►
From YouTube: Jenkins Platform SIG Meeting, Nov 19, 2020
Description
The Jenkins Platform SIG has met to discuss the status of multi-platform Docker images, Windows Docker images for LTSC 2019, membership in the OpenJDK Quality Outreach program, future of Docker packaging for Stretch, and adoption of the new Plugin Installation Manager tool. Full agenda and meeting notes: https://docs.google.com/document/d/1bDfUdtjpwoX0HO2PRnfqns_TROBOK8tmP6SgVhubr2Y/edit#heading=h.r97hloffu9u4
A
A
So
this
is
what
I've
got.
Is
this
readable
enough
for
everyone
on
your
screens,
I'm
using
a
tiny
size
screen
and
I'm
not
sure
what
works
for
you
and
what
doesn't.
A
Great
all
right
super,
so
what
we'll
do
is
we'll
review
the
open
action
items
oleg
had
had
suggested.
We
might
put
a
quality
outreach
program
topic
on.
I
would
hold
that,
if
he's
here
and
leave
it,
if
he's
not,
we
can
talk
briefly
about
it,
but
then
adopt
open
jdk
for
docker
on
debian
retirement
plan
for
stretch
is
the
topic.
C
A
Gareth,
I
suspect
we
need
to
talk
about
windows
and
docker
images
for
windows,
docker
images
for
ltsc
2019.
A
Sure
sure,
so
is
this
a
discussion
about
its
structure
or
what
what's
what's
the
topic.
A
A
I
did
the
sampling
of
the
data
for
the
docker
image
so
for
the
controller
image,
the
various
agents
and
those
samples
told
us
that
we
are
relatively
okay
right
now
for
operating
system
support
the
big
gaps
or
the
big
danger
zones
are
doc
or
debian
9
ends.
A
And
we've
got
some
outdated
java
version
on
the
debian
9
image.
It's
jdk8u
242
instead
of
272.
A
So
the
proposal
will
be
to-
and
I
think
I
mentioned
it
here-
that
the
proposal
is
that
we
are
on.
We
run
the
current
jdk
on
each
image
and
we
want
to
run
the
let's.
Let
me
put
it
in
the
list
here
proposal.
A
A
B
Yeah,
so
I
looked
this
because
we're
currently
just
using
the
repository
directly,
that
comes
that's
directly
forced
into
us
and
it's
using
just
open
jdk.
B
So
if
we
want
to
make
sure
that
we're
using
the
same
version
on
all
images,
we
would
need
to
update
the
install
and
use
the
adopt
open,
jdk
yam
repository,
which
I
can
put
together.
Br
for.
A
B
A
A
Tool
is
now
used
for
for
the
the
standard,
install
instructions.
B
It's
also
used
in
the
all
the
windows,
controller
images,
there's
no
install
plug-in
script,
it's
just
the
plugin
installation
manager,
tool.
A
E
It
raises
importance
of
fixing
the
remaining
regressions
in
plug-in
installation
manager,
because
there
are
bugs
in
this
tool
and
yeah,
so
they
become
it
becomes
official
once
we
support
yes
with
this
version,
it
might
lead
to
some
problems.
E
E
E
A
E
E
The
downside
that
plugin
installation
manager
tool
doesn't
fully
verify
the
dependency
tree,
so
it
means
that
in
some
conditions
you
may
end
up
with
a
broken
checkpoints
instance
or
stop
sync
up.
E
A
Well-
and
it's
important
important
point
that
you
know
that
may
lobby
that
we
should
not
add,
we
should
not
obsolete,
install
bashplugins.sh
until
we're.
We
have
more
confidence
that
all
the
use
cases
are
covered
well
enough
with
plug-in
installation
manager.
We
may
make
plugin
installation
manager
the
preferred
choice,
but
better
that
it
we
use.
We
have
it
still
available
to
use,
install
plugins.sh
great.
A
All
right,
okay,
I
had
the
action
item.
Oh
anything
else
on
installplugins.sh
alex
or
oleg.
B
B
A
And
have
we
seen
any
any,
can
shouting
outrage
or
concern
red
flags
raised?
Okay,.
E
Yeah,
the
script
was
duplicated
in
2015-16.
It
hasn't
been
really
updated
for
a
while.
A
A
A
E
Can
maybe
for
these
topics
it
makes
sense
to
create
tickets,
for
example
on
jenkins
scion.
Somebody
else
could
potentially
pick
them
up.
A
A
Jim
you
want
to
take
us
through
refinements
for
parallelization
and
multi-arch.
C
C
The
wrapper
is
really
utilized
for
ci
cd
implementations.
It's
pretty
easy,
you
just
say
hey.
I
want
to
build
debian
images
and
it
will
handle
all
the
tagging.
All
the
you
know,
votes
tagging
scanning
the
docker
scan
the
repo
for
docker
files
and
doing
all
that,
for
you,
it's
pretty
simple
to
use.
I
have
documentation
in
the
ci
folder
that
kind
of
goes
over
each
script.
The
parameters
that
you
pass
in
and
a
couple
examples.
B
C
Yes,
yes,
it
does
includes
the
verbose
tagging
and
then
keeping
pretty
much
all
the
same
tags
that
you
got
off
right
now.
I
mentioned
this
in
the
pr.
The
only
tag
I
didn't
carry
over
was
slim.
C
I
opted
more
for
just
clarifying
it's
debian
slim,
so
you
know
you
can
see
the
difference,
but
it's
easy
enough
just
to
go
in
there
and
add
in
slim.
It's
just
one.
If
statement
I
need
to
add
in
if
we
want
that
yeah,
the
the
only
other
thing
I
would
need
help
from
alex,
probably
on
the
parallelization.
C
I
did
not
touch
the
jenkins
file
as
I'm
not
really
that
verse
in
groovy.
That
much
so
I
would
need
some
help
kind
of
adding
the
parallelization
to
the
jenkins
file
and
pointing
to
the
right
scripts
to
call-
and
I
I
mention
these
two
points
in
the
the
pr
as
places
I
need
help
to
kind
of
finish
up.
The
pr.
A
B
Jim
has
done
all
the
work.
I
I've
done
some
talking
through
things
on
twitter
and
stuff.
I
I
need
to
review
the
vr
still
I'll
do
that
today.
C
There
shouldn't
be
any
incompatibilities,
I
I
tested
it
pretty.
Well,
it's
just
the
the
only
tag
that
doesn't
exist
or
carried
over
or
the
the
slim
tags.
B
Okay,
can
you
add
the
the
docker
head
repository
you're
pushing
to
just
so
we
can
review
that.
C
A
And
now
will
this:
will
this
alter?
How
we
do
how
we
do
our
actual
our
production
builds?
I
assume
it
will.
Yes,
yes,
it
will
and-
and
does
it
I
assume
it
continues.
The
the
push
to
jenkins
for
eval
of
the
s390x
image,
the
powerpc
image.
Is
it
also
pushing
an
arm
image.
C
Yes,
these
so
the
scripts,
you
can
run
images
and
tags,
the
publish
images
publish
tags
on
well,
they
need
to
be
run
on
arch
to
generate
the
s390
arm
or
power
images.
The
manifests
can
be
run
on
any
arch,
actually
constructing
the
manifest,
because
for
manifest
you're
just
pulling
down
all
the
images
they
don't
actually
need
to
be
ran
or
anything
like
that.
I
think
technically,
tagging
you
don't
need
to
run
on
a
different
arch
too,
but
yeah.
A
C
No,
no,
those
those
are
the
four
I
target
and
just
to
note,
I
did
not
touch
just
kind
of
like
have
the
build
scratcher
before
I
did
not
touch
the
windows
images.
I
kind
of
left
those
separate.
A
Okay,
but
but
that
is
currently
separate,
okay-
and
that's
a
good
one
for
this
for
gareth
to
just
report
for
us
on
how
his
progress
is
going
there,
I'm
not
sure
what
it
would
mean
yeah,
because
don't
I
do.
I
remember
correctly
that
the
windows
images
are
generated
from
powershell
scripts,
not
really
from
make
files.
B
Correct
so
the
windows,
I
believe,
already
matches
the
tag
that
jim
is
proposing
for
the
site
images.
So
I
think
we're
okay
in
terms
of
the
naming
and
the
tagging
there,
but
it
could
use
a.
C
Review
another
feature
I
did
add
in
that
might
be
of
interest
to
you
guys
from
seeing
the
email
chains
on
docker
hubs,
new
limitations
and
stuff.
Like
that,
I
know
I
think
you
guys
got
sponsoring
right
for
for
docker
hub.
Yes,
it's.
E
C
Okay,
sweet.
The
one
thing
I
did
add
was
there
is
environment.
Variable
I
managed
on
this
pr
you
could
you
can
actually
override
changing
the
the
image
repository
you
guys
pushed
to.
So
if
you
ever
want
to
start
publishing
for
multiple
repositories
like
cue,
o
or
github's
docker
repository
that
is
possible
with
the
scripts.
A
A
E
A
So
so
that
would
conceptually,
then
that
would
allow
custom
war
packager
to
be
used
to
build
an
s390x
or
an
arm-based
custom
war,
packaged
for
a
for
for
that
specific
architecture
so
allows
a
user
to
combine
controller
and
plug-ins
into
a
custom
war
that
is
platform.
E
A
C
E
E
We
had
quite
productive
discussions
with
opportunity,
k
jigsaw
team
at
that
moment
to
be
basically
able
to
support
java
10
and
java
11.,
but
we
haven't
been
really
active
with
regards
to
supporting
higher
java
versions,
and
this
topic
might
become
actual
when
the
new
java
lcs
is
released
so
far.
The
plan
is
it's
going
to
be
java,
17.,
so
half
a
year
from
now,
we
should
be
able
to
test
the
upcoming
lcs.
E
E
So
yeah
does
it
make
sense
to
support
java
17,
probably
yes,
but
it
means
that
we
would
be
supporting
three
lcs
baselines.
You
would
need
someone
to
actually
add
support
and
testing.
You
would
need
to
create
a
ci
cd
pipelines,
increase
cost
of
all
testing
across
the
agency
organization
and
if
it
happens
like
with
java
11,
when
we
have
less
than
two
percent
of
users
running
this
version,
I'm
not
sure.
What's
at
the
point.
A
E
E
So
there
is
a
reason
to
update,
for
example,
we
run
searching
say
on
java
11
and
if
you're
using
newer
versions,
we
would
have
even
better
garbage
collector.
We
would
have
even
better
diagnostics
tools,
so
in
principle
it
makes
sense
to
update,
especially
if
you
run
your
own
service
and
when
the
cost
of
operating
this
service
is
something
important
to
you.
But
if
you
ship
on
premise
version
so
that
others
run
it
basically,
others
yeah.
E
E
And
yeah
this
10
percent,
I
yeah-
I
just
made
up
a
number
but
yeah
the
principle
that
while
there
is
considerable
improvement,
it's
still
not
a
point
for
instances
which
just
work
at
the
moment.
A
A
And-
and
that's
a
that's
a
good
point,
for
example:
we've
got.
We
can't
fundamentally
support
the
adopt
open,
jdk
jdk
8
on
system
390z,
because
it
doesn't
use
just
in
time
compilation
but
java
11
does
use
it
and,
and
is
great
it's
so
that
you've
got
a
good
point,
that
there
is
a
there's,
a
motivation
on
s390z
and
other
places
that
java
11
just
behaves.
B
A
Good
all
right
anything
else
I
like,
should
we
should
we
call
for
a
decision
here,
call
for
it,
because
I
think
I
think
we've
got
the
right
people
on
the
call
to
decide
if
we
should,
if
we
should
continue
being
part
of
the
quality
outreach
program
and
just
then
give
them
the
answer
accordingly,
would
would
that
be
okay
or
do
we
need
the
governance
board
to
to
get
involved
in
that?
I
don't
think
the
governance
board
was
involved
in
the
initial
joining
of
it.
E
E
The
problem
is
whether
somebody
wants
to
contribute.
Even
when
we
were
doing
java
10
java
11
support,
we
organized
a
hack
first,
we
also
used
other
events.
We
had
more
than
30
contributors
in
total,
but
at
the
same
time
a
significant
part
of
changes
came
from
cloudbees
at
that
point
and
yeah,
I'm
not
sure
whether
there
would
be
investment
in
java
17
support
at
the
moment,
I'm
not
a
product
management
manager.
I
cannot
say
but
yeah.
A
All
right
and
and
they've
expressed
their
interest,
but
but
again
the
interest
we
got
from
oracle
was
oracle
cloud,
not
the
oracle
java
team.
So
so
even
they
may
not
be
that
that
focused
on
java
17
support
good
okay.
E
So
yeah
at
some
point,
I
wanted
to
just
poke
recent
java
versions
to
see
what
are
the
obstacles,
but
I
can
confirm
that
jenkins
doesn't
just
work,
so
we
will
again
go
through
all
these
jboss
mastering
through
is
updates
with
many
potential
regressions,
because
we
use
awesome
pretty
much
everywhere
across
the
project,
so
yeah
in
principle
having
let's
say
a
hack
fest
like
we
did
in
2017
for
java
10
would
be
nice
just
to
crunch
and
see
what
are
the
issues
likely
like
for
java
11
many
things
will
work
out
of
the
box
after
some
patches,
but
still
it's.
E
A
Okay,
so
I
think
what
I'd
like
to
do
is
propose
just
a
poll
here
of
the
of
those
in
on
the
call
say:
should
we
remain
members
of
job
equality,
outreach
and
increase
attempt
to
increase
our
efforts,
and,
if
so,
who
or
should
we
step
away
from
job
account
quality
outreach
and
admit
that
we're
not
able
to
participate.
B
A
So,
plus
one
from
from
mark
to
step
away
and
and
to,
I
should
state
it
differently
to
officially
state
that
we
are
dropping,
because
I
don't
want.
I
don't
want
the
job
equality
outreach
people
to
think
that
we're
we're
somehow
going
to
step
up
and
do
more
or
but
rather
let's
just
let's
step
out
so
alex.
I
took
that
from
plus
one
from
you
as
well
oleg
your
your
feeling.
C
A
E
A
B
B
Yeah-
and
I
think
we
decided
not
to
do
it
for
stretch
just
because
we're
going
to
be
deprecating
stretch
soon,.
C
Right:
okay,
yeah!
I
was
only
actually
asked
about
that
because
the
end-of-life
stretch,
I
think
kind
of
came,
maybe
not
for
lds,
but
the
the
basic
versions
and
I
don't
think,
adopt
open,
jdk
or
adoptium
produces
for
stretch.
I
think
they
moved
on
to
buster
for
the
lts
support.
A
A
So
any
objections
to
to
that,
so
we
just
we
allow
the
stretch
image
to
retire
as
its
current
configuration
used
in
open
jdk
and
it's
currently
using
open,
jdk
8u
242
currently,
and
it
will
just
stay
there
until
we
obsolete
it
remains
until
stretch
is
obsolete
or
they
upgrade
jdk.
They
do
the
upgrade
themselves.
A
A
A
Debian
alex
any
any
concerns
there
or
places
where
ooh.
This
is
this.
This
is
a
bad
plan,
something
like
that.
No
sounds
good
to
me.
Okay,
all
right
so
retire,
and
so
I
just
to
be
sure
I
think
what
that
means
is
we
will.
The
the
label
latest
will
switch
at
some
point
in
the
future
from
stretch
based
to
buster.
A
C
Mark,
if,
if
I'm
not
wrong,
I
thought
the
the
latest
actually
points
to
buster
right
now.
I.
E
E
Them
when
we
were
cleaning
up
the
terminology
at
the
same
time
alex
submitted
a
pull
request
for
buster,
so
it
was
also
integrated
and
later
we
retrospectively,
they
had
that
stretch
back
but
yeah.
They
believe
that
we
immigrated
everything
to
buster
by
default.
A
Yeah,
I
I'm
reasonably
confident
that
the
controller
images
are
still
are
still
shipping
stretch,
but
I'll
have
to
double
check.
A
Interesting
because
at
least
I'm
I'm
quite
confident
that
my
image,
that's
derived
from
the
lts,
the
lts
latest,
so
2.249.3
lts
is
in
fact
based
on
stretch.
So
all
that
we
may
have
this
may
indicate
that
we've
already
solved
our
problem.
I
didn't
realize
that
I'll
have
to.
A
I
thought
that
when
I
ran
my
tests
I
had
we
had
quite.
We
had
several
operating
systems,
several
tags
that
are
still
using
the
stretch
image
but
I'll
double.
A
A
A
For
each
image,
what
I've
seen
is
that
I
don't
monitor
the
centos
images,
for
instance,
because
they
just
aren't
relevant
to
my
use
cases,
but
there
are
probably
people
who
do,
and
so
we
would
like.
I
think
we
need
the
concept
of
a
maintainer
for
each
image
and
I
believe
that
the
github
code
owners
concept
may
help
us
there.
I'm
open
to
the
comments,
though,
is
that.
A
With
clearly
communicating
who's
who
the
image
owners
are
or
the
image
maintainers
are.
E
A
Okay,
great
all
right,
submit
the
jab
right
and
submit
the
jab.
C
Yeah,
so
the
like,
I
mentioned
before
the
pr
has
been
submitted.
The
whole
point
of
this
is
to
add
a
couple:
more
verbose
tags.
The
idea
is,
you
publish
the
super
verbose
tags
which
include
the
jdk
jvm
os
name,
os
version
jenkins
version
yeah.
I
think
that's.
Actually
it's
and
you
publish
those
and
then
you
re-tag,
those
with
the
the
tags
you
guys
are
all
used
to,
like
slim,
debian,
latest
lts,
and
basically
that
way.
C
And
the
the
new
proposed
scripts
do
all
that
plus
keep
all
the
tags
except
the
one
I
mentioned
slim,
which
that's
an
easy
fix.
If
that
is
needed,.
C
I
don't
really
use
the
agents
that
much
so
I
probably
can't
speak
for
that.
But
my
assumption
is
the
agents
probably
don't
need
it
as
much.
Unless
there
is
a
very
big
case
where
you
know
the
agent,
you
know
jdta,
jvm
and
os
actually
matter
a
lot.
A
Yeah
oleg
you've
got
lots
of
experience
with
the
agent
images.
What's
your
sense
there
in
terms
of
adoption
or
of
of
operating
system
preferences,
do.
E
A
Great
okay
well-
and
I
like
that-
that's
that
has
we
can.
We
could
certainly
use
the
controller
naming
pattern
if
someone
says
hey,
I
need
this
support
for
this
type
of
agent
great,
we
can
use
the
we
can
at
least
use
the
naming
pattern
that
jim's
created
if
we
needed
to
be
more
verbose
in
the
agent.
C
C
Okay,
because
the
if
and
when
you
guys
did
want
to
do
the
verbose
naming
the
scripts
are
written
in
a
way
where
it
scans
those
folders.
So
as
long
as
that,
like
folder
structure
is
the
same
on
the
controller
or
on
the
agent's
repos,
the
scripts
should
be
pretty
easy,
actually
just
to
carry
over
a
slight
modification,
not
a
whole
rewrite
or.
A
Great
excellent,
thank
you
very
much
for
your
work
on
that
jim.
Thank
you.
Thank
you.
Anything
else
on
simplified
labeling
of
docker
images.
No
okay,
next
topic,
gareth
windows,
docker
images
for
ltsc,
2019,.
F
Sure
yeah
not
too
much
to
update
since
last
week
on
this,
so
the
packer
builds
for
2019
ltsc
will
see
through
work
nicely.
F
All
those
images
have
been
published
and
all
of
those
jenkins
instances
are
updated
and
running
with
the
updated
images
they
have
some
tests
in
them
to
validate
docker
and
the
disk
size
is
correct,
which
is
some
common
failures
that
we
seem
to
have
on
those
which
is,
I
think,
it's
quite
good
and
windows
updates
are
installed
as
part
of
the
build.
F
F
So
that
seems
to
be
they
look
like
they're
going
in
nicely
and
then
there's
a
pr
for
jenkins
to
add
2019
ltsc
as
a
sort
of
docker
variant.
For
that,
that's
where
I'm
up
to!
I
think
that's
the
I
think.
That's
the
last
pr,
but
it's
it's
yeah.
It's
currently
timing
out
and
validating
stuff.
A
A
B
No,
I
know
they
have
a
problem
with
build
machines
like
they.
They
don't
have
a
lot
of
infrastructure
for
building
windows
images,
so
I'm
I'm
not
sure
how
things
how
that
will
get
resolved.
Okay,.
F
A
A
F
So
the
next
step
is
there's
a
jenkins
co
docker
pull
request
which
I'll
I'll
pop
in
the
chat.
F
F
Stuff:
okay:
I've
tried
extending
the
timeout,
but
it
doesn't
have
any
effect
for
me.
When
I
push
the
change
up,
I
think
I
think
that's
a
permissions
piece.
A
And
alex
any
objections,
if
we
were
to
extend
the
timeout
for
or
is
that,
is
that
a
timeout
that
is
system
wide,
no
job
on
on
ci.jenkins?
That
io
is
to
be
allowed
more
than
60
minutes.
B
The
windows
builds
like
we
have
for
the
linux
builds,
but
I
just
haven't
had
the
time
to
look
at
it:
okay,
great
all
right,
because
we
actually
do
build
we're,
building
eight
images
and
testing
them,
so
it
I
think
it's
eight
that
may
be
wrong
on
that.
So
it
does
take
some
time.
So
if
we
could
parallelize
that
it
would
be
good.