►
From YouTube: Velero Community Meeting/Open Discussion - Jan 28, 2020
Description
# Jan 28, 2020
## Agenda
- ### Status Updates
- Nolan (have to leave early)
- Working CRD restore issues
- I think all we have left is the int64/float64 conversion on some CRDs raised by Scott Seago on https://github.com/vmware-tanzu/velero/pull/1937
- CSI update
- Currently AWS, GCP, and Azure aren't up-to-date on their CSI drivers, partially due to the newer v2.0.0 images requiring Kubernetes v1.17
- We're going to work on implementing Velero against the v2.0.0 codebase and v1beta1 CRDs and help update their drivers where we can, using the demo hostpath driver for now.
- Ashish (also have to leave early)
- Same update as Nolan on CSI
- Path forward:
- Test Velero against the hostpath plugin.
- Currently figuring out the deployment of a CSI driver:
- All the various sidecars that need to be used
- What should a deployment of the CSI driver look like:
- This information is a bit muddled/ fragmented across the various drivers.
- On community support this week:
- Getting caught-up on PRs and open issues
- [carlisia]
- All caught up on PR reviews except 1 for the helm-chart and 1 for the aws plugin
- Updated the resources page with latest videos/blog posts/podcast about Velero. New entries are welcome if anyone knows of resources not yet included.
- Contributed to discussions and revisions of a submission for VMware internal conference on a Velero related topic, submission is today
- Participated in the DP WG meeting, great start
- Back to updating CLI design doc after this meeting!
- Steve
- lots of PR reviews
- working with @nrb on sorting out CRD backup/restore issues
- ### Discussion Topics
- [steve] do we want to release a v1.3 soon-ish with the CRD fixes and the other [~70 commits](https://github.com/vmware-tanzu/velero/compare/v1.2.0...master) in master since v1.2, and push the CSI integration to v1.4 (can still keep to an end of March timeframe if feasible)?
- thinking here is that it's easier to release everything in master than cherrypicking a few things, and we have a decent # of commits built up already, which I don't really see a reason to hold back
- [Antony B] - What about https://github.com/vmware-tanzu/velero/issues/2069 ? Is that still 1.3 or 1.4?
- We'll review and see if we can fit it in for 1.3 or have to move it.
- [carlisia] Do we want to move https://github.com/vmware-tanzu/velero/pull/2214 (migrate to go 1.13 + modules) to "Ready for review"?
- This doesn't look like it'd be part of Velero: https://github.com/vmware-tanzu/velero/pull/2119 ([design] Proposal of triggering backups based on Kubernetes events)
- ### Contributor Shoutouts!
- @cpanato - CLI auto-completion docs (https://github.com/vmware-tanzu/velero/pull/2208)
- @Sh4d1 - bump restic version (https://github.com/vmware-tanzu/velero/pull/2210)
A
B
But
the
fix
for
upstream
doesn't
look
huge.
It's
just
going
to
have
to
take
some
time
to
get
through
the
upstream
process,
so
I'm
hoping
to
get
that
in
today,
and
then
hopefully
we
can
get
that
out
and
Steve
has
a
discussion
item
for
later
on
exactly
how
we
get
that
released.
We
had
thought
about
doing
it
for
one
one
point
two
point
one,
but
we're
thinking,
maybe
just
do
at
one
point
three,
so
that's
been.
B
The
main
thing
I've
been
chasing
down
is
testing
down
on
various
versions,
testing
that
with
different
different
CRD
sets,
but
Scott
had
a
had
a
sample
one
that
ran
into
this
particular
issue
and
he
kind
of
pushed
back
on
us.
So
thank
you,
Scott
for
making
sure
we
address
this
and
then
the
second
major
thing
was
working
with
sheesh
on
tracking
down
current
si
si
support
on
the
big
three
cloud
provider.
B
Si
si
drivers
the
main
thing
there
is
about
two
weeks
ago
now
getting
going
on
three,
the
CSI
snapshot,
calming
controllers
had
a
new
update,
the
V
two
point:
zero
point:
zero
images
went
live
and
currently
no
CSI
drivers
support.
That
part
of
the
reason
is
that
these
new
versions
of
the
CSI
snapshot
right
si
si
snapshot,
api's
and
CRTs
need
kubernetes
1.17,
which
is
still
pretty
new
for
a
lot
of
providers.
B
I
think
ashish
has
net
update
against
that
and
yeah
we're
gonna
we're
gonna
try
to
help
these
cloud
providers
where
we
can.
Hopefully
we
can
go
through
and
and
update
their
the
code
imports
and
their
mo
manifests,
and
hopefully
that
takes
care
of
it,
but
part
of
that
will
be
on
them
to
get
their
offering
updated.
Do
you
have
good
where's
117?
B
So
that's
most
of
what
I've
been
doing
over
the
last
week
and
there's
been
some
data
protection
working
group
discussions
mostly
early
days,
they're
kind
of
figuring
out
how
we're
gonna
work
together
and
figuring
out
what
topics
we're
gonna,
we're
gonna,
tackle
and
and
starting
to
bring
together
kind
of
philosophy
and
what
things
to
protect,
whether
it's
the
kubernetes
metadata
or
and/or
volume
data,
and
things
like
that.
So
I
think.
D
D
I'm,
currently
kind
of
figuring
out
the
various
like
side,
cars
and
what
they
all
do
for
the
CSI
driver
and
also
how
do
we
want
to
bundle
up
that
as
a
deployment
looking
at
the
various
drivers
looks
like
there
is
like
a
lot
of
fragmentation
or
like,
for
example,
the
host
path
again
eyes
deployed
as
a
stateful
set
with
each
of
the
side.
Cars
are
not
a
side
car
containers
but,
as
individual
stateful
said,
with
all
running
on
as
privileged
pods
on
the
same
nodes
with
pod
affinity.
D
D
E
D
Been
able
to
like
follow
and
get
it
to
run,
but
I'm
not
sure
it
works
the
way
I
wanted
to.
So
that's
what
kind
of
got
me
to
think
about
how
if
the
deployment
is
set
up
correctly
and
if
I've
deployed
all
the
components
necessary
and
stuff
so
I
was
like
battling
through
the
yeah
mph
or
that
and
kind
of
understanding
what
exactly
it's
doing
behind
the
helpful
scripts
that
are
in
the
repo
and.
D
A
D
B
Ideally,
ideally,
we
won't
have
to
understand
it
beyond.
Maybe
support
community
support
quests
but
yeah
for
now
it's
it's
worth
knowing
but
yeah.
If
you
have
questions
and
and
if
things
could
be
explained
better,
it's
worth
talking
to
Shane
and
we
can
get
pr's
up
there
to
better
explain
documentation
if
things,
if
it's
just
that
things
aren't
clear,
we
can
explain
it
or
if
there
needs
to
be
architecture,
redone
or
things
like
that,
she
would
be
the
one
to
talk
to
okay.
D
Yeah
I
think
documentation
wise
we're
in
pretty
good
state.
It's
it's
just
that
the
documentation
is
not
there
in
the
place
that
you
would
expect
it
to
be
in
it's
not
in
the
repo,
but
it's
in
the
Associated
get
book.
So
it's
just
it's.
It
has
just
been
the
matter
of
figuring
out
where
that
documentation
is
you
you.
D
F
Hey
awkward
up
on
PR
reveals,
there
were
few
accepts
to
which
I'll
get
to
right
after
this
meeting,
I
felt
like
updating
the
resource
restore
space.
So
there
is
a
page
on
on
our
website
that
lists
blog
posts
that
were
written
by
community
members
in
any
podcast
or
video
related
to
Valero.
So
I
added
a
bunch
that
we're
missing
from
the
page
and
if
anybody
knows
of
any
resources,
that's
not
listed
there.
It's
it's
not
PR
at
this
points
if
I
won't
be
updated
today.
F
F
Last
week,
I
also
participate
in
the
data
protection
meeting,
which
I
think
is
great.
It
has
been
great
because
they
are
really
starting
from
the
basics,
defining
things.
What
things
are.
Nobody
is
jumping
ahead
and
assuming
that
everybody
is
one
on
the
same
page
as
far
as
what
things
mean
so
I
think
it's
they're,
starting
from
a
great
place.
There's
a
meeting
tomorrow,
9:00
a.m.
Pacific
time
to
is
in
there
for
in
thereafter
every
two
weeks.
If
anybody
is
interested
and
I
have
to
get
back
to
the
CLIA
design.
E
Right
along
Steve
yeah,
no
major
updates
for
folks
here,
just
lots
of
PR
reviews
and
spending
some
time
with
no
one.
Looking
at
the
co
D
the
the
multitude
of
CRD
backup
and
restore
issues
and
sorting
through
those
that's
those
have
been
the
main
things.
So
we
can
probably
just
jump
into
the
discussion
topic
and
I
bumped
mine
to
the
top,
just
because
I
know
Nolan
needs
to
leave
midway
through.
E
So
we
had
no
one
mentioned
before,
but
we
we
originally
were
talking
about
pushing
out
a
12.1
bug,
fix
release
with
some
of
the
CRV
issues,
but
I
was
I
was
thinking
we
might
just
want
to
push
out
a
1.3
once
those
bugs
are
fixed.
I
mean
we've
got
something
like
70
commits
in
master.
Since
we
released
1.2
you
back
in
November
and
I,
don't
really
see
any
reason
to
hold
them
up
and
we
haven't.
We
haven't
really
made
any
progress
on
CSI.
E
Yet,
given
the
just
kind
of
the
timelines
of
things,
so
I
was
sort
of
thinking.
We
might
just
want
to
push
out
a
1.3
once
this
bug
fixes
already
and
then
we
can
keep
working
on
the
CSI
stuff
and
we'll
just
you
know,
label
it
1.4
for
the
next
release
and
doesn't
necessarily
have
to
impact
the
the
timelines
in
any
way.
You
know
just
depends
on
how
much
progress
we
make
on
the
CSI
stuff,
but
today,
I
just
wanted
to
get
folks
thoughts
on
that
and
whether
that
makes
sense.
F
D
E
Yeah
I
mean
it
kind
of
just
be
a
shuffling
like
you
know,
we
can
still
release
1.4
at
the
end
of
March,
which
was
our
original
timeline
for
1.3.
So
it's
just
it's
just
sneaking
an
additional
minor
release
in
before
then
and
relabeling
that
release,
but,
given
you
know
given
where
we
are
with
CSI
I,
feel
like
it's
possible
that
that
timeline
will
push
out
a
little
bit
as
well.
G
B
I
I,
don't
think,
there's
a
huge
rush
on
getting
the
CSI
stuff
out
now
so
yeah
I'm,
fine,
cutting,
1.3
and
and
trying
to
hit
the
trying
to
still
trying
to
hit
the
March
timeframe
but
like
at
this
point,
the
CSI
stuff
is
very
much
bleeding
edge
and
beta.
So
we
can
get
it
there
and
and
iterate
on
other
stuff
as
well.
But
it's
it's
not
like
we're
trying
to
with
CSI
it's
not
like
we're
trying
to
deliver.
B
That's
not
trying
to
deliver
a
new
GA
feature
right,
even
even
if
we
deliver
the
CSI
stuff
in
March,
it's
that's
still
going
to
be
a
beta
set
feature
set,
so
yeah
I
I
really
want
to
get
this.
The
CSI
stuff
I
know
I've
been
saying
this
week
for
several
weeks
on
it
now
and
and
I
want
to
get
that
out
of
the
way.
B
Hopefully,
this
this
n64
float64
conversion
stuff,
is
the
last
of
it
and
we
can
get
people
doing
their
CDs
correctly
and
there's.
There
is
one
more
fix
that
we're
gonna
have
to
put
in
for,
like
Steve
and
I
had
talked
about
having
a
more
long-term
fix
for
the
CDs
that
we'll
have
to
do,
but
yeah
I'm
I'm
for
the
topic
at
hand
I'm
for
getting
a
1.3
out
instead
of
a
the
one
point
two
point:
one:
okay,.
E
So
I
I
may
be
able
to
spend
some
time
next
week.
Looking
at
some
of
the
refactoring,
we
talked
about
for
the
discovery
stuff,
but
okay,
so
that
I
think
that'll
be
our
plan.
Then
let's
sort
of
figure
out
what
the
timeline
is.
Gonna
look
like,
but
you
know
first
first
focus
is
just
getting
the
fixes
tested
it
in
and
then
we'll
figure
out
what
the
release
timeline
looks
like.
D
The
other
thing
that
we
might
use
we
can
use
community
help
from
is
understand
what
releases
our
users
are
using,
that
it
kind
of
help
us
for
all
like
time,
our
releases,
if
there
is
not
much
demand
or
if
not,
if
there
aren't
many
users
using
1:17
at
this
moment,
even
if
we
were
ready
with
CSI
there
wouldn't
be
much
value
that
we
would
be
adding
to
our
users.
So
knowing
where
our
users
are,
would
would
be
really
helpful.
H
And
also
I
had
a
question
regarding
this
so
12th
end
of
last
year
the
find
a
bug
possible
memory
leak
issue
and
at
that
time,
since
it
was
really
close
to
the
one
or
two
relays
that
was
marked
as
a
milestone
for
version
1.1
row
3.
So
with
this
change
now,
how
does
that
affect
some
of
those
like
that
back
I
think
it
was
20
69,
that's
the
the
issue.
Id.
H
E
Yeah
I'm
not
totally
sure
we'll
have
to
look
at
you
know
what
else
besides
suicides
in
there
I
mean
I
know
you
know
we
had
that
one
labeled
as
a
p1,
but
I
spent
some
time
looking
at
it,
but
couldn't
didn't
totally
get
to
the
bottom
of
it.
So
I'd
say
you
know.
One
of
us
can
probably
take
a
look
at
that
over
the
next
few
days
and
see
if
we
can
sort
of
sort
out
what's
going
on
there.
Ok,
if
we
can
get
a
fix
in
41.3,
will
will
do
that.
Ok,.
H
I
E
F
I
just
want
to
bring
up
these
four
requests.
Die.
Steve
opens
its
I'm,
not
sure.
If
we
need
to
guess
consensus,
if
we're
ready
to
migrate
to
1.3
I
tested
this
yesterday
and
seems
fine,
I
ran
the
I
pulled
this
behind
some
random
Idol
commissioning
he
works
perfectly.
Are
we
ready
to
go
with
this
or
no
I.
E
E
Let
me
take
another
pass
through
it
because
it's
been
an
I
think
two
weeks
since
I
last
touched
it
and
just
make
sure
that
there
aren't
any
other
duze
that
I'm
aware
of,
but
if
other
folks
have
some
time
and
could
contester
you
know,
do
a
quick
review
of
it.
That'd
be
helpful.
I
think
it's
ready
ready
for
review.
F
B
D
D
D
B
Yeah
I
haven't
looked
at
this,
that's
one
of
the
things
than
behind
on
his
PR
reviews
and
I
know.
I
know
on
on
had
like
what,
when
we
and
moved
some
some
of
the
CRD
definitions
around
he
had,
he
had
done
a
bunch
of
stuff
in
the
make
file
in
hack,
hack
directory
for
generating
the
CR
DS
and
like
there
was
some
stuff
with
the
client
code
gen,
and
all
of
that
he
was
turning
on
and
off
go
module
support
all
right.
So
does
this?
That's
probably
just
reading
PR
yeah.
E
I
wouldn't
say:
go
to
the
PR.
I
mean
I
broke
it
up
into
commits
they
kind
of
go
through
it
step
by
step.
So
there's
you
know,
there's
like
the
dev
there's
the
dev
build
and
then
there's
the
generation
stuff
for
the
most
part
everything's
running
with
modules,
but
there
are
some
quirks
to
a
couple
of
the
code
generators
around
where
they
need
to
be
located
and
where
they
expect
the
the
code
that
they're
operating
against
to
be
located.
B
E
F
You
and
I
have
the
next
one
too,
so
this
is
a
full
request
that
has
been
sitting.
That's
been
there
for
a
while,
and
we've
been
going
back
and
forth.
I
just
wanted
to
bring
it
up
in
case.
Somebody
has
seen
any
strong
opinion.
One
way
or
another
I
mean
I.
Think
at
a
higher
level
is,
does
this
going
to
Valera?
B
F
So
I
think
at
the
very
least
we
should
prune
this
and
see.
If
anything,
we
want
to
add
anything
from
this.
We
want
to
add
to
Valero.
I
personally
would
love
to
know
if
other
people
are
requesting
any
of
the
things
that
are
being
suggested
here.
I
would
not
add
any
of
it
if
if
there
are
not
more
requests
yeah,
so
if
people
know
oh
yeah,
I've
seen,
people
needing
this
or
I
have
I
needed.
I
I
Actually,
making
an
operator
that
uses
Valero
as
like
I'm,
basically
making
a
Valero
provisioner
as
like
a
dynamic
provision,
airfare
storage
class.
But
it's
a
lot
harder
for
me
to
make
like
from
scratch
a
storage
class
type
provisioner
for
Valero,
so
I
just
went
down
the
easy
path
of
whenever
there's
a
stateful
set
replica
increase.
Basically
I
have
a
custom
resource
definition.
I
That
is
a
wrapper
for
that,
and
then
it
performs
a
Valera
restore
it
labels,
those
persistent
volumes
and
as
a
claim
ref
for
a
PVC
that
I
know
it's
kind
of
deterministically
like
accounted
for
because
it's
a
stateful
set.
The
name
of
the
the
claim
ref
is
known
and
that
way
I
can
spin
up
a
bunch
of
these
blockchain
full
notes,
with
with
almost
real-time
data
and
I'm
using
Valero
as
like
the
back
end
provisioner
for
that.
I
So
so
I
can
definitely
see
where
somebody
would
want
to
make
an
operator
using
Valero
but
I
also
kind
of
agree
that
it
doesn't
belong
in
your
source
tree.
This
could
be
like
a
plugin
kind
of
thing,
although
not
not
really
the
plugins.
The
way
you
guys
use
them,
but
more
just
like
this-
is
how
you
hook
into
the
Valero
API
to
do
whatever
you
need
in
the
backend.
But
there
are
definitely
extensions.
People
are
building
on
top
of
Valero
no.
F
I
think
I
think
this,
for
example,
is
a
perfect
use
case
for
an
operator
because
for
us
to
do
understand
and
compiled
together
all
the
use
cases
of
where,
when
people
will
want
to
take
a
backup,
for
example,
if
X
happens,
I
want
a
backup,
but
the
other
person
if
Y
happens,
I
want
to
take
a
backup
for
us
to
incorporate
all
of
that
into
Valero.
I.
Think
is
it's
not
a
good
idea?
F
B
Yeah,
and-
and
this
is
something
I
had
told
now-
that
like
this
is
I,
can
I
can
totally
see
that
this
is
something
you
can
build
outside
of
Valero
without
existing
in
the
Valero
core
tree,
so
I
think
I
think
it
would
be
appropriate
to
set
a
deadline
like
if,
if
nyle
doesn't
come
back
to
it,
then
we
can
close
it
out
say
a
week
or
so.
If,
if
that
long,.
B
F
I
think
I'm
gonna
make
an
update
to
this
PR
and
say
from
last
week.
We
can
gather
some
strong
opinions
for
having
any
of
these
requests
in
the
lair.
Oh
and
we'll
wait
for
some
time.
We
are
not
going
to
consider
doing
any
of
this.
We're
not
going
to
consider
refining
it
either
because
it
just
doesn't
make
sense
to
go
in
the
Calero.
Anybody
disagree
with
this.
E
E
Just
a
couple
of
sins:
last
week,
seat
Bonanno
Carlos
has
been
doing
a
bunch
of
work
in
helm
and
then
the
core
Vallejo
repo
this
past
week.
He
got
some
some
really
helpful
docks
for
how
to
get
auto
completion
for
the
Valero
CLI
setup.
So
this
is
something
that
we've
had
support
for
for
quite
a
while,
but
the
docks
have
not
explained
what
the
prereqs
are
and
exactly
how
to
set
it
all
up.
So
he
contributed
some
really
extensive
and
well
organized
docks
for
getting
this
set
up
across
a
bunch
of
different
platforms.