https://kubernetes.io/docs/tasks/configure-pod-container/configure-liveness-readiness-startup-probes/
[marccampbell] Willingness / timeline for custom restore hooks? https://github.com/vmware-tanzu/velero/issues/2116
[mayank] patch release for plugin velero.io/change-pvc-node-selector
Contributor Shoutouts
@duyanghao - docs on community project to automatically annotate all volumes for restic backup
A
Hi
everyone
and
welcome
to
the
April
4th
2020
version
of
the
Bolero
community
meeting,
slash
open
discussion.
So
today
we
get
a
bunch
of
status
updates
per
usual.
If
you
have
any
questions,
please
write
them
in
the
chat
or
just
unmute
yourself,
and
if
you
have
any
discussion
topics
we
do
have
a
few
already
today,
so
we'll
try
to
get
through
those
as
soon
as
possible.
I'll
be
sharing
my
screen.
So
you
can
all
see
the
agenda
for
today
as
well.
B
Okay,
so,
yes,
oh
sure,
I
forgot
to
collect
the
PRS
from
home,
Oh
Adam
next
week,
because
I
think
some
of
them
are.
We
have
a
cluster
of
some
that
are
fixing
things
and
add
them
next
week,
but
in
any
case
we
had
some
activity
that
I
was
reviewing
and
testing
on
the
helm.
Repo
I
took
that
CLI
design
proposal
for
configured
to
change
the
installation.
You
can
break
it
down
into
two
separate
commands
and
I
added
different
issues
for
each
of
the
things.
B
Oh
yeah,
that's
the
the
first
thing
that
I'm
going
to
tackle,
which
is
to
change
the
behavior
of
an
arrow
to
dot
halts
the
server.
If
there
is
there
is
any
way
so
right
now,
if
there
is
even
one
invalids,
backup,
storage
location,
so
that
I
will
not
like
it
in
not
work.
So
we're
going
to
change
that,
so
filler
continues
work
and
continues
checking,
and
if
you
have,
if
you
don't
have
any
valid
backup
location,
the
worst
that's
going
to
happen
is
you
can
create
backup
both
of
ours
will
continue
working.
You
continue
tracking.
B
So
as
soon
as
you
have
one
that
is
valid,
it
will.
Let
you
make
backups,
and
the
advantage
of
that
too,
is
that
right
now,
if
you
have
two
to
backup
locations
and
no
one
is
invalid,
valera
will
not
even
let
you
use
the
other
one
because
they
just
won't
stop
working.
So
if
you
so
after
this
change,
even
if
you
haven't
any
invalid
back
applications,
you
can
still
use
any
one.
B
That's
valid,
and
in
part
of
that
is
I
added
a
topic
for
discussion
in
in
other
section,
because
we
might
be
able
to
use
some
the
work
that
Nolan
added.
He
has
a
PR
to
add
a
liveness
check
for
health
health,
each
healthiness
check
health,
let's
check
or
libelous
check
to
Valero,
and
that
might
be
useful
for
this.
B
But
we
can
yeah.
Let
me
just
finish
my
studies.
Oh
and
I'm
doing
a
presentation
for
Valera
today
at
this
meetup
in
Brazil.
Originally
it
was
going
to
be
kubernetes
Community
Day,
which
is
the
the
equivalent
of
cube
corn
in
different
regions
of
the
world,
because
they'd
seen
the
CNC
F
is
not
going
to
have
cube
corn
chip
cons
all
over
so
because
of
the
virus.
They
turned
into
an
online
Meetup
and
that's
indeed
the
other
presenter
is
also
someone
involved
in
Valero
Carlos
Patenaude.
So
so
that's
pretty
cool.
A
B
B
A
D
That
they're,
the
weird
thing
is
that
I've
actually
tried
to
do
a
technical
talking
Greek
one
time
it
was
the
3
PS
thing
ever,
because
all
the
terms
and
all
the
things
you
know
are
in
English,
but
then
you're
trying
to
explain
them
in
a
different
language.
He
was
creepy.
He
took
me
twice
as
long
to
prepare
because
I
had
to
figure
out
what
I
was
gonna
say
in
certain
situations.
So
I
appreciate
how
hard
this
good
job
kaliesha.
D
A
E
Yeah
everyone
community
support
this
week
definitely
take
up
some
of
my
time
over
the
past
week
done
lots
of
PR
reviews.
There's
lots
going
on.
As
far
as
the
CSI
work,
the
work
Carlie
is
doing
on
the
install
and
configure
and
various
other
things,
though
this
definitely
kept
me
busy.
I
did
find
some
time
last
week
to
do
some
hands-on
testing
on
the
CSI
snapshot.
E
Plugins
that
are
under
development
and
things
are
starting
to
look
pretty
good
they're
starting
to
sort
of
stabilize
I,
did
find
a
couple
of
bugs
that
I
logged
into
the
repo
so
potentially
we'll
find
some
time
this
week
to
try
and
help
out
with
some
fixes
there
and
still
plenty
more
testing
to
do
as
well.
So,
if
I
find
time,
I
will
continue
working
on
that
I
spent
some
time
doing.
Some
cleanup
on
our
plugin
example,
repo,
which
was
sort
of
in
need
of
some
maintenance.
E
D
A
A
E
F
E
B
F
F
So
that
may
be
something
we
could
investigate,
but
I
don't
want
to
take
up
a
ton
of
time
with
that
today.
I'm
just
gonna,
just
gonna
mention
that
that
maybe
we
could
make
this
a
template
repo,
but
I,
don't
know
what
that
entails.
Like
I,
don't
know
what
effect
that
would
have
on
that.
The
repo
itself
so
is.
F
Well
effectively,
what
we
asked
people
to
do
is
check
it
out
and
then
use
it
as
a
template
and
delete
code
and
write
their
own
stuff
in
it.
So
but
I
don't
know
what
the
github
like
I,
don't
know
how
the
the
template
workflow
works
within
github.
So
it
may
be
that,
like
you,
click
fork
and
then
it
does
a
bunch
of
templating,
but
I
would
have
to
look
into
how
that
works.
A.
F
F
Of
course,
I,
muted,
myself
and
lost
a
button,
so
I
got
the
CSI
structures
getting
uploaded.
That
should
be
ready
and
done
ray
for
review
now.
The
only
issue
I'm
aware
of
is
that
is
not
uploading.
Volume
snapshot
contents
and
that's
because
the
plug-in
is
not
labeling
them
right
now,
which
is
a
pretty
small
fix
that
we
just
need
to
make
in
the
plugin.
F
So
that's
ready
for
review
and
with
the
comments
on
screen,
I
think
I've
addressed
the
last
of
the
issues
that
Steve
had
raised:
cleaned
up
some
duplication,
some
of
those
functions.
There
are
some
dues
kind
of
sprinkled
throughout
the
code,
but
I
don't
think
any
of
them
are
crucial
to
getting
the
core
functionality
of
this
PR
done.
F
So
that's
finished
since,
since
I
got
that
kind
of
wrapped
up
in
and
and
am
waiting,
I
moved
on
to
looking
at
some
of
these
CRD
issues
that
have
been
reported,
these
two
were
filed
by
the
same
person
and
they
kind
of
split
them
up
so
I've
got
some
I've
got
a
couple.
Reproduction
cases
that
I
just
got
in
last
night,
so
I
need
to
take
those
and
and
actually
reproduce
them
now
and
while
I
was
waiting
for
that,
I
started
it
as
part
of
our
our
one
for
backlog,
I
started.
F
Taking
a
look
at
this.
This
restoring
a
namespace
resource
issue
is
this
bug
came
up
basically
because
this
user
wanted
to
specifically
restore
only
the
name
space
where
velaro
currently
just
skips
the
name
space
and
assumes
nobody
wanted
wants
to
restore
a
name
space
by
itself.
They
always
want
to
restore
namespaced
resources,
so
I'm
investigating
how
we
might
be
able
to
make
this
more
useful
and
in
the
course,
if.
D
E
D
F
And
and
I
think
what
I
what
I
wanted
to
see
was?
Can
we
can
we
accommodate
both
cases
both?
Do
our
lazy
restore
that
we're
doing
now
and
treat
the
namespace
like
we
would
any
other
resource
and
in
my
attempts
to
do
this,
I
actually
ran
into
another
bug.
So
I
put
this
on
pause
and
moved
on
to
another
bug
which
there's
PR
out
for
right
now.
This
is
only
master.
There's
no
released
code
with
this,
but
found
if
kubernetes
minor
versions
have
plus
symbols
in
them.
F
E
F
Yeah
and
and
I
think
this
was
I
think
this
was
on
gke,
so
I
think
they
had
their
own
compiled
version
like
you
can
see
there
they're
using
one
16.8
and
and
I
think
my
PR
may
need
a
tweak.
I
think
this
is
only
at
the
beginning
and
end
so
the
error.
The
error
message
specifically
said
that
you
can't
have
non
alphanumeric
characters
at
the
beginning
or
end
and
in
the
in
the
de
kubernetes
validation
package.
F
There
is
a
is
valid
label,
I
think
it's
function
name,
but
it's
basically
checking
is
valid
label
value,
not
just
label
name,
but
there's
a
there's
a
function
to
check
if
it's
a
valid
label
value,
so
my
PR
basically
adds
a
Valero
version
of
that
and
tries
to
for
internal
Valero
use.
It
tries
to
find
the
most
appropriate
value.
F
So
that's
what
I
did
in
my
PR,
but
if
an
annotation
works
for
our
use
case
here,
then
we
can
move
it
to
an
annotation,
but
right
now
this
is
not
released
code
and
it
doesn't
happen
on
all
platforms.
Not
all
platforms.
Add
this
plus
at
the
end,
so
yeah,
it's
it's
just!
This
just
happens
with
labels
that
Valero
is
generating.
E
Well,
yeah:
let's,
let's
check
out
the
annotations,
my
impression
is
that
they're
more
that
they
are
more
free
form
allowing
yeah
and
yeah.
It
may
just
make
sense
for
this
information
anyway,
since
we're
probably
not
using
it
to
filter,
but
we
will
check
it
out
see
if
that,
if
that's
a
possible
solution
as
well,
and
then
we
can
just
kind
of
how
to
move
forward
yeah.
F
Yeah,
if
that
fits,
if
that
fits
the
use
cases,
then
it
might
just
be
easier
to
move
those
into
an
annotation
Andrea
and
taught
us
this
valid
label
value
I.
Think
the
the
label
value
validation
may
make
sense,
because
right
now
we
are
doing
some
validation
on
label
values,
but
we're
using
the
label
name
function,
but
we
can
discuss
that
separately.
F
F
Think
the
only
change
I
want
to
see
in
this
PR
at
this
point
is
that
we
bump
the
version
of
the
back
up
format
to
be
one
point:
one,
because
it's
backwards
compatible,
but
it's
not
1.0
any
one
or
one
and
basically
I
just
wanted
to
bring
this
to
the
attention
of
the
core
team
and
carlesha
asked
the
great
question:
what
triggers
a
bump,
because
I
don't
know
that
we've
ever
laid
that
out,
because
this
is
different
from
the
Bolero
program
version.
It's
the
data
format,
so
I
wanted
to
bring
this
up
and
resolve
it.
F
E
F
E
I
honestly
think
it
was,
it
was
further
back
than
that,
but
I'm,
not
totally
sure
I
mean
I.
You
know
I
think
it
probably
makes
sense
to
do
a
minor
version
bump.
We
might
want
to
move
to
an
actual
like
summer
version,
but
but
I
don't
know,
but
since
it's
there
I,
you
know
I,
don't
think
it
would
hurt
to
bump
it,
but
we're
not
using
it
for
anything
right.
F
Now,
yeah
and
I
think,
like
the
reason
I'm
I
mentioned
it
here,
was
for
documentation
just
to
be
explicit
and
not
use
like
recently
in
the
docs,
because
recently
could
be
like
when
I'm
reading
this
a
year
from
now.
What
does
recently
mean
if
we
could
point
to
you
specifically
in
backup
format,
v
1.1?
This
is
when
it
changed.
F
So
maybe
we
just
say
like
when
we
if
we
can
come
up
with
some
rules
around
when
we
bump
the
diversion
I,
think
that
would
make
sense
to
me
this.
This
seems
like
a
good
time
to
bump
the
version,
but
it's
since
in
terms
of
semantic
versioning,
it's
backwards
compatible,
but
it's
adding
to
it
makes
sense
to
be
a
point
release,
not
full
major
format.
F
E
And
I,
you
know
I
think
we.
This
could
be
useful
in
the
future.
If
we
have
a
version
of
velaro,
that's
trying
to
deal
with
multiple
sort
of
different
historical
formats
and
it's
you
know,
at
least
with
the
minor
version
bumps.
It's
like
a
nice
shorthand
way
of
knowing
what
information
is
expected
in
the
tarball
rather
and
having
to
unpack
the
tarball
and
go
like
look
at
the
contents
and
sort
of
deduce
what
what
format
version
it
is.
E
So
it
could
definitely
be
useful,
and
you
know
it's
easy
to
easy
to
increment
it
for
now
and
have
that
data
in
all
the
the
backups
going
forward.
In
case
we
decide,
we
want
to
use
it
down
the
road,
so
I
think
it
makes
sense.
It
would
be
great
if
someone
could
maybe
take
a
stab
at
just
like
documenting
it
and
documenting
when
we
would
increment
these
things
so
that
we
can
at
least
have
some
information
out
there
about
it.
I
can.
G
G
Well,
I
guess
it
was
a
long
time
yet,
but
we
kind
of
have
been
circling
back
to
it
about
a
month
ago,
due
to
some
issues,
we've
been
hitting
in
our
migration
tool
and
it
has
to
do
with
passing
additional
items
back
on
a
restore
action
and
there's
a
race
condition
that
if
the
item
that
you
are
passing
in
as
an
additional
item,
easy
be
restored
first,
you
know
there's
no
waiting
or
blocking
in
Valero.
So
the
actual
item
they're
trying
to
restore,
may
fail.
G
If
some
items
aren't
there
so
I
know
there
was
some
talk
on
the
issue
that
we
tend
to
agree
with
about
a
potential
weight
function
being
used
on
the
plug-in.
This
way
that
you
know
you
kind
of
define
what
what
logic
you
want
to
take
in
terms
of,
if
an
additional
item
is
passed
and
the
restore
action,
how
X?
How
exactly
do
you
know
when
the
item
is
ready
and
when
you
can
stop?
This
is
waiting
and
blocking
function.
I,
don't
want
to
get
into
like
a
big
design.
G
G
I
mean
it
sounds
like
we
might
be
the
only
people
this
is
actually
biting
in
mainly
because
you
know
we
are
working
with
open
shift
and
we
have
some
open
chef
resources
that
care
more
about
this
than
others,
and
we
are
in
the
unique
case
of
generally
restoring
across
clusters,
so
we
don't
always
have
the
benefit
of
some
existing
resources,
even
if
they're
cluster
scope
turn
on
cluster
scoped
already
being
ready
on
restore
time.
So
we
are
one
of
the
few
people
that
this
is
really
biting.
G
E
E
G
Great
this
next
issue
that
Scott's
got
to
bring
out.
It's
probably
gonna,
take
precedence
over
this,
so
I
don't
know
when
we'll
actually
had
a
design
up
on
it,
but
I'll
make
a
comment
on
the
issue
that
you
know
will
circle
back
and
and
critias
on
doc.
First,
before
going
at
blending
any
changes,
I'm
Scott
you
want
to
take
over
the
next
one
sure.
H
So
an
issue
that
I
run
into
when
I'm
just
gonna
step
back
is
that
one
of
the
things
that
we're
doing
with
our
migrations
is
is
trying
to
allow
for
PV
copies
kind
of
an
incremental
migration.
So
we'll
really,
we
might
do
a
stage
migration
where
we
copy
the
PBS
over
but
not
shut
down
anything
on
the
source
cluster
and
then
later,
when
we're
ready,
she
makes
it
chained.
We
did
the
final
migration,
which
will
then
make
another
PV
copy
and
with
rustic.
This
is
pretty
much
working
for
us,
but
I
realized
with.
H
Snapshot
copy-
it's
actually
not
working
in
this,
because
when
we
try
to
do
that.
Second
migration,
the
said
the
second
restore
the
difference
between
rustic
and
snapshot,
is
that
the
actual
PV
is
restored
with
snapshot
rather
than
it
being
this
car
didn't
represent
reprovision,
and
so,
since
Valero
is,
you
know
seeing
ok,
the
PV
PV
C
exists
already,
so
we're
not
gonna
copy
them
in
we
get
the
warning
saying.
Cluster
version
is
different
than
backup
and
I
think.
H
The
result
there
is
that
the
plugin
for
the
EBS,
for
example,
database
plugin,
is
not
running
so
we're
not
getting
that
updated
PV.
The
issue
that
doing
linked
in
there
isn't
exactly
the
same.
I
think
it
looks
like
it
was
a
case
where
someone
did
it
restore
with
rustic.
Some
files
try
to
restore
again
and
it
wasn't.
It
wasn't
updating
so,
but
there
was
a
comment
saying
this
is
an
issue
for
snapshots
as
well,
so
I
think
I,
don't
know.
This
is
an
issue
that
others
who
run
into
I
know.
H
There's
a
larger
question
of
how
do
you
handle
restores
if
the
resource
already
exists-
and
this
is
kind
of
a
subset
of
that
I
guess.
But
you
know
it's
a
different
case:
I,
don't
know
if
it's
something
that
makes
sense
for
Valero
to
make
an
exception
for
for
PBS
and
figure
out
a
way
to
make
this
take.
There's
a
snapshot,
restore
happen,
I'm
just
not
sure
what
the
right
answer
is
but,
and
it
was
right
now
impacting
our
incremental
restores.
You
know
when
were
using
snapshots.
A
H
Right
exactly
and
what
we
want
to
happen
would
probably
be
to
have
I.
Don't
know
this
means
we
would
need
to
either
inside
or
outside
of
velaro
delete
the
TV
and
have
it
restore
again
or,
if
there's
a
way
to
just
trigger
the
restore
within
the
PV
I'm,
just
not
sure
what
the
again
the
wafer
flare
to
do.
This
I.
B
H
H
F
E
H
And
that
would
be
with
with
the
PV
possibly
already
existing,
or
is
this
the
specific
image
that
we're
hitting
is
where
you
know
if
the
PV
doesn't
already
exist?
It
works
fine,
but
if
the
PV
already
exists
with
a
previous
snapshot,
then
it
doesn't
because
basically
you
know
blur
is
just
discard,
the
PV
doesn't
restore
it,
and
so
then
we
don't
the
plugin
doesn't
get
triggered
to
to
run
the
you
know:
copy
snapshot
to
volume
and.
E
E
H
H
Gonna.
Add
some
new
files.
Do
a
second
migration
to
the
Andal.
Do
the
backup
and
does
the
rest
of
your
store
and
in
that
case,
rustic
will
update
the
existing
volume
was
in
your
data,
but
the
way
the
plugins
work
for
snapshots
that
doesn't
happen
because
because
the
PPA
already
existed
or
version
of
it,
so
we
don't
get
the
plug-in
doing
anything.
Yeah.
F
F
A
Main
issue
that
I
see
is
that
the
restored
PB
might
have
been
used
before
you
do
the
next
incremental
update
right,
so
we
would
have
to
have
the
the
snapshots
need
to
be
at
the
same
time
on
both
sides
at
the
same
time
frame,
because
otherwise
you
might
have
changes
and
in
the
restored
PV,
and
you
can't
overwrite
those
correctly
with
the
physical
block
snapshot
that
you
have
on
the
original
I.
Think.
F
H
Well,
I,
in
this
particular
case
you
know
just
because
the
way
we're
doing
the
migrations
it
doesn't
happen
to
be
in
use
in
the
sense
that
you
know
it's
kind
of
it's
almost
like
the
stage
migration
is
kind
of
pre
filling
PV
data.
So
the
idea
being
you
know,
I
might
run
a
staged
migration
today
to
copy
over
all
of
our
data
and
then,
when
we're
using
rustic
at
least
we
then
tomorrow
do
the
actual
migration
shut.
Everything
down
and
rustic
is
checking
this.
You
know,
so
it's
only
cotton
is
so
so
with
rustic.
H
We
had
we
had
made
upstream
yeah,
but
we're
using.
Where
is
basically
when
it's
when
rustic
is
copying
the
file,
it
actually
checks.
If
the
file
exists
already
in
the
two
or
the
same
file
incremental
migration
copy
there,
obviously
with
snapshot.
We
don't
have
that,
but
at
the
very
minimum
you
know
the
hope
would
be
that
if
I
did
that
staged
migration
yesterday
and
today,
I'm
doing
the
final
migration,
including
you
know,
all
the
running
applications
I-
would
want.
H
F
H
F
H
And
in
fact,
the
way
we're
currently
doing
it,
that
pot
is
even
that
initial
pod,
because
when
we
do
that
initial
migration
of
PB
data
we're
actually
deleting
those
pods
afterwards,
anyway,
just
leaving
the
PD.
So
it's
actually
not
even
mounted
at
that
point
and
then,
when
we
do
a
second
migration,
essentially
repeating
the
same
process
as
we
did
the
first
time
you
know,
maybe
the
next
day
were
again
for
rustic
we're
just
copying
those
files
that
have
changed
but
with
and
with
snapshots
again.
H
It
wouldn't
expect
the
incremental
part
to
happen
in
the
same
way,
but
at
the
very
least,
I'd
hope
to
be
able
to
get
the
new
new
bits
over
and
just
right
now,
that's
not
happening
because
we're
exists,
so
we
discard
it
and
I'm.
Assuming
because
of
that,
the
volume
snapshot
plugin
is
not
being
called
in
the
context
of
this
so
usually.
A
You
use
well
when
you
move
data
that
way,
you
usually
use
a
replication
mechanism,
not
a
backup
and
restore
mechanism.
Those
backups
are
supposed
to
be
here's
the
the
data
and
here's
the
restore
and
file
file.
Backups.
Of
course
they
can
be
incremental.
That's
easy,
yeah
block
backups
are
essentially
here's
the
big
chunk,
oh
yeah,.
H
I
get
that
and
I'm
not
saying
you
know.
I
expect
this
to
go
work
in
increments
away,
like
you
said,
but
what
I
would?
What
I
would
like
to
see
is
that
you
know
that
it
would
work.
You
know
that
you
would
get
the
updated
copy
at
the
end
and
the
answer
may
be
to
since
snapshots
are
not
incremental
anyway.
The
answer
here
may
be
not
to
only
do
the
rest
of
copies
for
the
first
one,
because
the
expectations
there's
a
mixture
of
rustic,
and
you
know
reasonable.
H
We
can
a
drastic
depending
on
where
we
can,
and
the
answer
may
be
to
just
not
include
the
snapshot
copies
in
these
incremental,
because
we
wouldn't
really
be
getting
that
incremental
benefit
anyway,
but
I
also
just
wanted
to
bring
it
up
because
I
didn't
know.
If
these
cases
were,
you
know,
if
we're
doing
a
restore,
and
we
have
a
snapshot
here
and
the
that
PV
exists
already.
A
A
If
you
don't
know
what
you're
doing
and
you're
trying
to
restore
things
incrementally
over
another
physical
block
device,
things
can
go
very
very
wrong,
so
I
would
be
very
careful
opening
up
that
optional
flag,
I'm
sure
it
could
be
doable
but
yeah.
The
way
for
incremental
migrations,
like
that,
the
rustic
overusing
file
that
works
or
replication
mechanism
that
works
to
just
my.
H
H
A
H
A
F
That's
not
a
very
helpful
comment.
I
think
everybody
agrees.
It
would
be
great
yeah
it.
It
just
requires
more
design
because,
like
like,
we
do
get
that
question
on
support.
I
would
say:
I
would
say,
I
don't
know.
Every
few
weeks
is,
is
velaro
incremental
and
we
have
to
be
like
well.
What
do
you
mean
by
that?
Because
it's
not
incremental
at
all
for
for
the
kubernetes
metadata,
it
doesn't
support
incremental
snapshots
on
cloud
providers
and
it
kind
of
does
on
rustic
it
doesn't.
F
I
A
A
B
So
and
if
we
have
of
lightness
check,
then
my
question
is
would
be.
Should
we
have
a
different
check
that
checks
for
readiness?
Do
we
I,
don't
think
I,
don't
think
this
is
relevant.
I
just
thought
I
would
bring
this
up.
I
take
one
check
would
be
sufficient
just
to
check
the
seis.
Febrero
is
ready
in
the
ready
state,
meaning
it's
ready
to
run
a
create
a
backup
and
then
so.
B
F
The
kubernetes
definition
of
liveness
is
that
the
pods
up
and
running
and
the
kubernetes
definition
of
readiness
is
that
the
pod
is
ready
for
network
traffic.
So
kubernetes
readiness
checks
are
you
we
for
adding
them
to
a
service
or
a
load
balancer.
So
in
the
kubernetes
context
in
if
we
had
like
these
checks
on
the
pod
definition,
a
readiness
check
wouldn't
be
relevant
to
Valero
because
we're
not
behind
a
load
balancer
we're
just
like
grabbing
Sierra,
DS
or
CRS
off
the
API
server.
F
F
So
basically
it
starts
another
local
webs
web
server
in
the
go
routine
and
you
would
hit
this
end
point
and
that
end
point.
When
you
go
to
there,
it
would
do
some
of
the
checks
we
do
on
startup,
so
it
would
say:
okay
do
I,
have
any
backup
storage
locations?
No!
Well,
then
that's
a
404
or
I
need
to
chant
work
change
to
some
other
error.
F
F
It's
taking
all
the
existing
checks
that
we
have
that
run
on
server
startup
and
it's
exposing
them
in
this
end
point,
and
if
you
just
ran
this
on
your
laptop
outside
of
kubernetes,
it
would
just
loop.
So
until
someone
hits
that
end
point
or
it
runs
in
a
kubernetes
server,
it
would
just
sit
there
there.
F
The
velaro
process
would
just
sit
there
and-
and
this
PR
would
rely
on
the
Valero
pod
definition
having
the
end
points
defined
and
in
the
kubernetes
api
server.
Actually,
specifically
the
cubelet
hitting
that
HTTP
endpoint
on
some
period
that
we
define
and
saying.
Are
you
ready
all
right?
Are
you
live
sorry?
All
right
are
you
live
and,
if
flare
ever
said
no,
then
the
cubicle
would
mark
the
pod
as
not
are
as
unhealthy.
F
If
you
want
a
demo,
I
can
do
a
demo
and
we
can
kind
of
see
it.
The
part
that's
missing
here
is
I.
Didn't
do
the
pod
definition
change
that
actually
puts
it
in
a
cluster
and
calls
into
it
and
and
says:
hey
poke
the
end
point
to
to
market
the
the
thing
I
like
about
this
liveness
check
versus
doing
it
ourselves
is
it
plays
into
the
existing
kubernetes
kind
of
the
tool
sets
like
cube.
This
will
make
it
so
cube.
E
F
B
F
Main
things
are
getting
the
deployment
definition
updated
so
that
it's
actually
used
in
I
think
it
might
make
sense
to
demo
it
just
so.
People
can
see
it
and
see.
What's
getting
returned,
I
know,
there's
there's
other
PRS
out
there
asking
for
like
hey.
Can
you
add
a
web
server
into
Valero
I
want
web
hooks.
I
want
a
proxy
for
the
API
storage
and
if,
if
that
stuff
ever
comes,
then
we
could
add
readiness,
checks
and
all
that
kind
of
stuff
for
load.
Balancer.
F
But
I
think
this
is
one
way
to
solve
some
of
the
issues
like
hey.
We
don't
check
for
backup
storage
locations
being
valid
after
server
startup.
We
can
have
the
checks
all
put
into
this
endpoint
and
poke
it
every
so
often,
that's
mostly
I
think
going
to
require
tuning
to
find
those
appropriate
time
periods,
because
the
the
one
thing
that's
kind
of
like
a
gotcha
here-
is
we're
going
across
the
internet
in
a
lot
of
cases
to
reach
those
backup
storage
locations.
So
we
can't
have
this
check
period.
F
B
F
B
F
Don't
know
I,
don't
think
it
hurt,
because
in
my
testing,
I
was
just
letting
it
sit
there
and
loop
and
then
I
applied
to
CR.
Ds
am
l
file
and
then
it
just
like
recovered
so
and
then
start
serving
so
and
like
I
applied
a
backup,
storage
location
then
deleted
it
and
then
reapplied
it
and
then
recovered
so
I
don't
think
it
hurts,
because
then
you
can
like
remove
the
Sierra,
T's
and
re.
Add
them
and
valerio
just
works
again,
so
it
takes
advantage
of
the
self-healing
stuff
in
kubernetes.
I.