►
From YouTube: Meshery Development Meeting (March 17th, 2021)
Description
Goodbye, Daylight Savings!
B
Okay
cool,
so
the
bug
which
was
which
we
are
currently
fighting
out
was
one
where
we
were
not
able
to
connect
properly
to
the
to
the
cluster
itself
right,
so
we
were
getting.
We
were
not
getting
the
correct
end
point
then
we
needed
to
connect.
So
that
was
the
first
thing
that
was
fixed.
It
was
fixed
in
pr
one
and
learned
f5,
which
you
changed:
change
the
service
to
load
balancer.
That
was
helping
fingers
of
it
in
case
of
mini
cube.
B
A
B
The
changes
over
here
now
like
we
need
to
make
small,
smaller
changes
in
which
they,
the
adapters,
need
to
use
the
libraries,
the
new
library
which
will
make
use
of
this
particular
version.
This
capability
mesh
it
right
so
yeah.
Basically,
yes,
but
we
need
to
update
the
machine
adapter
library
for
each
of
the
adapters.
A
When
you
well,
you
clarified
what
I'd
said
saying:
well,
not
just
for
services
but
for
adapters,
and
so
I
guess
I
was
just
reinforcing
you're
saying
that
this
utility
will
help
people
or
when
measuring
spins
up.
I
I
said
when
measuring
spins
up
a
new
service,
this
utility
will
help
figure
out,
what's
the
most
appropriate
networking
to
assign
to
that
service
and
you'd
said
yes,
but
for
adapters.
B
A
A
A
B
So
this
utility,
what
he
is
talking
about,
is
a
different
one.
I
guess
he's
talking
about
the
gate
as
client
utility,
I
think,
is
that
that's
the
one
right
yep
yeah,
the
client
utility
is
basically
nothing
but
it
instantiates
a
number
of
kubernetes-based
clients
like
like
the
kubernetes
client
itself,
discovery,
client
and
all
the
different
clients
that
are
needed
for
doing
commodity
specific
operations.
B
Yeah,
so
that
was
the
first
thing
that
was
tackled
and
second
was
the
percentage
which
we
are
getting
was
not
entirely
accurate.
So
now
we
are
getting
the
accurate
percentages.
So
let
me
just
just
pull
up
that
yeah.
Here
we
go
so
the
new
way
of
counting
the
percentage
has
been
implemented
where
you'll
probably
get
person
with
accuracy
of
two
and
yeah.
That's
it.
B
B
Sorry
about
this,
my
desktop
is
amazing,
yeah,
so
fun
fact,
good
question.
I
just
collaborated
on
this.
He
was
working
on
something,
so
we
did
live
share
together
and
yeah.
We
now
have
a
public
endpoint
called
smi
results
of
public,
in
which,
basically,
currently
it
is
my
id
in
which
all
the
results
which
are
gone
through.
My
particular
id
will
be
shown
over
here
publicly
to
anyone
which
will
basically
be
used
to
show
official
results
in
our
in
our
near
five
website
to
give
people
context.
B
Nerf,
I
o
has
a
pretty
table
in
which
we
show
the
in
which
we
show
the
results
of
the
conformances,
which
have
been
involved
right
over
here.
So
currently,
the
data
which
is
over
here
is
a
static
data
means
we
had
run
test
locally
and
the
results
which
we
have
gotten
were
added
over
here.
But
the
thing
is
now:
we
have
a
particular
endpoint
and
whenever
a
new
test
is
invoked
and
the
latest
changes
will
now
be
added
over
here
now,
the
functionality
in
the
website
is
yet
to
be
implemented.
B
B
A
Nice
yeah
and
you've
on
that
one
of
the
more
like
there's
a
couple
of
ways
that
this
could
be
done
and
one
of
the
more
likely
ways
is
that
is
that
the
measuring
provider
continues
to
be
used,
but
that
individual
one
account
per
service
mesh
is
identified
as
being
authoritative
for
that
service
mesh's
status
and
that
one
two
three
four
five
five
accounts
would
be
statically
configured
here
or
well
here
or
in
the
endpoint
yeah.
As
okay.
B
A
B
A
B
B
B
B
So
here
we
are
in
traffic
access.
For
so
here
is
where
you
are
not
working
properly,
and
if
you
want
to
dig
more,
you
can
check
out
our
dog
in
which
we
have
defined
in
a
better
detail.
What
each
particular
test
does
and
still,
if
they
want
to
dig
down,
then
we
can
point
them
to
the
learn
layer.
If
I
recover
where
the
tests
are
generally.
A
And
then
yep,
if
they
want
to
see
the
I
mean
one
of
the
issues,
one
of
the
the
things
about
this
is
that
they
may
well
understand
that.
Okay,
that's
that's
the
test!
That's
that
test!
Okay,
I
get
it,
but
I
thought
that
we
had
this
taken
care
of.
You
know
whatever
that
is
and
they'll
say
well,
so
why
isn't
it
working
then?
Because
I
think
I
thought
we
had
that
so
where's,
the
debug
log
or
like
where's,
the
what
specifically
is
failing.
B
B
A
A
We
need
to
be
able
to
tell
people
what
failed
like
yeah
failure
in
that
step.
That's
excellent!
That's
beautiful!
That's
the
right
thing
to
show
in
a
concise
way
here,
but
for
them
to
go
fix
it
and
for
them
to
like
it's
okay.
If
the
answer
is
well
hey,
you
need
to
go
ahead
and
go
look
at
the
logs
in
the
learn
layer,
5,
app
and
or
or
wherever.
Those
logs
are,
if,
if
we've
yet
to
expose
them
here
that
that's
okay
but
you've
got
to
anticipate
that.
A
B
A
A
A
Ideally,
the
second,
but
we
don't
need
to
your
point
about
like
getting
off
of
kudo.
We
don't
or
yeah
cuddle.
We
don't
need
to.
It
can
be
the
former.
It
can
just
be
like
you
know
here
here
here,
here's
the
log
gear
and
tell
whoever
is
working
on
this.
They
are
absolutely
intelligent
enough
to,
like
you
know,
step
through
the
whole
stack
trace,
so
to
speak,
the
whole
log
and
you
know,
figured
and
that's
fine,
and
now
it's
not
the
best
experience
and
that's
fine,
because
that's
entirely
appropriate
that
you
we
wouldn't
want.
A
You
wouldn't
want
to
invest
so
much
more
into
this.
It's
just
like
enough
to
get
us
by
so
that
we
can
then
have
convinced
everyone.
This
is
real.
We
we
in
fact
are
we've
been
talking
about
this
for
a
year
and
a
half
since
we
opened
the
issue
not
not
with
you
drew,
but
just
and
but
it's
serious,
we
we
were
serious.
We
are
serious.
A
This
works,
this
is,
is
here
you're
unable
to
do
it,
yeah
there's
going
to
be
some
rough
edges
and
things,
and
now
that
was
the
first
cut
and
that
architecture
you
know
drew
is
identifying
as
not
great
and
we've
actually
rewritten
that
stuff
we've
already
rewritten
it
once,
but
we
do
need
to
rewrite
it.
A
second
time
and
it'll
be
really
good.
The
second
time
in
order
to
guarantee
that
it'll
be
worth
it.
We
need
to
have
more
than
one
of
those
teams
beating
on
our
door.
Saying
hey,
we
need
this.
A
Our
customers
are
asking
us.
Are
we
compliant
also
that
that's
pie
in
our
face
when
you're
saying
it's
only
50
and
we're
the
ones
that
made
this
spec
and
so
they're
yeah?
And
it's
also
okay?
If
it's
like
hey
look,
it's
gonna
be
really
hard
to
get
that
that
log
exposed
here,
even
just
the
entire
log
like
here,
it's
okay.
If
the
answer
is
yep,
we
know
that
there's
an
open
issue.
What
you
know
you
can't?
No,
you
can't
just
let
it
run
and
come
back
that
doesn't
that's
the
tool.
A
Isn't
that
great
right
now,
here's
what
you've
got
to
do
to
or
or
here's
something
else
you
can
do
to
persist
those
logs
so
make
sure
that
you've
installed
fluentd
in
your
cluster
that
you're
persisting
the
logs
off
elsewhere
so
that
when
it
dies,
you
can
come
back
and
look
like
what
just
needs
some
answers
and
some
to
have
look
like
we've
done
our
diligence
to
say
this
is
how
the
tool
works.
This
is
how
you
sign
up
hey.
I
I
I
work
on
osm.
How
do
I,
how
do
I
get
osm's?
A
How
do
I
register
and
get
os?
How
should
I
use
this
in
our
pipeline,
and
how
do
I
schedule
this
so
so
I
grab
the
token
then.
How
long
does
that
token
last
for
me
to
then
send
my
results,
or
do
I
even
need
a
token
or
like
so
there's
just
a
there's,
a
lot
of
like
you
know,
rough
edges
that
we
need
to
tie
off
so
that
someone
can.
A
They
can
really
come
use
this,
because
when
I
don't,
we
don't
want
to
go,
tell
them
here,
use
this
and
then
they
they
ask
us.
The
first
question
like
oh:
why
did
it
fail
and
you're
like
well
geez?
I
don't
know,
let
me
think
about
that.
I
don't
oh,
I
guess.
We've
got
some
enhancements
to
do
and
we
haven't
even
written
that
down
or
you
know
so
so
whatever
that
answer
is,
please
go,
create
an
issue
and
and
identify
like
what
they
can
do
today.
C
A
And
this
is
good
by
the
way.
I
I
none
of
the
stuff
that
you've
done
well,
none
of
the
things
that
I
just
said
are
in
like.
Please
don't
interpret
those
negatively
the
you're
doing
you're
doing
well,
there's
a
lot
to
consider
here,
a
ton
of
moving
parts.
It's
it's
a
whole
function.
It's
it's
something
that
exercises
all
of
what
a
specification
is
against
eight
different
servicemen.
You
know
five
different
surface
meshes,
it's
a
ton
and
it
captures
them
and
like
it's
a
lot
to
have
considered
so
it
looks
really
good.
A
The
ui
here
looks
really
good
and
yet
those,
hopefully
people,
bang
on
our
door
about
those
those
next
couple
of
things,
I'm
like
hey.
Why
am
I
failing?
I
think
this
is
important.
How
do
I
fix
that?
Also,
I'm
I've
begun
to
run
this,
and
I've
got
results
that
I'd
like
I
want
to
go
fix.
I'm
we're
working
on
fixing
those
I'm
ready
for
my
results
to
be
published.
How
do
I
work
with
you
to
make
sure
that
my
results
are
published.
A
B
B
Okay,
so
the
update
one
will
be
the
layer
5,
the
new
conformance
that
they
stack
like
the
docker
image
has
been
published
and
updated.
The
adapter
library
is
still
yet
to
be
merged.
It
is
working
all
fine,
it
is
up
for
review
and
we
need
to
update
the
once
the
adapter
libraries
are
built.
We
need
to
update
each
of
the
adapters
to
use
some
adapter.
B
A
No,
that
makes
sense.
That
would
be
like.
Oh
a
one-time
thing.
Eight,
you
know
ten
eight
different
times,
but
like
yeah,
very
good.
According
to
my
watch,
we
have
14
and
a
half
hours
before
this
turns
into
a
pumpkin
without
having
those
other
those
items
that
you
just
mentioned
completed
so
we're.
I
think
it's
feasible
right.
I
think
we're
you
can
get
those
merged
today.
Maybe.
A
Nice
sorry,
everyone
else.
I
forgot
to
ask
for
feedback
from
others,
but
but
drew's
been
hot
hot
on
the
trail
here,
so
drew
one
other
item
as
you
go
to
wrap
it
up
is:
please
do
test
from
a
user's
perspective
so
because
I
assume
that
you're
you've
compiled
mescheri
locally
and
aren't
running
measurements.
A
A
There
is
a
new
edge
release,
that's
made,
and
so
you
can.
You
can
subscribe
to
the
edge
channel
the
edge
release
channel
and
you
know
test
your
changes.
Post
merge,
you
know
10
10,
15
minutes
or
so
after
it's
been
merged,
but
before
it's
been
merged,
if
you
want,
if
you're
a
reviewer
of
a
pr
or
if
you
as
the
individual,
that
has
made
the
changes
and
has
opened
a
pr,
you
know
that
there's
an
intermediate
build
that
happens
when
a
pr
is
opened.
A
A
A
A
B
A
Nice
cool,
yep,
you're,
then
you're
on
it
or
then
you're
aware
of
it.
I
guess
yep
it's.
What
is
it
measuring
ctl
measure,
ctl,
exp
in
experimental
mesh
validate
and
then
and
then,
who
knows
if
it
works
from
there?
One
of
the
issues,
I
I
shouldn't
say
who
knows
if
it
works
from
there?
That's
no!
I
don't.
I
don't
mean
to
say
that
with
respect
to
any
root,
I
I
mean
it
was
done
a
long
time
ago
and
he
did
it
quickly
and
and
so
a
lot
has
changed.
A
A
I'd
be
curious
as
to
what
that's
currently
set
to
or
drew
this
might
this,
or
I
shouldn't
say
orbit
and
also
as
individual
service
mesh
teams
come
to
say,
to
register
their
account
and
say
hey.
This
is
the
account
that
I
want
to
use
in
the
mesry
provider
to
send
in
our
official
results
that
you
know
part
of
that
that
hard
coding
of
their
you
know
that
that
registration
process
might
also
be
and
utkarsh.
If
you
give
this
some
consideration.
A
A
Wait
what
what
what
did
you
crash
did?
Who
did
you
just
say
I
don't
know
hold
on
yeah
hold
on?
Let
me
all
right.
I
think
I
think
I
think
we
have
I'm
gonna.
Take
a
note
here.
What
is
today
the
17th
march
17th
2021
rude
car
says
I
don't
know
for
the
first
time:
okay,
okay,
I
just
want
to
get
that
on
paper.
All
right,
okay,
anyway
drew
thank
you
man.
I
thank
drew
you
hit.
A
D
D
A
D
From
the
last
penalty,
like
last
time,
so
the
calendar
is
moved
down,
we
have
cards
above,
you
can
go
and
manage
your
profiles
and
you
get
to
see
the
total
results
test
results
that
are
under
this
profile.
You
can
turn
it
around
by
clicking
on
the
clock
card.
That's
confusing,
you
can
click
on
the
details.
D
D
D
And
this
is
the
big
review
I
mean
daily
view
of
the
test
started
run
because
all
of
them
may
like
all
of
them.
All
of
them
couldn't
have
been
shown
here
because
they
were
like
six
days.
So
you
can
click
on
this
says
that
would
run
in
one
single
day
at
what
time
and
you
can
click
the
run
test
and
you
will
get
your
same
old
kind
of
interface
so
that
so
now
you
don't
actually
have
to
manually
create
a
profile
in
case
you
just
want
to
run
an
office.
D
A
A
One
one
of
the
ones
that's
a
little
bit
harder
and
it
takes
a
while
to
get
to
a
lot
of
times
it's
just
especially
if
you
don't
have,
and
you
don't
have
full
mockups
that
show
all
of
what
would
ideally
be
there
what's
something
that
takes
a
little
while
is
like.
You
know
how
we
have
results
here.
We
have
results
and
updated
on.
A
It
is
that's
the
right
info.
That's
that's
great
information
to
show
someone
is
they
have
a
they?
Have
this
test
profile
a
how
many
times
how
many
test
results?
How
many
times
have
they
run
it
basically
like
how
many
results
do
they
have
for
this
particular
profile?
One
great:
when
was
the
last
time
they
ran?
When
was
the
last
time
they
updated
this
profile,
they
they
reconfigured
the
profile.
A
A
It's
good
data,
the
formatting
of
it,
though,
along
with
how
to
present
it
in
a
way
in
which
not
a
lot
of
white
space
is
or
feel
that
there's
there
is
white
space,
but
it
doesn't
feel
wasted,
and
by
that
I
mean
like
if
you
go
to
view
test
results
at
the
top
of
the
you
know,
this
is
white.
Space
makes
things
feel
clean,
but
on
the
upper
right
hand,
side
like
about
one
third
of
what
we're
showing
is
not
used
and
that's
important,
because
there's
a
hide
test
results,
button
that
takes
up.
A
So
if
you
were
to
take
that
button
and
move
it
up,
you
know
in
line
with
the
details.
Well
great,
that
would
save
some
white
space,
but
it
would
also
be
a
little
harder
for
users
to
follow,
because
the
nice
thing
about
what
you've
done
is
that
you've
left
the
button
in
the
same
location.
So
if
you
click
hide
test
results,
and
then
you
get
view
test
results
like
is
right.
The
user
feels
pretty
good
because
things
didn't
if
you
click
it
again,
there's
a
small
animation.
A
A
A
A
Those
are
awesome
details
and
they
are
so
incredibly
hard
to
read
and
dissect
and
understand
what
the
heck
those
things
mean,
and
it's
not
your
fault
or
you
know
this.
Isn't
this
isn't
your
fault?
This
is
like
this
has
been
here
from
the
start,
but
those
are
awesome
response
time,
histogram
at
max
target
qps
what
the
hell
is
a
qps.
A
A
A
Well,
if
we're
going
to
show
back
to
them,
here's
what
you
configured
and
here
was
the
actual
we
might
want
to
show
it
back
to
them
in
a
simpler
whoops
in
a
similar
format
as
to
how
they
configured
it.
So
like
again,
here's
the
the
test
profile.
Here's
the
test
profile,
configuration
format
that
form
these
values
are
grayed
out.
A
A
A
Before
we
move
to
hussein,
so
I
realized
now
that
we're
about
10
minutes
left
beautiful.
I
love
the
use
of
the
colors
here
austin.
If
you
have
some
information
architecture,
ideas
about
you
know
how
to
use
up
that
white
space.
How
to
convey
some
of
this.
One
thing
that
I
think
would
be
neat
to
see
is:
is
that
results
one
or
three
or
as
a
as
a
big
number?
D
A
A
This
is
helpful
getting
at
the
the
so
right
now
it
looks
like
one
thing
that
is
helpful
about
this
type
of
info,
like
the
kubernetes
cluster
and
the
nodes
for
it
to
be
presented
in
a
tabular
format,
really
much
more
legible
like,
and
so
it
might
be
that,
like
hey,
that
data
is
beneath
we'll
mock
this
up,
but
that
data
would
be
beneath
the
performance
chart.
It
would
be
underlying
like
hey.
This
was
the
performance
in
context
of
this
system,
and
then
yeah
here
was
your
perform.
Here.
D
Yeah
yeah,
actually
one
thing
that
I'm
yet
to
figure
out
is,
I
don't
know
if
anyone
noticed
or
not
in
previous
graph.
This
information
was
missing,
but
in
this
one
it
has,
and
it's
not
front
end,
which
is
skipping
some
information
either
it's
there
in
the
result
or
it's
not,
and
I
still
have
to
look
into
why
it's
in
some
results
and
why
it's
not
in
some
results.
I've
noticed
that
it
usually
happens
when
you
run
a
first
result:
it's
not
in
first
visit
and
but
in
when
you
run
subsequent
results.
A
A
C
C
If
the
other
thing
uses
api
dot
config,
so
both
of
them
have
this
context
info
or
thing
for
all
this,
that's
the
common
thing.
So
I'm
looking
into
how
can
we
basically
make
use
of
this
info
to
achieve
what
we
want
to
achieve
for
gke,
because
gk
is
a
conflict
which
is
a
bit
complicated
like
it?
It
has
many
steps
and
I
had
a
word
with
michael
as
well,
so
he
has
suggested
that
few
things
could
be
done
by
passing.
This
essay
account
name
or
such
things
as
input
parameters.
C
The
script,
failing
with
multiple
errors,
which
were
not
really
intuitive
to
the
user,
so
one
change
that
I
had
was
to
silently
overwrite
whenever
we
unzip
and
another
thing
is
about
install
path,
so
we
can
give
environment
variable
and
see.
Basically
what
is
happening
is
the
current
user
does
not
have
sudo
permissions
or
he
is
not
having
access
to
user
local
bin.
C
So
that
is
where
the
issue
was
so
in
such
cases,
if
they
choose
to
run
mastery
in
a
in
their
home
directory,
we
should
be
able
to
give
them
permission
to
install
it
there
so
that
the
machine
installation
won't
fail.
So
this
was
the
recent
thing
and
before
that,
I
also
had
worked
on
the
silent
flag.
Basically,
if
you
want
to
silently
start
docker
if
it
is
not
running
already
so
that
was
added
to
system
start
command.
C
Start
mystery
with
platform
kubernetes,
maybe
otherwise
how
how
to
identify?
What
is
the
user
user's
platform
without
asking
the
user
right
asking?
The
user
is
pretty
straightforward,
they
might
have
docker
desktop
as
well
as
kubernetes.
Also,
what
is
your
reference
is
something
which
we
can
ask
the
user
or
we
can
decide
it
for
ourselves.
A
A
We
do
want
to
fanatically
focus
on
time
to
value
on
like
hey.
I
ran
a
command
to
install
this
thing
and,
oh,
my
goodness,
it's
showing
me
statistics
on
my
pre-existing
service
mesh
like
it
like
you
did.
I
just
ran
the
bash
script
and
all
of
a
sudden,
some
stuff
downloaded
some
stuff
started
to
run
the
my
browser
came
up.
A
I
clicked
to
log
in
and
now
I
see
my
cluster
and
I
see
the
fact
that
I
already
had
istio
and
then
there's
a
grafana,
that's
already
connected
and
I'm
seeing
statistics
and
like
oh
wow,
oh
awesome,
if
we
get
it
right,
that'll
actually
happen
and
like
we
put
a
lot
of
effort
so
anyway,
your
efforts
here
don't
go
to
waste
like
people
have
that
moment
and
it's
those
things,
those
delightful
little
moments
that
have
people.
Turning
around
saying,
hey,
have
you
have
you
seen
this
tool?
A
You
should
you
know,
or
people
turn
around
saying
what
tool
you're
using
that's
pretty
cool.
It
is
well
you
know,
here's
the
thought
like
if
the
so
so
two
things,
a
bunch
of
things,
one,
the
silent,
installer
and
some
of
the
environment
variables
that
we
have
in
the
bash
script,
fantastic,
we're.
We
need
to
open
up
an
issue
to
make
sure
that
we've
got
that
documented
on
the
docs.
Yes,.
A
And
then
for
the
auto
discover
auto
identifying
what's
the
most
appropriate
platform,
for
you
know,
docker
kubernetes
is
like
well,
you
know.
One
of
the
approaches
can
be
can
be
today
we're
looking
to
see
if
there's
a
docker
service
available
tomorrow
we
can
be
looking
to
see
if
there's
a
kubernetes
conf
I
mean
we
look
for
a
kubernetes
config,
but
also
to
attempt
to
connect
to
it
and
ping
it
like
like.
It
might
become
somewhat
apparent
that.
A
If
the
docker
service
is
available,
you
know
like
we
might
be
able
to
make
get
50
of
this
automated
ish
or
some
certain
percentage,
but
then
yeah
the
other
half
the
time
or
the
or
the
the
the
default
switch
case.
If
you
will,
the
catch-all
is
okay.
Fine,
just
just
ask
them
you
know:
did
you
did
you
want
to
run
this
in
cluster
out
of
cluster?
You
know,
but
we
might
be
able
to
get.
C
A
You
mean
like,
as
their
default
preference
for
yeah
yeah,
it's
possible
yeah,
I
mean
it
is
yeah,
that's
possible
because
yeah
right
now
we're
just
we're
making
we're
internalizing
that
default
assumption
that
it
would
be
one
platform
versus
the
next.
If
we
include
logic
that
yeah
it,
I
that's
a
good,
I
don't.
I
think
that
we
might
find
it
some
6001
half
dozen
or
whatever.
A
It's
that
it's
potato
potato
in
terms
of
like
today,
a
user
is,
if
for
any
given
command,
they
are
able
to
specify
a
global
there's
a
global
flag
for
config,
I
believe
to
specify
like
a
different
anyway.
It
might
like.
I
don't
know
this
might
be
a
premature
optimization,
like
getting
user
feedback
on,
because
today
the
quick,
the
quick
of
it,
is
that
the
user,
you
know
in
a
very
human
readable
way,
can
go
in
and
you
know,
switch
out
their
their
platform
for
a
given.
A
We
should
get
some
of
these.
These
thoughts
on
paper,
I
think,
like
it
or
it'll
it'll
help
make
it
clear.
What's
I
don't
know
if
the
menu
pointed
you
to
a
design,
spec
or
or
if
he
had
suggested,
to
put
some
of
those
thoughts
into
the
mescheri
this?
This
main,
this
main
dock
or
not,
but
putting
some
of
those
thoughts
into
here
is-
might
help.
C
A
A
I
think
in
in
some
in
part.
That's
what
you
were
saying
with
respect
to
like
the
info
getting
getting
cluster
info
and
then
dissecting
what
the
distro
name
is
or
like
what
the
platform
name
is,
but
even
from
the
cncf
like
we
can,
I
can.
I
should
be
able
to
help
get
you
in
contact
with
people
that
can
like
readily
point
us
to
utilities
that
do
that.
C
Yeah,
so
basically
our
from
the
mystery
bash
script
point
of
view.
We
have
to
decide
the
preference
for
the
users
right
like
we
need
to
give
preference
to
either
kubernetes
first
and
then
we
look
for
docker
next.
So
this
station
is
with
us
basically
right
depending
upon
what
would
be
the
mostly
deployed
thing
right.
So
even
from
the
conflict
point
of
view,
if
you
see
in
eks
as
well
as
aks
and
gke,
it's
all
kubernetes
clusters
right
so.
A
Pervasive
support
for
kubernetes
as
a
platform
that
we
should
probably
just
switch
the
the
catch-all
use
case
or
the
default
to
kubernetes
as
like
to
in
cluster
deployments
and
stop
causing
ourselves
pain
with
nat,
networking
and
bridge
networking
and
other
you
know,
and
that
yeah
it
can
be
in
the
bash
script.
It
could
be
that
I
would
encourage
us
to
write
as
little
into
the
bash
script
as
possible
because
we'll
get
more
reuse
out
of
logic,
that's
written
into
both
mastery
ctl
and
mastery
server.
A
A
Oh
yeah
yeah
I
mean
mestre
ctl
needs
to
identify
what
platform
so
that
measures
server
can
be
deployed
on
that
platform,
and
then
you
know
between
the
two
of
them.
They
would
hopefully
share
some
logic.
Maybe
a
mesh
kit
that
identifies
cluster
type
but
also
makes
it
a
little
easy
to
potentially
switch
between
platforms.
Hussain
when
we
get
to,
we
need
to
create
a
design.
Spec
on
the
concept
of
environments
and
environments
can
be
a
kubernetes
based
or
a
docker,
based
and
and
having
a
control
and
having
a
construct
called
an
object,
called
environments.
A
A
Yeah
never
mind
I
was
going
to
say
they
could
get
mystery
up
and
then
reconfigure
part
of
their
kubernetes
connection.
But
I,
but
really
that's
not
so
much
true
like
it's
mostly
in
mastery
ctl,
where
they
need.
We
need
to
figure
out
how
to
actually
connect
to
kubernetes,
to
be
able
to
deploy
measuring
into
that
kubernetes
environment.
A
D
D
Gcp
offers
access
tokens,
and
I,
like
today,
told
me
that
distribution
also
does
the
same,
so
they
provide
access
tokens
for
authentication
and
that,
after
spoken
time
is
like
can't
be
renewed
using
google
cloud,
and
that
is
something
that
can
be
automated
and
that
is
automated
using
trying
to.
But
the
thing
is
that
you
don't
actually
install
g
cloud
sdk
on
system
while
mystery
gets
installed.
So
that
is
something
that
either
you
automate
or
else
so
the
process
automated.
D
If
the
token
had
expired,
then
the
plugin
would
come
into
play
like
come
into
play
and
it
will
try
to
like
renew
the
token
using
the
bind
written
like
using
the
binary,
which
has
been
specified
in
the
cube
content.
But
as
that
binding
doesn't
exist,
so
it
will
start
showing
error
that
hey
youtube
has
expired
and
I
cannot
access
this
boundary
to
refresh
your
token
other
than
that.
Gcp,
like
machine
guns,
run
fine
on
gcp.
I
deploy
this
uin
as
well
as
prometheus.
C
D
D
So
basically
it
was
kind
of
giving
a
false
error,
so
that
was
also
one
of
the
issues.
So
there
were
a
few
issues
which
we
can't
fix
very
easily.
One
of
them
has
like
would
be
fixed
once
through
three
hours
is
merged
and
a
few
are
there.
Like
I
mentioned
earlier,
I
then
actually
talked
about
importing
a
a
library
that
is
a
go
client
trying
flow.
D
It's
plugins
once
we
import
them,
then,
basically
it
has
authentication
plugins
and
it
can
authenticate
automatically
with
the
keyboardist
provider.
So
that
is
second
barrier.
We
just
have
to
import
it.
I
mean
not
difficult
this
g
cloud
sdk
was
the
only
thing
which
was
a
bit
not
a
value,
but
it
was
something
that
we
are
not
doing
again.
It
would.
A
We
should
poke
the
venue
to
make
sure
that
we're
getting
this
into
into
a
design
spec
into
the
right
design,
spec
you're
getting
on
paper-
oh
god.
I
so
desperately
want
to
keep
talking
about
this
well,
we'll
we'll
catch
up
more
in
ch
and
chat
and
in
a
design
spec.
This
is
good.
Abhishek
got
off
the
hook
without
having
to
show
a
measuring
operator
today.
So
next
time.