►
From YouTube: Velero Community Meeting/Open Discussion - Jan 21, 2020
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Hi
everyone
and
welcome
to
another
episode
of
the
the
Valero
community
meeting,
slash
open
discussion.
I
can't
speak
today,
we're
gonna
walk
through
a
bunch
of
status
updates,
and
then
we
have
some
discussion
topics
and
yet
we'll
do
it
as
we
usually
do
so.
I'll
share
my
screen
here,
so
everyone
can
see
the
agenda.
If
you
have
any
anything,
you
want
to
add
into
the
agenda.
I'm
gonna
share
the
link
as
well
so
started
here
there
we
go
and
first
up
for
status
updates.
We
have
Nolan.
B
B
It's
it's
trying
to
cover
a
lot
of
ground,
and
it's
it's
also
trying
to
figure
out
how
to
work
together
to
start
out
with
so
there's
gonna
be
a
lot
of
different
topics
that
they're,
covering
and
I
think
they're
going
to
try
to
split
off
into
smaller
groups
that
will
try
to
tap
tackle
individual
things
and
then
come
back
together
in
the
larger
group
on
an
alternating
weekly
basis
and
discuss
those
with
the
larger
group.
I.
B
Think
the
the
issue
that
we're
going
to
try
to
participate
on
is
kind
of
the
larger,
the
larger
goals
of
like
what.
What
does
backup
look
like
for
kubernetes?
Where
do
the
backup,
workflows
and
define
kind
of
the
define
the
high-level
problems
and
then
work
on
solutions
from
there
so
agree
with
community
on
what
the
high
level
problems
are
and
then
see
what
we
can
do
about
working
on
common
solutions
from
there
and
got
folks
from
Google.
B
Traditional
backup,
vendors,
like
Vienna
and
Veritas
and
yeah
I,
can
go
look
up
some
some
links
to
Docs
for
that
and
toss
them
in
this.
When
I'm
done
with
my
update
and
then
for
velaro
last
week,
I
mentioned
there
was
the
CRD
issue
on
restoring
customer
resource
definitions
and
just
still
working
on
this
pull
request,
because,
as
we
got
four
along
with
this,
we
actually
realized
there's
a
couple
different
issues
at
play
here.
So
we
have
this
with.
B
Get
it
done
tested
and
then
what
we
found
out
was,
while
this
is
useful,
there's
another
issue
at
hand
that
Valero
doesn't
actually
recheck
the
kubernetes
api
server
when
it
restores
a
custom
resource
definition.
So
what
will
happen
is
in
in
some
cases
you'll
restore
the
custom
resource
definition,
but
then
the
custom
resources
won't
show
up
in
the
same
restore
action,
so
you'll
have
to
call
you'll
have
to
run
velario
restore
again
and
then
your
custom
resources
will
show
up
so
we're
Steve
and
I
sat
down
and
and
thought
through.
B
Basically
after
we
restore
a
custom
resource
definition,
we
update
our
list
of
resources
from
the
kubernetes
api
and
then
continue
on
with
the
restore
so
hoping
to
get
that
done
this
week
and
get
a
v12
one
out,
and
this
this
goes
with.
There's
a
I
forget
the
PR
number
of
that
that
goes
hand-in-hand
with
another
PR,
that's
related
to
restores
and
then
finally,
for
CSI.
There's,
there's
not
a
whole
lot
of
update
on
that
front.
B
B
Maybe
honestly
I
don't
know,
I'll
have
to
look
at
it
again.
The
happier
solution
was
the
the
initial
pass
is
basically
I.
B
Think
I
think
I
I
need
to
clean
it
up
a
little
bit
because
there's
there's
some
literally
copy
pasted
code
in
there.
That
can
definitely
be
cleaned
out,
but
basically
what
the
the
code
does
there?
Is
it
reruns
it?
We
do
our
restore
loop
and
then,
after
that
we
re
query
the
kubernetes
api
server
and
say
hey,
what's
new
stuff,
that
we
didn't
that
we
see
now
that
we
didn't
restore
before,
and
then
we
run
through
that
we
run
through
our
tarball
again
and
restore
anything
that
we
didn't
see.
B
I'm
was
mostly
running
that
on
a
very,
very
small
test,
so
before
I
say
that
we
would
commit
to
that.
I
would
want
to
test
it
on
some
other
things,
but
that
may
be
an
option
and
then
we
could
test,
but
like
it
worked
for
a
small
case
where
we
were
where
I
was
testing
with
cluster
API,
a
cluster
API
control
cluster
in
kind,
but
I,
don't
know
if
that
is
generalizable
at.
C
C
B
So
the
issue,
the
issue
that
it's
tying
to
this
is
2159,
so
there's
reported
by
our
field,
engineer,
Duffy,
Cooley
and
so
there's
actually
like
three
or
four
issues
at
play.
There's
there
was
a
the
CRD
api
extension
version
mismatch.
So
we
had
v
1
beta
1
and
V
1.
B
B
Then
it
also
introduces
some
the
this
PR
that
we're
talking
about
now
has
some
Valero
logic
fixes
where
we
wait
for
the
custom
resource
definition
to
be
ready
on
the
API
server,
so
that
we're
we're
not
running
into
this
race
condition
where
the
custom
resource
definition
is
restored.
But
it's
not
accepting
custom
resources
yet.
E
B
So
this
this
is
to
make
that
a
little
bit
more
complete
we're
going
to
adjust
the
behavior
so
that
Valero
will
re
query
for
API
groups
and
and
then
resubmit
any
new
API
groups
how
how
exactly
we
do
that
is
kind
of
the
up
in
the
air
part
right
now.
Steve
and
I
had
a
couple
different
implementations
and
we
talked
about
because
we
didn't
want
to
go
in
and
and
rip
up
a
ton
of
restore
code
yeah,
but
there's
like
as
far
as
we
know.
B
B
B
So
that's
been
sitting
open
and
we
wanted
to
make
Valera
wait
to
restore
the
custom
resource
definition,
make
sure
it
was
ready
and
then
move
on
and
then
finally,
it's
rediscover
all
the
a
the
API
groups.
Each
time
we
restore
to
make
sure
we
have
a
complete
view
of
what
the
urban
is.
A
API
server
has
okay.
B
When
we're
trying
what
I
was
trying
to
avoid
was
like
doing
a
to
pass
thing
because,
ideally
like
this,
would
require
redoing
the
whole
restore
logic.
Ideally,
what
I
would
like
to
see?
Is
we
scan
through
the
whole
backup
and
do
our
like
include
exclude
logic
and
say
this
is
the
directed
graph
that
we
have
that
we
want
to
restore
before
we
do
any
kind
of
API
server
logic
so
that
we
know
up
front
what
we
want
to
do,
but
that's
that's
way,
bigger
changes.
B
That's
that's
been
something
we've
wanted
to
do
for
a
while,
I
think,
but
it's
just
not
gonna
happen
for
a
dot
patch
release,
but
yeah
for
this
we
want
to.
We
want
to
say
each
time,
Weaver
and
really
there
can
be
optimizations
here
to
say:
oh,
we
just
got
done
restoring
a
custom
resource
definition
or
we
got
done
restoring
custom
resource
definitions.
As
far
as
I
know,
that's
the
only
thing
that
modifies
the
API
groups.
So
then
we
could
say:
hey
update
discovery,
I,
don't
think
anything
else.
B
Modifies
API
server
groups,
but
that's
kind
of
where
Steve
and
I
had
landed,
and
we
talked
about
like
optimizing
it
so
that
we
can
move
some
of
that
discovery
and
there
was
the
higher
level
restore
loop
and
some
things
like
that.
But
I
think
those
are
implementation,
details
and
and,
like
Steve
said,
we
might
be
able
to
to
move
some
of
that
off
to
optimizations
later.
C
Yeah
so
handfull
things
so,
first
of
all,
I'm
the
community
support
person
for
this
week
so
updated
the
slack
Channel
topics
and
I'll
be
keeping
an
eye
on
that
and
trying
to
get
through
any
of
the
open
issues
and
PRS
that
we
have
in
github
last
week
worked
on
a
handful
of
different
things.
So
I
got
a
PRN
for
this
issue.
C
We
have
a
prometheus
metric,
which
is
a
gauge
metric
and
it
records
the
last
successful
time
stamp
for
each
backup
schedule,
and
the
problem
is
that
whenever
the
Valero
server
restarts
when,
when
a
new
pod
gets
created
or
when
the
pod
gets
restarted,
this
metric
gets
cleared
out
and
it's
never
reinitialized
so
took
a
look
at
this
and
put
in
a
relatively
straightforward
fix
to
just
re
initialize
this
metric
for
each
schedule
when
the
server
restarts
so
that
one's
ready
for
review.
Now,
whenever
everyone
gets
a
chance
to
look
at
it,.
C
Laughter,
wait
for
the
agenda
to
pop
back
up
yeah
there's
a
another
issue
reported
by
Dylan
Murray
from
Red
Hat,
where
rustic
restore
there
was
a
rustic
restore
that
failed,
but
the
velaro
restore
still
ended
up,
reporting,
successful
completion
which
shouldn't
happen.
So
if
a
rustic
restore
fails
and
the
overall
restore
should
fail.
C
So
I
took
a
look
at
this
and
it
looks
like
there
was
a
race
condition
and
some
of
the
code
that
was
essentially
collecting
results
from
the
rustic
processes
and
reporting
them
back
up
to
the
main,
restore
so
I
changed.
How
that
worked?
A
little
bit
basically
got
rid
of
some
of
the
code
that
had
the
race
condition
and
put
in
what
I
think
is
a
more
straightforward
solution.
So
I've
tested
this.
It
looks
okay
to
me,
but
this
one's
also
ready
for
review
and
every
folks
get
a
chance.
C
And
then
yeah
spend
time
looking
at
the
cigar
dearest
or
issues
that
no
one
talked
about
and
also
reviewed,
Carly
CEA's
initial
design
PR
for
the
install
and
configure
UX
there
and
had
some
discussions
with
the
team
that
I
think
mostly
covers
it
and
few
other
random
things.
Pr
is
an
issue
like
in
issues.
That's
about
all
for
me.
E
Yeah
so
as
Nolan
mentioned,
I
wasn't
in
the
office.
I
was
in
Palo
Alto
last
week
or
most
of
last
week,
but
I
was
able
to
make
some
progress
on
like
the
GC
pcs
driver
CSI
driver
that
was
kind
of
I
was
I.
Wasn't
able
to
install
that
or
set
up
my
project
to
install
this
year
as
a
precursor
to
the
CSI
driver,
so
I
kind
of
worked
around
that
and
unblocked
myself,
yeah
I
will
be
installing
the
CSI
driver
on
spinergy
ke
cluster.
E
D
I,
don't
think
anybody
mentioned
that
yesterday
was
a
you,
have
us
holiday
and
also
a
holiday
for
VMware
folks.
So
today,
I'm
going
to
start
with
me,
PRS
I
think
I
did
drop
the
ball
on
that.
The
last
two
days
of
last
week
and
I
focused
on
this
PR.
Last
week
we
also
had
a
meeting
between
the
Valera
members
core
members
and
we
recorded
it
is
an
ad
that
I
added
the
link
it's
posted
on
on
our
YouTube
playlist.
If
anybody
was
is
interested
in
going
through
what
we
discussed.
D
Everything
isn't
that
PR
I
do
need
to
update
the
PR
with
the
comments
that
were
made
already,
and
one
thing
that
I
have
to
do
is
right,
I'm
working
on
this
right
now.
What
changes
in
the
world
that
we're
going
to
need
to
do
to
make
this
wife
function
and
now
so
one
thing
that
came
out
of
our
meeting
is
that
this
seer
life
clone
is
too
big,
get
ups
compatible
and
it
might
be
a
matter
of
making
it
compatible
with
customize
or
was
also
looking
at
how
customized
works
for
any
rates.
A
C
I
just
wanted
to
quickly
mention
I,
know
everyone's
busy
and
working
on
stuff,
but
we
got.
We
got
a
bit
of
a
PR
backlog,
so
if
everyone
can
take
some
time
this
week,
preferably
earlier
and
then
we
can
and
just
go
through
things
and
do
some
reviews,
that'd
be
great.
I've
tried
to
keep
them
on
the
Zen
board,
somewhat
ordered,
so
the
ones
that
are
at
the
top
of
the
review
QA
pipeline
are
the
ones
that
are
kind
of
ready
for
merge,
assuming
that
they
are
okay
on
review.
C
F
I'm
doing
good
so
I'm
working
in
on
open
up
this
project
and
we
have
a
better
plugin
for
our
sister
data
engine.
Okay.
So
now
we
are
working
on
the
data
protection
project
and
we
are
considering
to
use
valor
role
for
snapshot
and
restoring
the
snapshot
so
recently
in
value.
We
have
one
change
regarding
the
period
enemy
so,
but
that
change
our
value
is
not
considering
giving
them
from
the
snap
shuttle
from
the
plugin.
F
So
in
C
store,
we
are
creating
some
storage
assets
when
restoring
the
snapshot.
So
this
storage
storage
assets
have
name
which
is
similar
to
the
PV
nom.
Now,
while
renaming
the
PV
value,
creates
PV
with
the
valid
of
clone
some
view
ID
and
here
in
systole,
we
have
this
story
assets
with
some
different
name.
So
here
comes
the
usability
issue,
because
user
won't
be
able
to
figure
out
the
storage
ourselves
from
the
PV
memory.
F
B
F
E
C
E
G
B
F
A
F
B
H
H
The
program
running
into
is
I'm
trying
to
run
some
command
before
I
backup,
let's
say
like
a
my
sequel,
port
and
right
now,
with
the
pre
backup
hook,
I
can
actually
run
the
command,
but
the
issue
is
that
they
say
once
who
reference
some
secret.
This
is
some
credentials
to
connect
to
my
sequel,
server
and
I
want
to
read
those
secret
from
within
the
hook
command.
H
C
Hey
Anthony
I
would
I
would
think
I
mean
you
could
one
option
would
be
to
just
have
that
secret
mounted
into
the
pod
that
you're
running
the
hook
in
so
that
you
can
just
read
it
as
a
file
or
an
environment
variable
or
something
you.
You
could
also
have
your
hook.
Just
have
your
hook
execute
a
bash
script,
that's
included
in
your
part,
and
that
you
know,
if
you
have
a
script,
it
may
be
easier
to
actually
write
the
code
to
query
the
secret
as
the
two
thoughts
I
had
off
the
top.
My
head
yeah.
C
H
C
Nothing
specific
now
so
I
mean
the
you
know,
part
of
the
thinking
behind
adding
the
hooks
capability
was
to
give
users
the
ability
to
sort
of
configure
their
configure
commands.
That
would
enable
them
to
do
that,
but
we
don't
have
any
any
specific
code
within
bolero
to
deal
with
that.
No
okay,
yeah.
A
C
Sure,
first
up
we
have
J&L
tea,
I.
Think
John's
on
the
call
and
put
in
a
PR
last
week
for
updating
some
permissions
on
I.
Think
maybe
I
have
a
typo
here
in
the
description,
but
updating
permissions
on
the
velaro
config
directory
and
file
just
to
make
things
a
little
bit
more
secure,
so
appreciate
that
change,
and
then
we
had
a
Segoe
Scott
from
Red
Hat,
who
also
put
in
a
PR
to
fix
a
bug
around
rustic
repositories
with
a
the
last
maintenance
time
field.