►
From YouTube: Velero Community Meeting/Open Discussion - May 19, 2020
Description
May 19, 2020
Status Updates
[ashish]
Community support last week.
Quite a few restic related issues/ mostly questions.
Migrating our plugin repos to use go modules:
velero-plugin-for-microsoft-azure
velero-plugin-for-gcp
velero-plugin-for–aws
Moving our CI from travis to Github Actions:
Started with velero-plugin-for-microsoft-azure. Once this is merged other plugins and velero core should be straight forward.
Testing 1.4 release
Testing CSI in Azure
More testing this week.
Started on blog post for CSI velero integration. PR this week.
Update our helm charts to support Velero feature flags
Started working on a Octant plugin for velero over the weekend.
Not fully functional just- scaffolding.
more to com
PRs, ideas and suggestions welcome!
[carlisia]
k8s Slack #velero channel now has a @velero-maintainers user group
Migration of BSL CRD to kubebuilder:
All CRD manifests, and generation of manifests ported to kubebuilder (simpler to do all of them imo, but optional)
BSL api client and code generation ported to kubebuilder
Current status:
basic CRUD works on BSL and backup
need to fix failing tests
steve
on community support this week
getting ready for 1.4 release
1.5+ planning
stephanie
1.5 release planning
velero roadmap planning
nrb
1.4 testing
Blog post for v1.4.0
Some reviews
Discussion Topics
vitta
Would like to have a high level walk through of this PR https://github.com/vmware-tanzu/velero/pull/2543 and the procedure of next steps
cblecker
Non-existent items should not cause backup to be PartiallyFailed
areed / marccampbell
Discussion of RestoreHooks proposal @ https://github.com/vmware-tanzu/velero/pull/2465
Governance Model review
I may be gone to at another meeting by the time this comes up. folks can view the PR and comment.
PR at https://github.com/vmware-tanzu/velero/pull/2541
A
Hi
everyone
and
welcome
to
this
week's
episode
of
the
malaria
community
meeting
and
open
discussion
today.
We've
got
a
bunch
of
status
updates
that
we
want
to
go
through,
and
then
we
have
a
few
interesting
discussion
topics
as
well
from
community
members,
the
Vista
and
crystals.
So
let's
dive
into
it,
and
if
you
have
anything
you
want
to
add
to
the
meeting
notes.
Please
do
so.
I'm
gonna
share
the
link
to
the
meeting
notes
in
the
chat
there
and
then
I'll
share
my
screen.
So
you
all
can
see
what
I'm
looking
at.
B
So
I
was
in
communities
for
last
week,
not
quite
a
lot
of
issues.
It
was
pretty
quiet,
mostly
questions
about
rest,
ache
and
stuff.
So
I
had
some
time
to
do
other
stuff,
I,
moved
or
I
have
peers
open
to
move
all
our
plug-in
repositories
to
use
Co
modules,
so
I
have
those
peers
out
and
then
I
also
started
working
on
like
moving
or
moving
our
CIS
from
Travis
to
github
actions.
As
an
example,
I
did
the
Microsoft
Azure
plugin
and
once
that's
merged.
B
Everything
else
should
be
pretty
straightforward
and
then
the
other
things
were
yeah
I
saw
an
issue
being
opened
up
for
the
for
the
feature.
Flags
on
helm,
use,
feature
using
feature
flags
on
our
help,
repositories
or
hem
charts
I
have
a
PR
out
to
fix
that,
and
last
week
they
were.
There
were
also
like
a
couple
of
asks
for
like
a
dashboard
for
velaro.
So
over
the
weekend,
I
spend
like
a
couple
of
hours
putting
together
or
getting
started
on
like
octant
plugin
for
Valero.
B
C
Slack
ID
and
the
tool
that
does
that
automation
was
not
accounting
for
this
in,
but
they
they
have
a
periodic
change
in
their
stuff
and
you're,
going
to
be
added
to
that
group
as
well,
and
oh
so
I
got
to
the
end
of
the
tunnel
with
migraine.
The
yes
LCR
d2q
builder,
in
having
filler
work
into
ends,
and
now
pretty
much
I
just
have
to
change
the
tests.
So
I'm,
assuming
we
still
want
this
and
I
should
have
a
PR
for
it
tomorrow.
Latest.
C
D
C
C
C
D
Not
too
much
for
me,
I
was
out
the
second
half
of
last
week,
so
not
a
lot
of
update
since
last
meeting
I'm
on
community
support
this
week
and
basically
just
getting
ready
for
the
1.4
release,
quick
update
on
that.
We
are
currently
planning
to
push
that
out
next
Tuesday.
So
next
monday
is
Memorial
Day,
it's
a
holiday
for
folks
in
the
u.s..
D
E
Everyone
so
yeah,
like
Steve
mentioned,
we
are
working
on
doing
the
1.5
plus
release
planning.
We
are
also
looking
forward
to
the
future
towards
getting
at
least
a
three
month.
Roadmap
for
Valero
put
together
a
lot
of
eerily
interesting
items
on
the
roadmap,
including
running
concurrent
or
close
to
work.
Our
pods
improvements,
some
communication
work,
rustic
work.
Of
course
that's
come
up
a
few
times.
We
are
also
looking
toward
the
future
for
potential
multi-tenancy
support
working
further
on
those
CSI
snapshot
or
improvements.
E
We
also
have
up
on
there
a
lot
of
the
improvements
coming
out
of
the
you
know:
community
some
of
the
install
configure
X
that
currently
she
has
been
working
on
some
in
advance,
restore
scenarios,
performance
and
scale
improvements,
some
snapshot
or
improvements
as
well.
As
you
know,
even
some
plug-in
and
infrastructure
improvements
and
AZ
and
region
backup
restore
a
lot
of
really
interesting
kind
of
topics,
and,
to
that
end,
we're
also
looking
for
opportunities
to
make
sure
that
this
roadmap
is
visible
and
surfaced
community.
E
We're
working
on
that
right
now
the
team
does
have
a
longer
form
sort
of
roadmap
document
put
together
right
now.
We've
got
a
lot
of
those
topics
that
I
mentioned
listed
on
that
document
and
we're
working
on
the
team.
Prioritization
of
that
and
once
we've
gone
through
those
exercises,
we'll
settle
on
a
format
and
kind
of
let
everyone
know
where
they
can
find
in
those
details.
E
Yeah,
that's
a
great
question,
so
we
are
also
looking
for
providing
details
on
how
we
can
get
feedback
from
the
community
on
the
road
map
release,
because
we
definitely
want
to
hear
from
everyone
on
those
pieces.
We
are
looking
at
kind
of
publishing,
more
of
that
are
making
it
more
visible
and
github,
potentially
maybe
will
create
a
document
and
publish
it
there
so
that
the
community
has
the
idea
and
provide
the
link
here
in
that
community
meeting.
F
So
last
week,
I
don't
have
a
lot
to
go
into
what
today,
mostly
one
point
for
testing
the
link.
There
is
just
a
spreadsheet
of
test
cases
that
are
throwing
together.
I
was
gonna,
do
it
in
a
wiki
page,
but
markdown
and
tables
are
kind
of
hard
to
read
and
edit,
so
at
least
bring
the
editing
side
they're
hard
to
line
up
columns.
So
I
just
heard
together
a
spreadsheet,
so
I've
I've
made
some
scripts
that
were
running
through
different
scenarios
against
each
of
these
kubernetes
versions
and
running
five
backups
per.
F
The
main
use
case
I
want
to
hit
is
when
we
don't
have
any
feature
flags
on
to
make
sure
experience
is
good
for
folks
using
it
in
the
common
case,
but
we
do
want
to
make
sure
our
CSI
mode
and
this
API
the
enable
a
API
group
versions.
Flags
are
working,
but
they
will
be
behind
these
flags,
so
those
new
that
new
functionality
won't
be
affecting
those
people
they're
not
working
on
the
1.40
blog
posts
got
some
edits
from
the
team
and
then
doing
some
reviews,
but
man,
the
huge
update
for
me
this
week.
A
G
G
G
G
And
then
the
next
lines
are
about
like:
how
can
we
wait?
Can
we
do
it
like
basically,
as
part
of
create
snapshot,
call
when
it
is
done
from
the
server,
so
the
snap
shorter
creates
it,
and
once
the
call
is
done,
the
whole
new
server
can
actually
pick
it
and
persist
it
to
a
backup,
location
and
user
can
can
actually
get
it
by
using
label
search
or
by
going
to
the
SIA
the
backup
name
on
volume,
name
which
user
can
know
like?
What
is
the
progress
happen?
G
What
is
the
progress
on
the
thing
or
endorse
if
it
fails
the
status
part
that
kind
of
details,
and
similarly
goes
in
the
same
lines
that
restore
part?
Also,
we
have
the
CR
and
then
basically
here
we
create
the
world
from
snapshot,
so
the
snapshot
ID
is
what
will
provide
it,
and
this
will
be
same
as
the
one
even
in
the
world.
All
describe
backup,
output
in
the
backup
name
and
then
status
shows
how
it
is
progressing.
G
D
G
G
G
G
So
only
thing
I
would
say,
machine
is
like
the
plugin
has
to
be
things
like
it
can
be
right
now.
It
is
designed
such
a
way
that
it
doesn't
need
to
depend
on
it.
So
so
a
little
server
can
actually
think
I
mean
find
that
if
no
such
CR
exists
there
is
no
need
to
push
it
to
the
object
storage.
So
that
is
also
fine.
D
F
G
G
F
F
G
That's
exactly
the
case:
yes,
okay,
so
the
point
with
a
little
blinking
you're
doing
is
we
send
the
data
I
mean
plug-in,
does
send
it
to
the
object,
storage
stores
it
there,
so
we
I
mean
is
how
like
one
is
based
on
ZFS,
which
does
this
streaming?
The
ZFS
send
and
receive
streams
to
the
object,
storage
and
it
also
does
incremented
that's
the
key
term.
So
CSI
think
we
had
to
get
anything
near
theory
and
another
thing:
the
value
actually
uses
the
entire
application
perspective.
Doers
I
mean
the
CSI.
G
B
G
So
it's
like
so
other
ways
like
the
server
itself
can
be
a
regular
polling
so
that
it
works
with
non
communities,
need
to
plugins
also
but
it'll
be
too
much
of
changes
and
the
other
is
having
the
restrictions
on
the
name
so
which
is
like
the
whatever
creates
snapshot,
latent
value.
That
is
the
snapshot
ID
that
can
be
used
to
know
what
is
the
here?
It's
created,
that's
another
restriction.
I
think
this
sounds
better
currently.
G
D
Yeah
so
I
think
you
know
all
the
maintainer
certainly
will
need
to
take.
Take
a
slightly
more
detailed,
read
through
this
and
add
any
questions
and
comments
that
we
have
based
on
that,
and
you
know
anyone
else,
who's
interested
in
this
as
well.
Should
the
should
feel
free
to
to
put
a
review
on
the
PR,
so
you
know,
ideally,
we
can
through
that
just
sort
of
get
in
sync
on
intended
design
and
once
we
can
get
the
design
approved,
then
we'd
be
okay
to
move
forward
to
implementation
I.
D
Definitely
you
know
the
the
first
alternative
that
you
mentioned
around,
adding
another
method
to
the
volume
snapshot
or
interface.
That
was
that
was
one
thing
that
kind
of
occurred
to
me
as
it
as
another
way
to
do
this,
so
you
know
I
see
what
you've
written
here
and
I
definitely
want
to
think
about
it.
A
little
bit
and
just
make
sure
that
you
know
going
with
a
CRD
approach,
makes
sense
versus
verses.
That.
G
H
It
looks
like
there
was
some
discussion
and
then,
if
she
just
kind
of
stalled
out
without
an
exact
consensus
from
folks
I
wanted
to
bring
it
back
up
again
and
see
if
there
was
able
to
be
a
consensus
reached.
I
know,
there's
also
a
PR
open,
that's
related
to
this.
That
would
again
fix
my
specific
case,
but
as
I
thought,
I
the
hit
where
the
you
know.
If
a
pod
is
the
impending
status,
don't
try
and
backup
it
don't
associate
PVCs
for
backup
like
attach
to
the
pod.
D
Yeah
so
I
mean
I
can
definitely
jump
in
with
my
opinion.
Here,
I
mean
I,
I
I
personally,
don't
think
we
should
flag
these
things
as
as
warnings
or
cause
the
backup
to
go
into
failure,
and
we
we
did
talk
about
this
before.
But
you
know,
if
you
have
a
pod,
that
references,
a
I
guess
are
a
couple
of
different
cases,
but
the
easiest
one
for
me
to
think
about
is
a
pod
that
references,
a
PVC
that
that
doesn't
exist
in
the
cluster,
and
so
you
know
previously
some
folks
said
okay.
D
So
if
you
have
a
pod,
that's
supposed
to
be
using
a
PVC,
but
that
PVC
doesn't
exist.
We
should
we
should
warn
the
user
who's
trying
to
create
the
backup
somehow
and
I
guess
my
you
know.
My
personal
opinion
is
that
it's
not
full
arrows
job
to
flag
sort
of
invalid
configurations
and
we
should
just
back
up
what's
there
and
if
the
pod
is
referencing
a
PVC
that
doesn't
exist,
we
should
we
should
just
detect
that
back
up
the
pod.
D
I
Yeah
I'm,
just
saying
in
terms
of
you,
know
the
who
said
that
you
know
we
shouldn't
be
warning
or
pretty
a
producing
there.
You
know,
from
my
point
of
view
whether
we
log
it
as
a
backup
warning
or
not
is
less
important
to
me,
but
seeing
it
as
a
failed
backup.
You
know
as
an
issue
because
our
migration
tool,
you
know,
detects
that
failure
and
says
hey.
We
can't
move
forward
due
to
the
backup
failed,
so
yeah,
so
I
guess
I'm
in
total
agreement.
I
D
Well,
yeah
thanks
for
the
input
and
then
I,
you
know
I
think
there
are
like.
There
are
a
couple
of
different
ways
we
could.
If
we,
if
we
want
to
make
this
change,
that
we
could
go
about
it,
I
mean
we
could
you
know
we
could
focus
on
the
specific
case
of
a
pod.
That's
you
know
either
pending
or
referencing
a
non-existent
PVC
and
checking
for
those
conditions,
or
we
could
change
the
backup
code.
D
That
gets
basically
gets
additional
items
returned
from
plugins,
which
I
think
is
where
this
tends
to
come
up
and
basically
any
time.
An
additional
item
is
returned.
Falero
could
check
and
see
if
that
item
actually
exists
before
trying
to
back
it
up
and
and
if
it
doesn't
exist,
basically
not
not
logging
in
there
or
not,
creating
an
error
and
a
backup
failure.
I
D
B
Application
is
not
in
a
state
that
they
expected
it
to
be
because
they
expected
the
backup
to
be
complete
and
everything
should
be
fine,
but
that
wouldn't
that
would
be
like
our
violation
of
what
the
customer
or
the
user
moneth
might
expect.
So
that
is
that
is
something
that
I
think
we
should
kind
of
think
about
here.
It's
it's.
We.
F
F
Because
I
yeah
I've
struggled
with
this
issue
because
that's
a
valid
point
like
a
user,
could
be
like
hey
why
the
backups
s
completed,
but
my
application
still
isn't
working
because
I
don't
have
everything
but,
on
the
same
token,
I
agree.
It's
Stephen
standpoint
that,
like
it's,
not
Blair's
job
to,
say
you,
your
clusters
must
complete
years
and
you're
missing.
Stuff
Valero
is
just
like
up
ever
I'm
backing
everything.
That's
here
you
just
happen
to
be
pointing
at
some
things.
That's
non-existent.
I
I
That's
that's
what
my
thinking
there
and
also
the
question
is
on
restore.
Can
you
restore
this
thing?
You
know
because
if
we're
saying,
instead
of
calling
it
a
failed
backup
of
because
successful,
if
you
go
back
in
then
and
restore
this
back
up,
do
you
get
everything
that
you
started
with
or
the
things
that
you're
actually
missing
because
of
I
think
that
would
also
be
relevant
to
actually
see
you
know
in
practice,
with
with
the
particular
case
we're
talking
about
here,.
J
J
I
Of
the
point
to,
instead
of
ignoring
these,
these
issues
listed
as
a
warning
on
the
on
the
on
the
backup.
So
if
you
want
to,
if
you
want
to
look
into
this,
you'll
see
the
list
of
warnings,
but
it
doesn't
list,
they
still
have
completed
backup.
So
you
can
consider
it
a
success,
but
you
have
access
to
those
warnings.
If
you
want
to
look
it
and
actually
see
what
happened
there,
yeah
I.
C
Was
just
going
to
ask
that
if
we
changes
to
make
a
sort
of
the
backup
is
successful
in
a
case
like
that,
what
would
be
the
best
in
in
Scot?
You
just
said
the
warning
will
be
good.
Raphael
I
wanted
to
get
your
perspective.
What
would
be
the
most
useful
way
for
us
to
let
the
user
know
would?
Would
it
be
maybe
a
separate
phase
to
indicate
that
hey,
we
expected
to
find
this
where
you
wasn't
there?
Oh,
just
a
warning
or
something
else,
warnings.
J
Are
good,
the
thing
is:
maybe
we
have
to
discuss.
Different
flavors
of
successful
will
be
a
knob
or
something,
but
in
large
clusters,
when
you
have
thousands
of
objects,
and
then
you
do
all
nothing
successful.
It's
kind
of
hard
right
back
at
my
experience,
I
had
absolutely
no
control
what
the
users
dead
in
their
namespaces
and
was
really
weird
stuff,
and
then
you
just
saying
that
you
know
your
backup
failed
because
of,
like
you
know,
I
miss
computer
pod,
it's
kind
of
like
maybe
a
little
bit
misguided.
J
C
J
J
D
I
Is
also
the
case
where
again
for
restores
right
now,
those
warnings
are
also
shown
you
describe,
and
if
we
talked
about
this
in
previous
weeks,
how
the
the
describe
behavior
is
different
on
backup
in
our
store.
If
the
backup
described
could
include
the
warnings
as
well,
it
will
be
a
little
more
visible.
I
F
Also
have
warnings
like
if
you,
if
you
do
a
whole
cluster,
restore
velaro,
often
generates
and
warnings
for
its
own
CDs.
So
like
backup,
storage,
applications
and
backups
like
the
custom
resource
definitions
already
exist
in
the
cluster
and
valera
will
try
to
take
those
out
of
the
backup
and
put
them
in
so
like
those
those
warnings
which
is
really
clean
up.
Somehow,
because
you
don't
want
those
be
false
flags.
F
But
I
in
terms
of
phases,
like
I
I
mention
about
adding
a
ton
of
gradiation
cuz
I
in
in
in
talking
about
like,
should
it
be
completed
partially
completed
a
personally
failed
like
it
or
actually
abodes,
like
I'm
I'm
thinking
about
the
difference
between
partially
sunny
and
partially.
Very
importantly,
sunny
partly
cloudy,
like.
B
C
I
I
We
could
here's
the
warnings,
you
know
if
you
do
it
again
go
get
the
same
warnings,
probably
which
is
what
we're
talking
about
here
and
and
when
you're
talking
about
you
know
Valero
fitting
into
another
larger
infrastructure
where
the
tool,
that's
you
know,
running
a
backup
in
a
restore
and
determining
based
on
that
status
field.
You
know,
what
do
we
do
with
this?
Is
this
good?
Are
we
gonna
have
to
do
till
use
it
or
to
do
it
again?
Are
we
okay
to
do
restore
now,
and
it's
that
you
know
completion
field?
C
I
think
the
distinction
that
Scott's
helping
make
here
is
that,
if,
if
a
failure,
if
there's
a
start
as
a
failure,
the
failure
has
to
be
a
very
little
failure
in
our
viler
failure,
because
the
cluster
was
not
in
the
States
was
missing,
something
there
was
in
a
case
like
this
right.
There
was
something
that
was
supposed
to
be.
That
wasn't
is
not
a
valeera
failure.
Well,
there
was
capable
of
packing
everything
up.
A
D
A
A
K
D
Yeah
and
just
to
chime
in
there
I
mean
I.
I
know
that
that
is
theoretically
an
issue.
I
mean
I
will
say
that
our
our
kind
of
entire
rustic
restore
implementation
assumes
that
it
can
just
you
know,
essentially
exactly
it
doesn't
literally
exact
into
the
pot,
but
essentially
can-can
exact
into
the
restored
pod
and
and
wait
for
the
rustic
restore
to
be
completed.
But
I
know
that
I
think
maybe
Scott
has
seen
the
issue
where
sometimes
that
doesn't
work
correctly.
D
K
D
Mean
well,
we
obviously
need
to
resolve
on
the
approach
for
that
item.
It's
it's
been
a
couple
of
weeks
since
I
went
through
this
I
can't
remember
if
there
are
any
other
open
questions
that
we
need
to
resolve
or
or
things
that
are
still
open,
but
I
think
you
know
if
we
can
just
go
through
and
make
sure
that
each
of
the
comments
or
questions
are
sort
of
addressed,
and
we
can
continue
to
discuss
the
approach
for
that
specific
item,
which
seems
to
be
the
the
key
one
to
resolve
at
this
point.
F
A
Thank
You
Ashish
all
right
last
item
on
the
agenda.
Here
we
have
gotten
a
few
questions
regarding
Valero's
governance
model
and
how
maintain
errs
and
contributors
are
viewed
within
the
project
and
we
are
putting
up
a
PR
for
a
proper
governance
model
and
describing
the
the
maintainer
x'
and
the
roles
of
the
maintainer
have
as
well.
A
B
A
A
We've
had
some
internal
discussions
about
this
and
we,
of
course,
want
to
embrace
the
the
community
and
welcome
external
maintainer
x'
for
the
Valero
project
here.
So
if
you
would
like
to
become
a
an
external
maintainer,
or
rather
a
maintainer
for
the
Valero
project,
that
would
be.
That
would
be
fantastic.
So
we're
we're
putting
this
in
place
to
make
sure
that
we
yeah
we
played
by
the
the
same
rules.
Everyone
and
that's
that's
the
plan
going
forward.
A
If
you
look
at
the
files
here,
we've
got
a
governance
file
describing
how
the
project
will
be
governed
or
lessening
all
the
code
repositories
for
for
the
project.
Of
course,
the
different
community
roles,
users,
contributors
and
maintainer
x'.
We
might
have
a
fourth
role
in
there,
depending
on
some
conversations
that
we've
had
and,
of
course
how
voting
happens
if
that
needs
to
be
done,
but
most
of
all
really
important
here
how
decisions
are
being
made.
The
project
proposal
process
you
all.
A
Then
going
into
the
the
maintainer
file
here
you
will
see
the
the
maintainer
list,
the
github
ID,
the
affiliation
of
the
organization
affiliation
here
and
then
other
contributors
and
stakeholders
within
the
project
that
might
not
be
maintained.
Errs
like
me,
I'm,
not
a
maintainer
but
of
course
I'm
here
to
facilitate
discussions
and
hopefully
help
out
with
the
the
community
portion
of
things.
A
A
A
Alright
I
think
that's
it.
We
don't
have
any
contributor
shoutouts
this
week
either
we're
gonna
do
a
big
one
for
for
the
1.4
release,
as
we
said
so
hold
off
with
that
until
next
week,
where
we'll
have
a
long
list
of
awesome
people,
and
is
there
anything
else
anyone
wants
to
add
to
the
meeting
today.