►
From YouTube: Distribution Group Conversation
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
Perfect
I'll
continue,
then
I,
I,
I'm,
DJ,
mountainy,
I'm,
a
senior
distribution
engineer
on
the
distribution
team
and
currently
I'm
taking
over
for
Martin
as
interim
engineering
manager
on
the
team
and
jumped
yeah.
So
here's
the
agenda
for
the
slides
today
I'll
go
ahead
and
jump
right
in
see,
as
I
mentioned
I'm
temporarily
in
as
the
interim
distribution
manager.
Martin
was
the
manager
here,
as
well
as
being
the
manager
on
the
delivery
team.
So
for
the
moment,
he's
gone
ahead
and
focused
on
that
team
and
I'm
stepping
in
to
take
over
the
direct
reports.
A
For
now
until
we
hire
some
more
permanent
into
this
position
in
terms
of
changes
in
the
team.
In
addition
to
that,
at
the
end
of
January,
we
had
a
team
member
leave
the
team,
so
our
team
size
shrunk
by
one,
and
so
we
are
now
actively
hiring
for,
like
I
mentioned
the
Distribution
Manager
position,
as
well
as
a
product
manager
for
the
team
to
take
over
from
Joshua
who's,
an
interim
role
there
and
we're
also
looking
to
hire
another
one
too
cloud
native
engineers.
A
In
terms
of
other
team
process
changes
over
the
last
two
months
at
the
beginning
of
January,
we
introduced
a
new
process
in
the
team
related
to
the
project
containers
we
had
so
the
maintainer
x'
and
the
entire
team
get
together
once
a
week
for
this
summary
meeting
in
terms
of
what
has
gone
into
the
product.
So
this
was
in
order
to
keep
the
maintains
aware
of
the
changes
going
into
the
product.
As
now,
we
have
multiple
on
the
Baskerville,
once
we've
moved
to
having
a
couple
different
maintained
errs
merging
separate
from
each
other.
A
A
A
So
over
the
past
two
months,
we've
had
two
notable
changes
come
out
of
the
results
of
this
both
of
these
happen
to
be
in
the
omnibus
project,
but
we
recently
had
an
open
work
day,
change
that
fixed
some
of
our
some
technical
debt
and
return
in
regards
to
our
tests.
So
several
months
ago
we
had
changed
our
our
resources.
We
use
in
omnibus
to
use
custom
resources
or
a
newer
version
of
chefs
custom
resources,
and
we
hadn't
gone
through
and
updated
our
tests
to
make
use
of
the
good
features
that
came
with
that.
A
So
it
last
a
while
we've
added
20
to
30
dependencies,
so
I
won't
that
their
sin
request
adding
those,
but
we
also
had
a
fairly
large
update
to
the
running
cookbook.
We
used
to
run
a
hist
in
its
system
and
system
processor
that
we
use
in
omnibus
we
brought
in
about
three
years
of
upstream
changes
on
that
change.
It
was
quite
a
big
change,
including
adding
some
niceties
to
testing,
but
also
to
process
restarts
and
also
additional
process,
signalling
that
we
were
looking
forward
to
we've.
Also
in
the
latest
release.
A
11.8
we've
updated
the
container
registry
to
the
latest
version,
bringing
in
so
in
the
previous
update.
We've
done
this
in
the
previous
release.
We've
done
a
step
update
towards
this,
so
the
two
combined
have
brought
in
about
two
years
of
docker
distribution
changes
to
the
registry
to
bring
us
more
features
and
more.
We
can
do
with
the
registry,
including
things
like
monitoring,
which
will
be
looking
into
soon.
A
I've,
looked
at
a
couple,
other
omnibus
editions
here
that
went
in
go
on
to
the
gitlab
home
chart
accomplishments,
so
we've
had
quite
a
few
documentation
improvements
and
help
from
the
documentation
team
in
access.
Specifically,
we
also
now
have
a
new
location
for
the
docs.
So
right
now
it's
findable
under
a
feature
flag.
But
if
you
go
to
the
slash
charts
directory
on
Docs,
you
can
see
the
documentation
there
in
terms
of
the
home
charts.
The
last
two
releases
have
been
largely
bug,
fixes
and
documentation.
A
Basically,
what
we've
done
is
that
we,
the
operator,
is
in
a
experimental
state
where
you
can
optionally
enable
it,
and
we
have
now
done
that
in
our
CI
and
we're
working
on
fixing
the
bugs
that
we
found
as
a
result
of
that,
but
one
of
them
being
clean
up
when
things
go
wrong
of
the
job
just
creating.
So
the
the
feature
we're
iterating
on.
A
There
is
no
downtime
updates,
so
we
have
the
the
featuring
kind
of
an
experimental
states
and
we're
working
on
knocking
down
the
bug
spiel
to
enable
that
by
default,
so
we've
made
some
progress.
There.
We've
in
test
increase
the
test
coverage
on
the
operator
on
a
separate
issue
with
in
making
city
progress
on
refreshing,
our
gke
marketplace,
image
where
the
charts
are
deeply
deployed
from
the
cheeky
marketplace,
and
one
of
the
other
things
I
want
to
call
out
is
community
engagement
on
this
project.
C
Go
for
it.
The
goal
of
this
whole
presentation
is
to
for
the
interaction
your
your
reading,
your
slides.
We
can
read
your
slides
way.
Faster
I
read
their
slides
in
the
first
five
minutes
of
the
presentation,
so
we
wanna
the
maximum
amount
of
presentation
is
10
minutes
and
if
you
do
know
five
sativum
better
that
the
idea
is
to
keep
this
interactive.
C
Okay,
I'll
jump
onto
the
next
slide,
no
you're
you're
over
your
over
the
10
minutes
already.
So
let's
go
to
Q&A
and
I'm.
Actually
thinking
about
not
having
presentation
at
all,
we
have
the
same
thing
with
the
board
meeting.
So
it's
not
new,
it's
very
normal
that
if
you
have
slides
you
start
reading
them
and
what
we
did
in
our
board
meeting.
C
We
have
no
presentations,
there's
a
slide
deck
of
140
slides
and
you
can
ask
question
to
use
that
like
I
have
a
question
about
slide
number
X,
Y
or
Z-
and
we
talked
about
that-
and
this
is
just
like
that
for
first
person
chat.
The
first
group
conversation
this
week
and
therefore
want
to
go
over
time.
So
I
think
it's
just
very
human
nature
to
to
when
you
present
sliced,
you
start
discussing
what's
on
them
and
we're
probably
gonna
change
that
okay.
D
I
love
the
work
on
anything
days.
This
was
a
you
know,
it's
a
huge
thing
to
like
not
improve
like
various
things
that
are
just
like
bugging
people,
and
it
makes
everybody
feel
good.
I
think
you
should
shout
out
to
it
like
specifically
who's
who's
working
on
each
of
the
different
things
and
then,
when
you're,
instead
of
doing
when
you're
doing
the
slides
or,
if
you're
not
doing
the
slides
like
just
get
the
shout
out
to.
Actually,
though
people
are
doing
it.
A
Yeah,
that
sounds
like
a
good
idea,
so
in
terms
of
in
terms
of
these
specific
two
items
so
on
our
team.
Palestine
car
was
working
on
the
the
first
item
and
it
actually
includes
did
quite
a
few
different
merger,
quests.
So
I
think
he
did
between
four
and
six
merger
lists
to
get
that
closed,
and
then
he
in
bomb
on
our
team,
was
the
one
who
did
the
changelog
scripts.
E
A
Yeah
I
think
I
think
that
the
big
thing
is
having
it
documented
so
that
that
the
team
members
feel
comfortable
doing
it
I.
Think
generally,
we
accept
the
idea
or
like
the
idea
of
people,
seeing
issues
or
seeing
things
they
want
to
work
on
in
other
areas
of
the
product
and
reaching
out
in
doing
that.
But
when
you
have
a
lot
of
other
stuff
scheduled
on
your
plate,
it's
not
necessarily
easy
to
know
that
you
can
do
that
or
to
feel
comfortable
doing
it.
F
A
DJ
your
question,
but
like
you're
doing
a
lot
of
omnibus
stuff
and
a
lot
of
cloud
native
chart
stuff.
How
do
you
in
the
team
divide
that
up?
Is
it
a
lot
of
commonality
or
they're
completely
separate,
and
when,
if
ever,
do
you
want
to
look
at
sort
of
splitting
distribution
up
into
omnibus
and
cloud
native?
If
that
makes
sense,
so.
A
G
In
an
answer
mark,
if
you
don't
mind
well,
BG
is
returning
yeah,
please.
So
we've
observed
that
we
have
a
number
of
items
that
are
actually
overlapping
and
mostly
most
of
them
are
related
to
actual
supporting
of
the
features
that
we
ship.
So
any
libraries
any
dependencies
that
we
have
between
the
two
projects
are
going
to
be
the
same
between
the
cloud
native
charts
and
also
the
omnibus
charts
sorry
Abdullah's
package.
G
We
also
seen
that
whatever
you
kind
of
learn
packaging
wise
in
one
of
those
you're
going
to
most
likely
turn
it
into
you
know
usable
thing
in
the
other,
so
while
the
two
are
not
strictly
overlapping,
the
the
knowledge
that
is
gathered
in
one
and
the
other
can
be
shared
between
the
two
and
we
found
that
automating.
Some
of
these
things
helps
us
and
the
knowledge
just
kind
of
stays
within
between,
without
the
need
to
have
like
all
these
syncing
issues.
Where
you
know,
oh
now,
we
added
the
support
for
this
feature.
F
G
H
Yeah
I
have
a
question
I
put
in
the
chopped
above,
but
it's
kind
of
curious
that
as
Dalton
a
bus
package
grows
in
size.
It's
arty
up
to
like
500.
Something
bags
have.
Are
there
any
discussions
about
having
like
a
minimal
install
where
they
say
the
core
functionality
and
then
download
a
different
kind
of
kind
of
similar
to
how
Linux
does
in
the
mid
and
fall.
A
Yeah,
so
we
do
have
an
issue
open.
We
have
had
issues
open
to
explore
that
sort
of
thing,
and
it
has
been
something
that
we've
considered
at
the
moment.
We
also
have
issues
open
specific
task
items
to
further
reduce
some
of
the
install
size.
We've
done
quite
a
bit
of
that
in
the
past.
Over
the
past
year,
tried
as
new
as
new
stuff
has
come
in.
We've
tried
to
maintain
the
same,
a
similar
package
size,
so
we've
had
to
go
through
and
be
more
strict
about,
ripping
stuff
out.
A
We
do
have
a
couple
pending
issues
still
remaining
to
get
the
last
of
what
we
had
seen
as
actionable
based
on
initial
evaluation
knocked
down.
It
has
been
something
that
we've
we've
looked
at,
but
I
don't
know
that
we've
come
to
any
conclusions
on
whether
we
want
to
do
that.
I
know
at
the
moment
it's
it's
kind
of
nice
having
it
as
one
package,
both
in
terms
of.
H
I'm
not
either
but
they're,
just
just
you
could
solve
that
big
big
Powerball
right,
but
then
they're
like
yeah,
the
full
DVD
versus
the
net
install
this
was
especially
when
you
start
looking.
Other
components
will
be
coming
like
shared
secrets,
for
example,
there's
some
other
stuff,
that's
coming
in
in
the
next
few
months.
At
all,
that'll
add
to
that
tar
ball
size.
G
Now
that
needs
to
behave
in
a
certain
way
compared
to
the
the
bigger
item.
So,
yes,
I,
think
ideal
would
be
if
we
could
keep
this
under
control.
You
know
like
keep
the
size
under
control,
keep
the
simplicity
under
control
as
long
as
we
can
so
that
we
don't
end
up
doing
this
plate,
because
those
plates
are
actually
costly,
maintenance,
wise.
B
Elaborating
there
shouldn't
I
say
that
if
you,
you
know
Kelly,
if
you
do
our
things
or
have
customers
like
paint,
like
particular
pain
points
like
if
you're
more
like,
like
the
more
rationale
to
try
and
do
this
and
the
more
like
hard
data
points
around.
That
would
be
great
I'm.
Just
because
it
is
costly,
it
would
be
a
fork.
It
would
be
sort
of
not
complete,
install
if
you
will,
because
most
of
our
set
does
default
on
and
so
yeah.
B
F
We
definitely
don't
want
to
promote
sort
of
you
know
using
a
scaled-down
version.
We
definitely
do
need
to
realize
this
competitive
pressure.
They're,
like
kitty,
is
one
of
these
things
where
they're
getting
a
lot
of
traction,
because
it's
a
smaller
you
know
more
memory,
efficient
kind
of
thing
and
so
having
a
slimmed
out,
like
obviously
being
able
to
disable.
F
Various
portions
is
really
important
to
be
able
to
keep
memory
down,
but
correct
me
if
I'm
wrong
in
terms
of
like
the
bundle
size,
if
you
do
disable
it
using
you
know
the
collab
Darby
or
whatever
at
least
it
doesn't
take
a
memory
like
ram
memory.
It
takes
up
disk
space
and
it
means
it's
slower
to
if
you're
slogging
it
around
or
whatever
backing
up
or
something
like
that
it
takes
longer,
but
it
shouldn't
have
any
impact
on
sort
of
ya.
Runtime
Ram
right,
not.
G
C
G
C
Think
we
should
focus
on
memory
like
there's
lots
of
things.
We
can
focus
on
the
thing
that
people
pay
for
is
memory.
If
you
go
to
digitalocean
you're
going
to
be
memory
constraint
on
this
space
constraint,
so
it's
much
more
important
that
we
have
a
way
to
if
there's
a
way
to
install
gate
lab
and
get
kind
of
gitche
equal
features
are
just
making
a
single
change
in
the
leftover
b
or
the
parameter
during
installations.
That
will
be
super
helpful
and
I.
C
Think
the
memory
I
know
memory
is
very
visible
because
your
distribution
team
is
sugar.
Packets
and
people
complain
about
that
too.
But
the
thing
that's
hurting
us
in
the
marketplace
is
memory,
memory
and
memory,
yep
memory
and
CPU
time,
but
you
know
we
have
we
opportune.
Oh
no
people
don't
complain
about
CPU
time,
it's
memory,
memory
and
memory,
and
that
is
what
people
are
paying
for:
we're
we're
losing
the
people
that
just
run
a
single
instance,
and
they
have
very
few
users.
It's
it's
all
about.
C
B
Nothing
I
know
we
have
some
global
switches
like
you
can
turn
off
all
the
Prometheus
monitoring
stuff,
with
like
one
single
flag,
for
example,
which
I've
done
all
the
exporters
for
me
Tisa
itself,
and
things
like
that.
So
we
have
some
of
those
tools,
but
we
don't
have
one
global
switch
to
sort
of
make
it
like
the
minimal
bear.
Get
web
install
and
I'll
open
an
issue.
Now
we
can
recon
track
that,
but.
G
D
The
by
far
the
biggest
memory
consumer,
actually,
one
of
the
things
I've
always
thought
about
is,
is
how
can
we
trigger
some
of
these
switches
from
from
the
rails
interface?
So
can
you
know
a
lot
of
this?
Is
you
have
to
log
in
to
the
machine
with
us,
SSH
and
and
edit
a
config
file
and
reconfigure
like?
Maybe
we
should
really
be
thinking
about
having
a
little
background
trigger
that
looks
for
these
things
and
Auto
reconfigures
for
the
admin,
though
the.
G
Binary
inside
of
omnibus
that
would
allow
us
to
do
those
type
of
things,
and
we
could
consider
something
that
would
Auto
reconfigure
on
certain
events,
the
biggest
problem
and
the
biggest
time
seeing
there
would
be
the
the
separation,
the
Linux
user
separation.
So
what
a
user
that
is
running
a
service
can
actually
cheat
right.
Unicorn
is
running
get
user,
whether
it
can
actually
do
something
for
the
register
user.
That
could
cause
some
security
concerns,
but
that's
definitely
worth
investigating.
F
B
I'm
g
ke
no
on
any
other
ones,
which
is
why
we're
going
through
and
adding
new
KS.
Basically,
we
we,
you
know,
you
know,
I
we
you,
as
the
charts
deployed
the
container
images
for
the
latest
version
of
master
and
then
running
it
on
QA
against
it,
and
we
stood
a
mirror
that
functionality
across
the
different
cloud
providers,
and
so
worst,
we've
established
sort
of
like
guidelines
for
official
support.
B
One
of
them
is
a
permanent
cluster
which
we
can
execute
tests
in
and
make
sure
that
any
changes
we
do
wake
our
tests
across
the
full
list
of
officially
supported
instances,
as
opposed
to
like
the
language,
which
is
we
test
on
gke.
Others
should
work
type
thing,
so
we
can
get
rid
of
that
and
say
we
know
it
works
on
these
providers
will.
I
F
I
A
Yeah,
if
we
can,
we
can
hear
you
yeah
so
in
terms
of
I,
don't
know
that
we
have
a
document
tracking
it
at
all.
What
we
do
have
is
certainly
for
some
of
them.
We
have
the
team,
aware
of
them,
but
yeah
on
on
others,
particularly
things
like
open
soos,
for
example,
and
stuff,
like
that,
we
probably
don't
have
a
good
handle
on
on
looking
ahead
at
that
at
the
moments
that
be
something
that
would
probably
be
good
to
look
into.
I
I
Packages
and
like
what
the
turnaround
time
is
associated
with
that
relative
to
win
the
distribution
becomes
available
right.
I
know:
Stables
was
the
go-to
middle
this
year
for
Debian
I,
don't
understand
the
customer
base
yet
well
enough
to
say
that
that's
important
or
not
that
we
really
have
that
available
versus
as
we
we
support
it
in
a
a
two
or
three
month,
time
frame,
yeah.