►
From YouTube: Velero Community Meeting/Open Discussion - July 7, 2020
Description
July 7, 2020
Status Updates
nrb
Community support last week, still playing catch up on GH issues though
Review blitz yesterday, but did not get them all
End in sight for our internal build process work, hope to give more time to this open source work!
Restore hooks meeting yesterday, thanks everyone for attending!
carlisia
BSL controller PR: https://github.com/vmware-tanzu/velero/pull/2674
Worked on adding the BSL validation as webhooks; talked to @nrb and decided we won’t do it
Did and doing PR reviews
Finished reviewing the volume backup/restore progress design doc
Next: converting ServerStatusRequest to controller-runtime
ashish
Community support this week:
PR to fix a CRD backup regression in 1.4. PR Issue
Tested against "v1.16.9" "v1.17.0" "v1.18.4"
Will be included in the release for 1.4.1
Working on internal build setup. Should mostly be done, need a few finishing touches
PR to fix the file permissions on the manifests in the backup tarball.
0755 to 0644
PR 2685
Discussion topics
Phuong Demo of Resource ordering
nrb: master branch will be changing to main. We need to work on making sure docs redirects will work, but it looks like in-flight PRs will not be redirected unless GitHub introduces some sort of PR redirection feature. https://github.com/vmware-tanzu/velero/pull/2677 is tracking that work.
vitta: PR 2543
Contributor Shoutouts
@danielthrasher: updated acceptable values on cron schedule for day of the week from 0-7 to 0-6
@mauilion & team: Builder Image handling
@tbatard & team: Add linters
Everyone who attended the restore hooks meeting yesterday!
A
Hi
everyone
and
welcome
to
another
episode
of
the
valera
community
meeting
and
open
discussion
today
as
July
7th
2020
and
per
usual,
we're
gonna
start
with
a
bunch
of
status
updates
and
then
go
into
the
discussion
topics.
Let
me
see
here,
I'm
gonna
share
this.
You
should
see
my
Google
Chrome.
Is
that
correct,
yep,
alright,
awesome
so,
first
off
we
have
Nolan.
B
B
Also.
I
just
want
to
let
everybody
know.
I
know
in
the
last
few
weeks
have
been
each
time
and
this
meeting
I've
been
saying:
I'm
working
on
internal
build
process
work,
but
that
the
end
is
in
sight
on
that.
Hopefully,
that
will
be
done
this
week
or
next
one
way
or
another.
It
will
have
to
be
done,
so
that
is
wrapping
up
and
then.
B
A
B
Think
here
there's
a
couple
issues
that
we
are
probably
going
to
do.
A
one
point,
four
point:
one
four
I
know
she
she's
gonna
update.
Do
an
update
on
the
on
his
that
we
want
to
get
that
one
out
and
I
think
there
may
be
a
a
small
rustic
resource
limit
increase
that
I
would
like
to
put
in
the
141
fix
as
well,
but
that
there's
no
PR
for
that
yeah!
That's
just
something
in
my
head
that
I
would
want
to
fix.
C
C
C
So
so
after
I
got
that
PR
done,
I
started
working
on
added
the
same
validation
but
as
web
hooks
for
creating
updates
and
then,
as
I
was
doing
it.
I
realized.
It
was
going
to
be
probably
impossible,
because
the
backup
storage
location
depends
has
an
external
dependency,
which
is
the
actual
store
which
we
can
only
bring
up
an
instance
of
that
if
the
server
would
the
plug-in
that
relates
to
that
store
is
running
there
anyway.
The
hats
conversation
with
Nolan
and
he
agreed
it's
not
worth
attempting
to
do
that.
C
So
in
the
the
way
the
controller
is
working,
is
it's
pretty
sufficient
for
what
we
want
to
do,
which
is
just
keeps
our
daily
working
on
PR
reviews.
I
finished
that
the
cap
restore
progress,
design,
doc,
I.
Think
I
agreed
with
the
previous
reviews
and
just
ask
them:
ask
him
to
move
the
the
approach
to
to
add
alternate
to
the
alternatives
section
which
is
big
deal,
but
unless
Ashish
also
wants
to
do
a
review,
I
think
it's
it's
ready
to
to
merge
pretty
much
either
after
we
wait
for
him
to
move
to
the
stuffed.
C
C
B
B
C
D
I
found
an
issue
that
we
had.
That
was
a
regression
in
1.4,
where
we
weren't
backing
up
the
CR
DS
in
the
right
place
or
in
the
right
in
the
expected
place
for
restores
to
work.
So
I
figured
out
what
was
causing
the
problem
and
I
have
a
PR
out
to
fix
that.
I
verified
that
this
works
against
116,
9,
117
and
118
4,
and,
as
Nolan
mentioned,
we
will
be
cutting
our
1.4
dot.
D
1
release
to
include
this
fix
and
and
all
the
other
changes
that
have
gone
in
from
1.4,
not
from
the
time
we
cut
1.4
other
than
that
I
was
working
on
some
of
the
internal
build
set
up
stuff,
which
should
be
mostly
done,
need
a
few
finishing
touches
on
that
to
call
it
complete
and
move
on
other
than
that.
I
also
Halla
fix
to
change
the
permissions
on
the
manifests
that
we
included
in
the
backup
turbine
we
currently
or
we
had
those
permissions
or
those
manifests
as
executables,
which
they
so
I
added.
A
D
A
E
E
A
E
E
The
first
backup
will
not
specific
and
then
and
the
second
one
on
the
second
backup
I
will
run
a
backup
with
the
order
optional
and
with
with
the
order
specify.
So
we
can
see
the
difference,
I
already
type,
all
the
back
so
I
don't
have
to.
But
you
can
see
that
this
one.
It
is
us
straightforward,
backup,
Bellaroma,
writes,
backup,
create
and
the
name
of
backup,
eps
backup,
no
order
the
names
paid,
the
names
into
how
matters
are
the
answers
normal
and
so.
A
E
C
E
So
this
one
is
the
new
thing
here
order
and
you
can
see
this
is
the
it's
just
a
string
in
this
string.
I
specify
first
of
all,
I
want
to
order.
The
I
want
to
specify
the
order
of
the
pot
that
I
need
to
backup,
and
in
this
part,
I
ordered
I
want
to
back
up
the
secondary
first
remember
in
the
first
one.
E
You
can
see
that
if
I
don't
do
that
and
then
the
finally,
then
the
secondary
right
in
this
case
I
want
to
back
up
the
secondary
first
then,
and
the
primary
and
I
don't
care
about
amateur,
so
the
aperture
will
be
a
bit
will
be
like
later
then
also
I
also
specify
the
order
of
the
persistent
balloon
thing.
So
in
this
case,
I
want
the
PVC
to
will
be
back
up
first
into
the
BBC
one
in
as
right.
E
Now,
if
I
don't
do
that
I'd,
if
I
don't
specified,
you
can
see
that
the
PVC
to
be
back
up
later
at
the
PVC
one.
Let
me
see
if
we
can
find
it
or
out
of
this
one
thing
is
yeah
right
here.
You
can
see
that
right
now,
if
I
don't
specify
order
that
PVC
one
will
be
back
up.
First
then
PVC
to
put
me
back
up
there.
Alright,
so
now,
I'm
gonna,
I
gonna
run
this
command.
D
E
E
Listen
persistent
go
to
claim
belong
to
this
pot,
then
go
to
belong
to
that
pot.
Being
back
up,
I
won't
dumb
one
here
and
now
you
can
see
their
secondary
I
mean
the
primary
right,
the
secondary
go
back
up
and
then
the
primary,
and
then
you
can
see
that
it
is
PVC
and
it's
TV
on
so
back
up
here
and
then
because
I
don't
specify
a
bitter
in
my
in
the
order.
So
it
will
go
after
this
first
after
the
one
that
inside
the
inside
of
us,
the
order
has
specified.
E
But
now
you
also
see
that
the
PVC
tube,
then
I
will
be
back
up
first
before
PVC
one
mean
back,
so
that
is
the
a
or
the
chain
that
caused
this
to
back
up
in
a
specific
order
that
I
specify
in
this
command
right
here.
So
so
you
can.
I,
this
is
the
they
often
leave
the
only
new
chain
and
health,
and
you
can
see
that
it
can
work
for
both
thoughts
and
resistant
guru.
E
E
B
E
Damon
said-
or
it
doesn't
matter
actually
within
the
name
same
namespace,
but
yes
in
my
case
it
is
inside
the
it
can
be
belong
to
one
staff
who
said:
oh,
it
can
belong
to
multiple
staff
who
set,
but
the
way
I
use
it.
In
my
a
consistent
backup
is
that
we
first
free
process
over
the
part
and
and
and
sequence
and
put
them
into
a
safe,
specific
sequence,
and
then
we
take
snapshot
of
them
one
by
one.
D
E
D
E
What
I
do
is
that
in
during
the
backup
process,
the
current
code
will
list
ow
it
with
a
list
of
all
the
MTT
belong
to
specific
resource
hive.
It
create
a
list
of
it,
so
my
code
chain
is
I,
only
reorder
them
in
the
order
that
I
want
that's
it.
They
are
all
which
is
yeah.
So,
for
example,
you
can
see
that
in
my
demo,
I
have
three
parts
right,
but
I
only
concern
about
what
two
of
them,
the
third
one,
the
arbiter
I,
don't
care.
E
So,
if
you
can
see
in
the
demo,
the
secondary
part
will
be
back
up
first,
then
the
primary
will
be
back
up
and
everything
else
that
I
don't
care
for
would
be
just
go
with
the
whatever
the
order
item.
So
I
don't
chain
that
order
I.
Only
kid
the
order
of
the
pot
that
I
concern
and
I
put
them
at
the
beginning
of
the
list.
The
rest
of
the
list,
which
is
remain
the
same
that
it
was
it,
was
in
the
current
code.
E
D
E
E
Put
them
all
in
a
map
and
I
go
through
the
order.
If
I
found
the
item,
I
put
the
item
in
the
still
result.
The
sorted
item
result
right.
If
I
don't
find
I
think
I
should
help.
If
I
don't
find
the
item,
I
should
have
give
a
warning
or
something
but
right
now,
I,
don't
and
then
after
I
run
through
this
loop.
E
So
when
I
run
the
second
loop,
the
loop
but
from
I
looked
through
all
the
original
item
right,
if
it
already
there,
if
it,
if
it
is
not
in
the
map
anymore,
that
mean
that
I
read
pick
it
up
earlier
then
I
ignore
it.
Then,
if
it
not
that
I
just
put
it
into
the
final
item.
So
this
way,
whatever
the
order
in
do
reason,
I
we
result
I
will
be
resolved
in
the
result.
E
When
they're
in
they're
sorted
item
right
and
this
one
will
guarantee
that
it
will
be
Oh,
everything
that
inside
is
out
of
this
could
be
pick
up
first
and
put
it
in
the
specific
order
in
the
resorted
item,
then
everything
else
will
be
put
it
in
with
it
in
later,
in
the
same
order
that
it
come
in
the
reason
idea,
so
that
this
is
one
function
call-
and
this
is
this
being
caught
in
one
place
at
the
very
end
of
the
function.
We're
here
so
in
this
function.
E
Right
here
is
this
is
a
function
that
to
a
specific
group
they
call
get
resource
item
and,
at
the
end
of
it,
I
just
need
after
update
collectible,
it
will
put
inside
a
variable
as
items
and
is
it
is
it
an
array
and
I
said
if
there
is
a
the
order.
Alright,
for
this
tag,
then
I
just
sorted:
that's
it
yeah,
that's
the
code
chain,
so
it's
very
low,
no
channel
regression
because
of
that
means
is
cool.
B
B
E
In
my
in
mines
used
case,
we
back
up
one
namespace
at
a
time.
Okay,
so
everything
that
we
are
talking
about
is
only
within
one
namespace,
but
I,
also
because
you
can
see
that
when
I
specify
the
order
right
it
each
of
the
item,
it
comes
with
the
name
space
last
item
named
man,
so
I
think
it
should
work
for
across
multiple
names.
B
I'm
curious,
like
I,
think
I
think
in
most
cases
cluster
scope
resources
you
can
back
up
by
using
like
a
a
namespace
tree
source
as
a
pointer.
So
as
an
example,
you
can
use
a
persistent
volume
claim,
in
a
names
case,
to
point
to
a
persistent
volume
yeah,
but
there
may
be
cluster
cluster
scoped
custom
resources
that
we're
not
thinking
about
so
I
just
want
to
raise
that
as
a
as
a
potential
thing.
E
Actually,
it
might
work
because
we're
in
my
implementation,
what
I
do
is
I
only
certify
the
name
and
mine,
the
name
that
I
have.
There
is
a
combination
of
name
space.
Last
the
item
space
in
the
persistent
volume
case.
For
them
it
doesn't
have
a
name
spare
associated
with
it,
so
it
just
it's
just
a
name
of
the
persistent
balloon
itself,
and
so
how
many
times.
B
E
E
E
I
think
it
suit
looked
because
the
the
oh
I
need
to
do
is
just
do
up
here.
I
mean
you
know,
compare
strings,
yeah
and
I
own.
So
right
now
my
implementation
right
now
is
the
exact
matching
right.
I
have
to
specify
the
exact
name
of
the
resort,
but
I
was
thinking,
maybe
in
the
future.
I
would
able
to
implement
like
a
regular
expression
there.
So,
for
example,
instead
of
specify
the
primary
pot,
zero,
family,
1
and
so
on,
support
I
just
say:
let's
collect
all
of
the
primary
part.
E
A
B
B
B
We
would
like
to
get
some
redirection
in
on
that,
so
I
need
to
work
with
Jonas
and
Abby
on.
We
can
set
up
rules
for
redirecting
from
from
master
domain
on
there.
The
other
issue
that's
going
to
come
up
is
that
any
open,
PRS
that
are
against
master
will
likely
not
work.
I,
don't
know
the
exact
failure
mode,
but
I
know
that
github
has
not
yet
implemented
a
like
PR
redirection
or
a
branch
redirection
feature,
so
we'll
have
to
get
them
reopened
against
main,
instead
of
master.
B
I'm
also
gonna
have
to
go
and
like
edit,
some
other
files
that
have
been
merged
since
so
wanted
to.
Let
folks
know
that's.
Coming
generally,
we
would
like
to
have
discussion
on
some
changes
or
really
almost
where
changes,
but
this
one's
not
really
negotiable.
So,
just
giving
you
heads
up,
I
I
do
realize.
B
B
This
wanted
to
highlight
again.
Please
take
a
look
at
this
list.
This
issued
two
six
five
four,
please
take
a
look
at
this
list
and
let
us
know
if
there's
something
not
in
this
list
that
you
depend
on.
Please
comment
on
this
issue
and
let
us
know
so
that
we
don't
move
it
into
an
internal
and
break
you.
D
B
Our
brew
install
is
actually
installing
a
source
tarball
from
a
lease
from
from
a
semantic
version
to
release
not
from
master.
It's
not
pointed
at
a
master
branch.
It's
pointed
at
a
tag:
release
okay,
so
I
think
some
especially
this.
This
definitely
would
have
been
a
problem
with
go
tooling
before
go
modules,
because
a
lot
of
go
tooling
was
pointed
directly
at
master
for
a
long
time,
but
with
how
we're
handling
the
brew
release
right
now
it.
It
is
less
of
an
issue
because
it's
pointed
directly
at
a
github.
B
B
Chocolaty
install
procedure
does
very
similar,
well,
I.
Think
that
one
grabs
the
binary
from
a
tag
release
but
I
think
that's
a
good
point
that
we
we
should
look
at.
That
I
was
not
thinking
of
syncing.
This
chewer
release
I
was
thinking
of
doing
this
kind
of
out-of-band,
but
do
let
me
know
if
that's
just
to
be
transparent
with
my
thought
process.
I
was
thinking
of
doing
this
separate
from
the
1.5
release
that
we
were
going
to
do
this.
B
Once
we
make
sure
the
docs
were
working
in
a
in
a
in
a
draft
PR
or
whatever
in
the
net
laughs
I
preview
mm-hmm,
then
we
would
go
rename.
The
bran
make
sure
the
the
links
in
the
net
laughs,
I
preview
are
working.
I
might
have
to
update
my
PR
to
merge
it
and
then,
and
then
do
the
rename
and
fix
up
any
open,
PRS
yeah.
D
B
D
B
Okay,
yeah
that
I
actually
like
that
suggestion
that
we
don't
simply
rename
it
and
that
way
we
can
I
think
we
do
have
to
can't
we're,
probably
not
going
to
keep
master
around
for
a
long
time.
So
we
have
the
mouth
for
some
some
time
period
like
yeah
a
few
or
maybe
right
right.
We
say
this
set
time
period,
we'll
keep
both
yep.
A
A
F
F
B
D
I
am
almost
done
them.
I
will
give
you
that
major
update
on
the
only
major
thing
that
I
have
in
the
in
my
review
is
like
moving
the
creation
of
the
CRS
into
the
plugins
itself
and
not
a
little
create
it.
That
way,
we
can
like
one
minimize
the
change
this
in
code
bolero
and
also
not
create
these
see
ours
for
volume,
snap
shutters
that
don't
support
this,
so
that
was
the
only
major
thing
other
than
that.
B
Anybody
who's
developed
on
velaro
may
have
noticed
that
when
you
run
make
the
Builder
image
just
constantly
excuse
me
got
rebuilt,
which
led
to
long
run
times
and
definitely
finally
fixed
that,
and
that
is
super.
Super
nice
and
I
think
he
worked
with
the
team
of
some
some
Cooper
Nerys
architects
here
at
VMware
and
then
Tony
petard
and
the
team
of
communities.
Architects
worked
on
getting
some
go,
linters
I
forget
the
project,
the
whole
Suites
name,
but
he
got
these
linters
added
to
not
only
yeah
going
once
he
are
going
see.
B
The
repo,
but
also
added
to
our
CI,
which
is
super
nice,
so
huge
thanks
to
Tony,
I,
think
Mike.
Trudeau
ball
was
involved
and
Duffy
as
well.
So
if
you
contribute
make
a
contribute
to
PR,
you
may
see
the
winter
yelling,
and
if
this
applies
to
everyone,
if
you
see
some
limping
rules
that
are
kind
of
onerous
and
you
think
may
change,
we
have
this
golang
CI
dog
Amal,
which
we
kind
of
just
accepted
the
defaults
right
now.
B
But
these
are
definitely
the
place
to
debate
the
code
code,
standards
and
code.
Formatting
things
in
my
opinion,
is
not
individual
PRS,
but
NPR
on-the-go
laying
CI
y
mo
file,
just
my
opinion
yeah.
Thank
you
very
much
to
those
contributors.
Thank
you
to
everyone.
Who's
attended,
the
restore
hooks,
meaning
yesterday.
That
was
very,
very,
very
useful.
B
As
a
very
quick
summary
there
we're
going
to
basically
try
to
merge
the
two
approaches
that
we've
discussed,
putting
in
it
containers
and
in
place,
restore
hooks
to
kind
of
meet.
Both
these
cases
and
also
everybody
who's
submitted,
PRS
and
has
paid,
has
had
patience
with
the
team.
Your
patience
is
very
much
appreciated.
So,
thank
you
very
much.