►
From YouTube: Harbor Community Meeting - March 24, 2021
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
A
Let
me
go
to
the
community
agenda.
Yeah,
hey
everyone!
So
welcome
to
the
harvard
community
meeting
today's
march
24th.
Yes,
today,
mostly,
we
just
have
two
items.
The
first
thing
is
declarative
config.
You
might
have
heard
of
this
there's
a
there's,
a
harper
issue.
A
Here,
that's
very
popular.
You
know
talking
about
some
of
the
issues
around
configurations
in
harper
today
and-
and
you
know,
we
talked
about
how
we
are
planning
to
make
this
a
little
better
in
the
upcoming
2.3.
A
Do
a
quick
summary,
okay
yeah!
So
essentially
we
made
a
pretty
big
change
to
1.8
before
harvard
version
1.8,
but
for
hardwood
version
1.8.
A
You
know
I
could
configure
harper
to
use
external
and
some
other
things
via
a
config
file
or.
A
Variables
and
then
we
took
that
away
and
then
so
with
that
you
lose
some
flexibility,
and
I
guess
people
really
didn't
like
that.
A
If
you
look
at
the
number
of
thumbs
up
on
this
issue,
had
some
pretty
good
discussion
throughout
the
issue
and
if
you
read
the
thread,
I
think
you
know
some
other
customers
or
users
are
not
wanting
to
use
the
ui
at
all
right
and
they
want
to
apply
configs
that
live
in
their
their
console
management
in
cicd,
and
so
one
of
the
biggest
problems
that,
with
the
current
design,
I
think
is
that
you
know
configurations
that
you
make
via
api
can
still
be
overwritten
by
the
front
end
right.
A
The
web
console
so
there's
a
bit
of
a
config
drift
in
that
sense
and
for
larger
companies
enterprises
they
have
you
know
compliance
or
auditing
requirements
to
be
able
to
track
all
the
configurations
made
right.
So
if
the
conflicts
live
in
the
std,
then
you
can
see
exactly
who
made
the
push,
what
what
changes
they
made
and
they
could
get
the
entire
history
of
all
the
changes
made
in
that
harper
instance
in
the
history
of
that
heartbreaking.
A
Sincere
so
yeah
I
mean
that
would
be
very.
That
would
be
very
possible
with
the
audit
configs
with
config
files,
but
with
apis
it's
difficult
to
validate
the
settings.
You
know
that
you're
making
your
api
like
it
would
be
difficult
to
do
a
lot
of
introspection
of
the
api
itself.
You
know
there's
solutions
in
ansible,
there's
other
solutions
but
like
for
this
issue
right-
and
I
talked
about
this
towards
the
end,
which
is
our
thinking.
A
Is
that
we'll
we're
attempting
to
find
out
the
exact
set
of
configurations
that
should
be
managed
by
con
to
buy
it
by
ci
and
then
we're
drawing
that
line
so
you're
not
modifying
it
in
both
config
and
in
api
right?
So
maybe
we
have.
A
Maybe
we
have
some
sort
of
like
base
configurations
right
things
that
are
meant
to
be
managed
once
or
not
as
frequently
by
the
harper
system
admin.
Maybe
it's
just.
Maybe
it's
just
based
on
persona
right,
like
the
harvest
assignment
which
deals
mostly
in
you
know,
off
control
in
setting
to
backing
store.
A
All
of
these
things
should
be
managed
by
a
config
and
then
the
more
you
know:
user
phasing
controls
for
the
other
hardware,
personas
like
the
developer
or
the
guest
things
like
that
could
be
managed
by
by
by
gui
by
api,
that's
sort
of
the
basic
idea
behind
what
we're
trying
to
do
here
with
with
this
dick
ticket
anything
else
you
want
to
add
daniel
or
stone.
I
know
your
stone
is
working
on
the
actual
implantation.
C
Yeah,
can
I
share
that
my
screen?
Yes,.
A
C
C
Sorry
I
have
changed
my
desktop.
I
have
to.
C
A
But
yeah
maybe
like
breathe
through
a
thread
and
see
what
they're
asking
like
at
the
last
comment
right
they're
asking
for.
I
think
more
than
what
we're.
A
E
All
right,
vadim,
you
had
a
question.
D
I
think
also
the
kubernetes
infra
guys
were
also
looking
after
or
looking
to
find
a
solution
for
themselves
regarding
configuring
harbor
from
from
their
ci
as
well.
So
they
had
also
a
few
requirements
there
and
I
think
alex
you
mean
that
they've
also
wanted
to
create
projects
and
things
like
that
from
configuration
as
well.
B
A
What
they're
asking
for
is
what
they're
asking
for
is
deploying
multiple
instances
like
across
different
cloud
providers
all
over
the
world
right
to
serve
the
kubernetes.
D
A
There's
a
ticket
called
cross
collar
cross
cluster
harbor
for
the
kubernetes
project,
so
yeah
I
mean
I
guess,
they're
going
to
be
they're,
going
to
be
doing
a
bunch
of
things
and
they're
not
going
to
be
wanting
to
to
configure
all
this
by
hand
on
the
web
console
and
they're
going
to
be
using
the
same
configs
across
all
the
different
instances.
A
You
know
what
they
want
to
do
is
bootstrap
the
installation
of
multiple
hardwares
on
different
cloud
providers
and
then
using
a
single
config
file
to
propagate
all
the
same
settings.
Right
like
all
the
the
I
imagine
the
project's
name
are,
are
set
up
in
a
way
that
you
know
works
with
their
ci
cd,
so
yeah.
I
think
that
that's
the
the
they'll
be
part
of
the
users
I
was
talking
about
in
terms
of
not
wanting
to
use
the
ui
in
any
way,
possibly.
F
But
I
have
to
interrupt
because
that's
a
different
requirement
product
is
not
configuration.
Project
is
user
data.
A
F
You
don't
clarify
that
this
proposal
will
not
cover
that
issue
and
I
don't
think
they
ask
the
right
question.
There
are
different
ways
to
do
that.
They
can.
You
know,
set
up,
however,
once
it's
running
and
call
some
api
to
populate
the
data
and
that's
a
totally
repeat,
leakable
solution,
but
yeah.
We
can
discuss
that
in
some
other
when
we
have
a
chance,
but
I
don't
think
that's
the
same
requirement
like
this
one.
A
A
On
on
on
the
kubernetes,
kate's
or
decades
uniform
working
group
meeting,
there
was
some
pushback
on
using
harbor
because
of
the
the
non-declarative
nature
of
harper.
You
know
in
reference
to
this
ticket
in
particular,
but
yeah
they
were.
They
were
wanting
more
than
what
we're
trying
to
cover
here,
but
you
know
what
we'll
do
it
in
iterations?
The
first
step
is,
I
think,
daniel.
Do
you
have
the
list
of
the
actual
things
that
will
lock
down?
C
Okay,
can
you
see.
C
Yeah
yep,
okay,
let
me
introduce
the
background
of
the
issue
in
the
early
release
of
the
hubble.
C
Maybe
in
1.5
or
7.6,
the
hubble
configuration
can
be
initialized
by
an
environment
variable
like
hubble
or
the
cfg
to
initialize
the
carbon
configuration,
but
with
the
growth
of
the
hubble
and
more
and
more
configuration
item
added
to
the
hubble,
and
this
cld
file
seems
uneasy
to
maintain,
and
then
we
decide
to
separate
the
system
configure
and
the
user
configure
for
system
configure,
it
can
be
maintained
in
hardware.a
can
be
maintained
in
harbor,
yama
and
the
for
the
user
configure.
C
It
only
can
be
maintained
in
a
web
console
of
the
hubble,
but
as
we
discussed
in
in
the
some
ci
cd
background
and
devops
and
most
of
the
users
prefer
to
use
the
script
or
command
environment
file,
it
is
a
very
inconvenient
to
change
the
configuration
by
ui
or
reside
api.
Here.
C
Another
way
we
can
and
the
option
two
is
that
we
can
provide
a
multiple
environment
variable
in
the
core
container.
So,
for
example,
we
can
provide
the
host
mode.
I
double
based
in
and
advanced
search
dn
in
this
eme
file
of
the
core
container.
C
Once
it
is
once
the
user
setting
is
set
up
is
is
set
by
the
environment
variable
then
the
configuration
configuration
is
changed
to
reader-only
model
and
only
and
can
be
changed
by
the
environment,
variable
and
restart
the
hubble
restart
the
core.
C
Container,
this
is
the
configure
panel.
When
the
configuration
item
is
in
reader
only
mode,
then
the
configure
panel
becomes
really
only.
This
is
the
only
configuration
panel.
C
In
this,
in
a
configuration
manager
we
have
to,
when
there
is
a
link,
configure,
json
environment
variable
here,
we
will
detect
the
environment
variable
and
if
it
is
said,
and
then
it
will
set
all
the
configuration
item
here
and
the
the
config
that
will
only
flag,
we
are
become
true,
and
this
is
the
option
one
and
for
option.
Two.
C
The
user
setting
will
be
set
by
environment,
variable
multiple
environment
variable.
It
is
defined
in
a
metadata
list,
a
manager
list
here,
for
example,
if
we
need
to
define
a
adapter
based
unit,
we
should
use
an
environment
key
to
define
adaptation
and
if
we
need
to
define
whatever
filter,
we
should
set
up
an
environment.
Add
up
a
filter
to
set
up
a
list
in
one
user
settings.
C
Once
the
environment,
once
this
feature
is,
is
enabled
and
triggered
by
the
this
environment
is
set,
then
all
the
settings
and
the
reader
only
flagger
will
become
true
and
all
the
configuration
in
the
ui
is
on
is
read
only
and
the
api
to
the
the
configuration.
A
C
Method,
the
impact
of
this
feature:
if
this
configuration
is
initialized
by
the
environment,
then
all
configuration
item
in
a
in
the
ui
or
addressed
api
is
read
only
cannot
be
updated.
F
C
F
B
C
But
for
for
others
for
other
settings,
if
they
didn't
want
to
set,
they
just
keep
it
blank.
A
F
C
F
B
C
B
F
C
Yeah,
I
think
for
this,
for
the
first
option
it
will,
it
will
have
required
a
user
to
maintain
json.
F
Yeah,
and
in
addition
to
that,
can
you
clarify
if
let's
say
we
choose
this
init
configuration
approach,
will
this
value
be
persisted
in
the
database
or
not.
C
No,
this
one,
the
configuration
and
the
indian
configuration
is
provided.
C
And
once
yeah,
I
think
it
is
correct,
because
once
you
have
updated
the
configuration,
I
think
it
is
better
to
to
store
in
a
single
store
right.
F
F
C
C
A
F
I
mean
I
mean
different
start.
If
I
restart
hover
without
setting
this
environment,
then
all
the
data
is
gone,
but
if
I,
if
I
there
may
be
the
case
that
user
set
this
environment
variable
and
later
he
decide,
I
would
you
know,
enable
the
configuration
api
and
it's
okay
for
me
to
let
add
mean
to
edit
my
setting.
B
F
D
B
D
I
think
the
you
know
from
from
from
the
past,
most
problems
arise
if
you
have
kind
of
a
data,
synchronized
and
duplicated,
so,
for
example
like
if,
if
the
user,
normally
that
case,
if
you
deploy
application,
you
set
the
environment
variables
and
you
do
redeploy
and
you
change
the
variable
and
change
the
applications.
But
if
it's
in
the
database
it's
copied
into
the
database
and
then
it's
not
clear
which
value
is
actually
the
source
of
truth,
is
it
the
database
or
is
it
the
value?
D
D
H
D
D
F
F
But
yeah:
well
maybe
we
may
need
to
think
about
this.
I
I
don't
have
a
quick
answer
but
yeah,
probably
if
you
think
it
through
j
stone,
you
know,
keeping
things
in
the
environment
variable
will.
D
Yeah
but,
for
example,
the
the
point
that
you
mentioned,
you
just
remove
the
environment
variable
and
then,
in
this
case
you
expect
the
data
to
persist
there.
You
know
it's
kind
of
a
contradicting
to
to
overwrite,
always
the
the
database
from
the
init
configuration
file,
so
you
know
like
if
it's
empty,
then
it
will
overwrite
with
empty
values.
This
would
be
the
correct
behavior,
but
this
is
not
what
you
actually
expected.
You
expect
that
if
it's
not
there,
it
should
not
reset
it.
Basically,
so.
A
D
That's
I
mean
the
thing
is
like
it
there's
a
valid
point
in
that
definitely,
and
it
makes
sense
for
the
user
not
to
break
and
and
not
to
have
surprises,
but
on
the
other
hand,
it
will
open
up
a
lot
of
edge
cases
that
need
to
be
handled
along
the
way
and
and
find
and
find
solutions
for
that
you
know
it's
this,
I
think,
on
from
a
maintenance
point
of
view.
It's
it's
a
lot
of
effort
for
for
little
value.
F
Satisfy
the
requirement
yeah,
but
I
think
we
need
to
ensure
one
thing
that
if
users
use
this
innate
config
to
initialize
harbor
and
use
it
for
a
while,
then
this
harbor
instance
will
always
need
a
new
configuration.
If,
if
this
environment
is
gone,
if
it's
gone,
harbor
should
fail
right
away
to
to
to
explicitly
like
user
know,
I'm
not
working,
because
I
need
the
unit
config,
because
I
was
initialized
with
this
in
its
config.
We
need
to
do
this.
If
we
do
not
exist
the
value
yeah,
maybe
it
should
not
be.
F
F
If
a
user
chooses
to
configure
hardware
using
this
way
as
proposed,
then
we
need
to
make
sure
after
restarting
or
upgrade,
you
know
this
innate,
config
or
config
override
always
exists.
C
I
think
we
have
to
add
a
configuration
item
to
indicate
that
it
is
in
a
need
by
configure
json
right.
C
F
D
F
Workflow
you
mean
ignore
config.json
or
ignore
the
database
setting.
I
I
think
it
would
be
safer
if
we
ignore
this
config
json,
because
there
are
already
data
in
the
database,
otherwise.
F
C
It
will
cause
some
problem
if
we
have
to,
for
example,
the
devops
change
the
environment
variable,
but
he
did
not
find
that
that
the
this
configuration
item
is
not
a
change
in
the
ui
right.
D
Yeah
so
because
no
the
normal
workflow
is,
if
you
upgrade,
and
you
didn't
had
the
config
json
file
set
before
and
you
now
explicitly
set,
so
the
user
knows
about
the
the
variable
and
he
sets
it
because
he
wants
to
configure
it
from
from
the
from
the
config
file
until
I
think
it
said,
because
if
you
don't
do
nothing,
it
just
upgrades,
so
it
will
be
kind
of
a
sunny
path.
You
know,
if
you
don't
change
anything,
you
do
the
upgrade
and
it
will
work.
D
But
if
you
say
after
the
upgrade,
you
want
to
do
the
config
file.
Now
you
set
the
config
file
and
then
it's
kind
of
explicit
decision
to
say:
okay,
I
wanna
have
it
the
country
file
so
that
the
environment
variable
has
the
last
word
in
there,
because
this
is
a
kind
of
explicit
decision.
You
need
to
set
this
variable
as
as
an
administrator,
it's
not
something
that
that
is.
There's
no
default.
B
F
F
C
Because
when
you,
when
there
is
an
adapt
user
in
a
in
a
user
table
and
now
you
and
just
change
the
internet
configure.json
to
change
it
to
db,
it
is
not
allowed
right
right
yeah.
I
think
we
can
fail
when
the
there.
A
D
Once
you
have
users
in
the
database,
you
cannot
change
the
auth
mode.
F
B
C
F
F
D
A
F
A
All
right,
I
think
we
had
a
good
yeah.
I
think
today.
F
A
Should
document
all
the
cases
like
what
happens
when
you
make
the
switch
midway?
What
happens
when
you
wanna
change?
Some
of
this
stuff,
yeah
it'd
be
good
to
document.
All
this.
All
these
scenarios
stone.
I
wanna.
A
Over
I
wanna
talk
about
nidus
now
and
give
lion
some
time
to
talk
about
what
knight
is
the
the
demo
that
he
created
and
yeah
like
what
we're
planning
to
do
with
nidus.
A
So,
okay,
not
just
do
you
wanna,
take
it
from
here.
Thank
you.
G
G
G
So
the
the
the
benchmarking
result
shows
that
the
performance
improvements
compared
with
the
oci
image,
like
as
the
image
size,
increases,
the
continuous
startup
time
of
nether's
image
remain
was
very
short.
So
if
you
want
to
know
some
details
about
neither
just
to
go
to
the
github
repo,
but
there
are
also
open
source
projects,
you
can
get
more
details
of
documentation
about
that
so
to
in
order
to
leverage
the
benefits
of
of
nadas.
G
The
first
thing
that
we
have
we
need
to
do
is
convert
the
uoci
format
container
image
into
another
format.
Container
image.
Another
provides
a
tool
that
names
95
95
to
do
that.
So
today's
dynamo
is
just
about
integrate
net
it's
file
into
hardware
and
the
basic
idea
is
to
leverage
the
hardware
job
service
capacity
and
the
general
workflow
is
when
a
user
push
an
image
into
arbor.
G
Once
a
deep
down
hardware
call
will
send
a
notification
to
the
job
service
and
then
the
job
service
will
trigger
a
convert
job
in
the
backend
to
convert
the
incoming
image
and
push
back
to
harbor
court.
To
finish
the
workforce,
I
have
to
say
that
it's
not
a
official
integration
scenario
just
by
the
showcase
to
demonstrate
that
that
harbor
can
have
the
capacity
to
integrate
with
nadas.
G
So,
let's
start
the
timer.
G
So
after
logging,
with
the
project
or
system
admin,
there's
a
new
items
called
automatic
convert
is
quite
similar
with
the
auto
scan.
So
when
intense
that
the
hardware
will
automatic
to
convert
the
incoming
darker
image
so.
G
Yeah,
so
here
we
have
all
over
lettuce,
but
I
haven't
remembered
can
work
so
I
let
me
make
that
and
let
me
push
that
image.
G
G
So
yeah
yeah
you,
as
you
can
see,
there
is
a
new
image
created
by
the
nethers
that
we
with
the
suffix
status
in
its
type
this.
This
artifact
is
created
by
the
next
fight
and
push
back
to
harbor
court.
So
so
you
can
see
there
are
two
artifacts
in
the
same
people.
You
can
still
mine.
G
It
manage
this
artifact
as
as
others
in
hardware,
but
this
is
an
another
format
image
that
is
for
my
image
and
another
thing
here
that
for
the
existing
image
already
pushed
into
hardware,
we
can
provide
a
new
api.
That's
called
the
notify
to
convert
the
existing
ocean
from
my
image.
I.
G
G
A
G
So
you
just
can
see
that
hardware
can
manage
the
another
image
you
can
push.
You
can
convert
your
ocx
from
my
image
into
the
menus
in
my
image
and
then
my.
G
G
G
Yeah,
I
I
I
don't
know
another
environment
on
my
machine
like
what
you
have
to
set
up
matters
tools
and
the
continued
debugging
to
launch
the
madness
for
my
image.
A
A
A
G
A
Some
background
on,
like
what
can
we
do
with
this
like?
What
is
our
goal
in
integrating
against
it?
Besides,
hey,
we
can
support
nice
images
and
you
can
you
can
store
those
on
harbor
and
you
can
you
can
pull
those
off
harbor
there's
a
lot
of
value
in
that,
but
I
was
just
wanting
to
understand
a
little
more
about
knight
as
itself,
because
you
know
you
spent
quite
some
time
on
it.
G
E
So
I'm
looking
through
an
older
issue
here
from
a
couple
of
a
few
months
ago,
yeah
january
over
in
the
repository
there
talking
about
how
to
use
knightonitis
with
harvard
as
the
registry
back-end,
it
seems
like
that
was
already
working.
So
this
is,
but
this
is
the
transformation
integration
here,
but
is
that
right.
G
E
So
we
we
could
essentially
say
that
yes,
nidis
works
with
harbor
as
well,
just
like
we
do
with
singularity
in.
A
A
Got
it?
Thank
you
volume.
I
think
you
know
that
is
helpful.
I've
I've
heard
a
lot
about,
and
I
just
recently
on
some
of
the
conferences,
and
so
we
should
continue
to
explore
that
integration,
but
it's
good
that
you
know
these.
Can
we
can
host
and
carry
these
things
on
harper?
So
it's
good
great
first
step.
A
A
2.2.1,
I
believe
we
do
monthly
patches,
so
I
know
this
month
telling
do
you
have
a
an
update
here.
I
Yeah,
actually
we
have
our
say
out
last
week
and
currently
we
are
working
on.
I
mean
we
observed
a
few
new
issues
here,
so
we
are
currently
evaluate.
What
do
we
need
to
include
them
into
2.2.1?
A
I
A
A
E
No
not
right
now
we
will
have
a
little
little
announcement.
Mid-April.