►
From YouTube: Meshery CI Working Group (May 11th, 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).
B
B
B
B
B
And
then
there's
a
sense
says:
cush
isn't
on
I
was
just
talking
to
him.
It
makes
me
wonder
if
he
isn't
hanging
on
a
different
Google
me
so
I'll
poke
him
real
quick,
but
while
I'm
doing
that
just
a
call
for
topics
as
we
get
up
and
going
so
I
think
ID
I
think
you
you
were
mentioning
one
so
feel
free
to
pop
pop
them
into
the
meeting
minutes
that
you
want.
B
B
Boy
I
think
I
think
I'm
still
waking
up
here,
so
it
did
do
as
I
bring
this
one
up.
This
was
submitted
by
a
relatively
new
community
member
Springville
and
I
think
he
had
corrected
I
guess.
Maybe
this
sort
of
indirectly
came
up
I
think
he
had
corrected
some
indentation
issues,
I
think
with
the
yeah
mall
that
we
have
that
describes
our
work,
clothes
and
in
in
that
process
it
I
think
became
obvious
to
us
that
between
the
various
jobs,
if
I'm
using
the
right,
github
action
or
not
I
shouldn't,
say
action.
B
Within
the
project
itself,
we
don't
always
need
to
run
all
of
the
jobs
depending
upon
what
changes
are
included
in
a
pull
request,
as
we
go
to
break
down
those
individual
jobs
so
that
they
can
be
called
independently.
It
seems
like
part
of
breaking
those
down,
might
introduce
some
repetitive
tasks
when
they're,
when
they're
are
called
when
they're
all
called
and
they're
all
invoked.
Maybe
some
of
those
tasks
only
need
to
be
invoked
or
could
be
invoked
less
than
we
do
so.
I
think
I
think
that's
the
open
question
here.
B
B
E
God
he
machines
to
cover
the
workflow
as
I
as
I,
had
told
in
the
last
CI
meeting.
That
sin
is
the
gate
of
action
still
that
mature.
So
it
asked
to
streamline
the
number
of
jobs
which
are
running
a
PR
to
to
observe
the
matter
at
all
here.
So
since
we
were
discussing
about
released
after
and
labeling
of
yard
on
that
day,
so
I
suggested
that
a
free
label
PR
according
to
their
title.
B
E
Areas
no
I
didn't
do
this
because
we
need
to
label
according
data,
provides
an
action.
Data
provides
a
way
to
observe
the
PR
title
with
the
PR
metadata
and
we
can
use
an
if
condition
for
each
and
every
workflow
and
the,
if
condition
we'll
check.
If
the
PR
is
having
a
certain
title,
then
only
it
will
go
on
and
run
the
certain
work
action
for
this,
and
if
the
PR
title
does
not
matches
with
the
with
the
condition
of
which
we
have
specified
in
the.
If
so,
it
will
not
run
okay.
B
B
B
Okay,
this
may
be
related
or
maybe
I'm
conflating
two
things.
There
was
I.
Think
part
of
what
why
this
discussion
initially
came
up
was
like
the
use
of
cash.
I
think
was
like
the
example
of
tasks,
sort
of
the
the
fact
that
we're
setting
up
this
cash-
like
you
know
each
each
and
every
job
it
may
it
seemed.
B
B
Okay
and
it's
not
when
like
lint
passes,
but
there's
an
air
check
that
fails
that
as
a
someone
as
you're
looking
at
the
PR,
you
can
see
that
the
failure
is
in
the
error
check
portion,
but
not
in
the
limp,
like
the
the
lint
check
passed.
So
it's
nice
in
that
regard
that
these
are
broken
out,
but
where
that
becomes
less
efficient,
maybe
is
the
the
notion
that
we
need
to
essentially
repeat,
like
this
whole
set
up
this
whole
test.
The
whole
job
set
up.
B
It's
really
just
this
last
portion
about
the
actual
check,
that's
being
run,
that's
different.
If
we
were
to
combine
these
checks,
it
seems
like
the
negative
aspect
of
doing
that
would
be
that
you
wouldn't
necessarily
know
right
away
whether
it
was
a
link
or
an
error
check
that
failed,
because
they
would
be
on
the
same
job.
E
B
B
It
might
make
sense
that
it's
called
out
separately
so
that
a
reviewer
can
easily
see
a
required
check
failed,
and
so
you
know
the
chair
shouldn't
be
accepted,
whereas
we
group
together
each
of
the
the
jobs
each
of
the
checks
that
are
acceptable
to
fail
if
there
are
any,
if
all
of
those
are
grouped
together,
that
that
might
be
the
right
balance.
I,
don't
know
I'm
just
just
thinking
out
loud
right
now,
given
that
we've
got
a
clean
builds
like
this,
it
might
be
the
case
that
we
would
say.
B
Okay,
so
I
think
what
we'll
do
is
I'm
gonna
leave
this
as
an
open
question.
Just
the
the
lack
of
kind
of
conversation
and
perspective
here
makes
me
think-
hey.
Maybe
we
just
haven't
had
enough
time
to
think
through
this,
so
I'll
try
to
take
some
additional
notes,
but
maybe
we'll
leave
this
open.
Let
people
kind
of
weigh
in
on
this
question.
B
F
Yeah
I,
just
I,
just
asked
it
that
in
particular
case
that
job
contains
I'm
working
with
the
agile
devops
pipeline
in
the
last
weekend.
So
that's
have
jobs
and
either
the
bid
and
the
job
they
have
tasks
available.
So
there
is
a
for
one
job,
one
job
there,
three
tasks
and
one
job
has
two
tasks.
So
I
can
say
that
I
have
a
pipeline.
I
have
built
a
quote,
compile
that
into
it
run
the
linting
desk
and
now
the
testing
involved.
F
So
each
my
one
build
job,
has
three
tasks
associated
with
it,
so
I
think
we
can
make.
We
can
break
down
the
tasks,
a
job
into
smaller
tasks,
so
smaller
tasks
there
and
then
we
able
to
find
out
if
none
of
those
tasks
earlier
phase
doesn't
allow
that
to
continue
the
next
steps
and
also
if
that
task
is
passing,
we
can
diddle
the
output
of
the
next
task
of
that
passing
test.
So
he
doesn't
have
to
rebuild
the
entire
thing.
So
he
have
some
output
parameter
going
on.
G
E
E
H
B
That
got
added
over
here.
The
spell
check.
Maybe
this
to
do,
is
Naveen
on
Kara.
Maybe
the
thing
to
do
is
have
someone
who's
and
what
that
looks
like
and
how
a
lot
of
our
okay,
maybe
what
I'll
do
is
we
can
comment
on
the
the
open
PR
and
ask
for
a
kind
of
a
recap
and
presentation
of
how
what
that
looked
like,
given
that
we've
been
running
it
in
the
sandbox
for
a
while
good,
because
I
think
our
initial
concern,
if
I
recall,
was
I'm
going
to
false
positives.
B
E
E
It
is
so
I
opened
this
issue
recently
to
add
UI
testing
using
Cypress.
We
have
various
cui
free.
We
have
various
these
frameworks
to
do.
Front-End
testing
like
mukajee's,
selenium
chai,
and
there
are
many,
but
there
are
problem
with
mukajee-
is
that
it
is
quite
so
it
is
quite
powerless
and
with
selenium
it
is
quite
heavy,
so
Cypress
has
been
in
the
industry
for
quite
a
while,
and
it
has
around
16
to
17
K
k't
upstarts,
and
it
is
also
open
source.
E
It
is
quite
developer
friendly
and
it
hides
around
each
and
every
UI
action
which
we
can
think
like
click
forward,
scroll
or
the
filling,
the
form
and
etc
so
I
have
just
initialized.
The
Cypress
testing
I
have
made
our
UI
ready
to
accommodate.
Cypress
tests
and
I
have
also
written
a
simple
test
which
is
over
here.
So
this
is
just
check.
If
Mishra's,
you
is
loading
or
not,
it
will
visit
up
the
local
Lowe's
5000
coming
to
the
port.
E
I
have
had
to
make
the
changes
to
the
port,
because
the
Cypress
occupies
3000
port
and
if
you
would
run
Cypress
test,
the
the
development
server
will
not
will
not
fire
up
because
the
3000
port
was
already
occupied.
You
can
also
run
the
Cypress
test
on
local.
You
have
to
just
check
out
the
PR,
and
you
have
to
just
do
and
p.m.
runs
side.
E
E
E
Over
here
you
can
see,
Cyprus
starts
the
server.
The
server
has
been
start
once
the
server
is
started,
Cyprus
rinses
test.
So
here
are
all
the
tests
which
have
been
passed
and
it
gives
the
result
also
Cyprus
bundles,
its
bundles,
the
package.
So
if
I
have
run,
if
I
have
rained
the
Czech
ones
and
if
recent
PR,
which
also
includes
the
tests
so
Cyprus
install
its
image
again,
it
will
be
taking
up
its
image
from,
and
the
location
of
the
cache
is
over.
Here.
Home
run,
Akash
Cyprus
and
the
address.
E
That's
quite
it
once
the
community
approves.
We
can
merge
the
PR
and
then
start
writing
tests
for
you.
Why?
Because,
as
I
have
seen,
if
you
will
fire
up
the
UI
without
running
the
gulang
server,
there
are
multiple,
multiple
pages
which
you
can
exist
like
the
dashboard,
the
performance
page
and
the
settings
page
the
preference
of
speech.
So
this
can
be
a
good
start
to
introduce
testing
in
Mishra.
B
I
E
Also,
the
Cyprus
abbreviations
are
quite
easy
like.
If
you
want
to
click,
you
can
just
use
the
dot
lick
function.
If
you
want
to
grab
an
element,
you
can
use
the
classes
of
the
element
to
stuff
specify
the
function.
So
if
one
goes,
FN
goes
around
the
documentation
of
Cyprus,
you
will
see
they
are
quite
easy
and
any
newcomer
can
write
some
tests
with
it.
So
it
won't
be
a
big
task.
It
has
a
very
simple
learning
curve.
It's
not
quite
steep.
B
B
E
Actually,
I
have
been
playing
with
it
for
some
time
able
to
grab
the
material
USS,
which
way
we
have
been
specified
and
coming
to
the
fragility
of
the
test
says
we
lack
the
testing
of
the
server
and
since,
in
recent
days
we
observed
that
result
result
page
was
broken.
The
sort
function
was
not
working,
so
this
kind
of
issues
can
be
easily
solved
with
Cypress
as
we
can
mimic
the
click
function
and
the
click
function
does
not
require
server
to
be
running,
and
since
we
are
using
next
days
for
a
UI
and
exit
server-side
rendered.
E
So
if
you
will
just
run,
if
you
just
run
the
UI
and
you
won't
fire
up
the
server,
you
are
able
to
see
that
there
are
multiple
pages
which
you
can
visit
the
UI
and
you
can
even
play
with
mishra.
There
are
some
issues
like
you
can't
run
the
performance
benchmark
or
you
can't
authenticate
your
user,
but
there
are
around
many
functions
which
still
available
without
the
server
being
run
on
and.
E
E
B
Right,
I
think
that
yeah
I
think
you
of
the
answer
to
my
question
about
you.
The
tests
being
fragile
I
think
you
spoke
to
part
of
it
cause
I.
Think
in
general
you
were
saying
that
I'm
having
having
an
end-to-end
or
an
integration
test
capability,
whatever
that
it's
Cypress
or
other,
would
be
really
helpful
and
would
capture
a
number
of
issues
even
just
locally
that
you
can
you're
able
to
test
out
various
aspects
of
mysteries
functionality
without
necessarily
running
the
server
but
yeah
even
running
the
server
locally,
as
isn't
much
of
an
issue
either.
B
So
I,
totally
again
and
I
totally
agree
that,
like
hey
we're,
it's
almost
it's
a
little
bit
nerve-racking,
sometimes
to
merge
a
PR
because
we
don't
have
a
robust
set
of
and
n
tests
of
various
tests.
I
think
the
thing
I
was
trying
to
get
out
about
the
test
being
fragile
is
that
some
of
the
tools
will
record
your
clicks.
B
E
To
the
point
that
about
the
location
of
the
cursor
and
location
of
the
buttons,
so
since
the
Cypress
Cypress
can
also
support
the
location,
the
native
way,
but
Cypress
gives
you
gives
your
selector
like
I.
Don't
it's
good?
The
dog
get
function
and
the
dog
get
function.
You
can
specify
the
class
name
for
the
element
ID
and
it
will
return
you.
The
it'll
tell
you
the
number
all
the
array
of
elements
which
it
found
and
from
this
you
can
use
dot,
get
dot
finders,
you
can
use
the
dot,
find
function
and
in
dot
fine.
E
You
can
specify
what
should
be
the
HTML
written
over
the
button
or
the
class
which
he
which
it
has
script
and
we
can
easily.
We
can
easily
use
use
the
click,
I
click
function
or
the
transform
function,
so
there
are
around
30
or
35.
Such
selectors
I
then
go
through
all
of
them,
but
I
went
through
the
one
which
which
we
can,
which
can
be
quite
handy.
B
F
F
So
there
is
here
that
we
have
to
rely
on
the
clicks
button,
Mouse
moves
and
other
button
kind
of
items,
and
these
are
moved
within
the
page
drastically
where
they
are
going
and
we
are
not
sure
the
login
betting,
where
goes
to
the
top
or
the
bottom
of
the
center,
and
how
the
website
of
I
think
there's
a
challenge
here
is
to.
It
is
one
things
that
we
keep
in
mind:
that
it
doesn't
impact
the
performance
type
of
performance
of
our
app.
F
The
page
that
we're
loading
in
is
necessary,
the
page
that
allows
and
and
number
and
number
two
that
if
we
most
of
the
time
there's
the
front
end
page,
it's
they
rely
on
the
server-side
rendering
for
the
initial
page
to
be
downloaded
and
represent
every
faster
way.
So
I
think
quite
a
few
challenges
here,
but
I
think
one
of
the
biggest
challenge.
That's
I
faced
over
the
year.
F
That's
if
we
start
introducing
the
unit
testing
of
UI
testing
and
this
increase
perform
style
and
quiet
through
drastically
and
then
which
things
we
have
to
remove
these
kind
of
thing
so
I
think
it's
a
good
step,
because
we
have
to
have
to
do
some
kind
of
us,
the
unit
testing,
for
that.
But
it's
nice.
It's
it's
time
to
be
investigated
properly,
whether
it
is
impacted
the
performance
time
of
the
ad
or
not.
F
Good
question
yeah,
yet
the
pointers
are
asking
here:
the
last
I
am
working
with
the
angular
app
for
quite
some
time.
So
this
is
a
bundles
that
we
build
the
compiled.
They
have
a
bundle
file
initially
generated.
So
when
we
start
introducing
the
testing
there
that
last
time
for
the
angular
project,
so
when
we
starting
they
load
the
librarian,
forgetting
what
was
first
kind
of
that.
So
it's
increase
the
bundle
size
of
my
app.
So
that's
something
again.
F
Yes,
for
my
part,
I
think
we
should
go
for
it
and
that's
great
step
so
moving
forward.
We
have
two
takes
how
to
handle
this
kind
of
how
to
handle
particularly
testing
in
a
project,
particularly
if
it's
continuous
integration
but
I.
Think
in
that
time
you
should
start
exploring
that,
but
they
are
an
advantage,
is
and
I
think
terms
of
good
would
be.
Oh
good
should
go
for
it.
B
Good,
my
understanding
right,
if
we
were
to
move
forth
with
the
proposed
use
of
Cyprus
right
now,
is
that
the
Cyprus
libraries
would
be
included
in
the
basically
in
in
the
UI
bundle
and
that
individual
users,
you
know
that
their
web
agents
I'm
download
when
they
initialize
a
session
with
massery
and
depending
upon
how
large
those
are.
You
know
how
learn
if.
E
J
We
can
have
NPM
install
pradhan
NPM
install
dev,
so
when
we
run
the
test
is
gonna,
be
the
MP
mr.
Davis
to
be
an
option
and
then
so
it
in
the
final
bundle.
No
development
library
should
be
included,
like
any
library
needed
for
tests,
for
instance,
or
we
can
yeah.
There
is
only
two
boxes
like
rodent
tester,
prevent
them
from
what
I
know,
but
yeah.
E
B
J
B
To
me,
it
seems,
like
you
know,
the
various
tasks
that
are
exercised
throughout
the
build
process
that
that
one
of
those
would
be
to
you
know,
build
using
Cypress
execute
the
tests
that
are
defined
if
those
tests
pass
do
another
build
at
the
project
without
Cypress
and
that
artifact
produced
in
that
that
second
build
that
that's
the
artifact
that's
released,
such
that
Cypress
isn't
included
at
all
in
the
released.
Artifact
am
I
thinking
of
that
wrong.
J
B
B
And
that's
actually
a
good
point
that,
like
hey,
you
know
as
we
get
into
further
granular
control
over
our
builds,
that
it
may
be
the
oh
yeah
I,
won't
let
go
when
we
gotta
do
a
release
day.
The
part
of
the
workflow
is
either
in
book
is
invoked
based
on
the
presence
of
a
release,
tag
and
so
yeah.
It
would
seem
I.
B
H
E
E
Is
the
dependency
or
the
or
the
node
modules
which
we
have
specified
now
are
not
currently
packaged
with
the
release
artifacts
so
whether
it
is
of
250,
mb
or
whether
it
is
of
hundred,
and
it
should
not
make
any
change
in
to
the
size
of
the
wizard
which
we
are
currently
making.
But
in
later,
if
we
start
dealing
the
mystery
server
and
if
we
start
by
giving
you
a
also,
then
there
we
can
have
a
discussion
about.
How
should
we
deal
with
this
yeah.
B
Okay,
yeah,
that
makes
sense,
actually
the
it.
If
we
do
merge,
this
PR
probably
will
affect
the
size
of
the
measure
e
releases
because
the
while
the
only
thing
that
we're
doing
a
good
job
of
versioning
today
is
mastery
CTL
and
putting
those
artifact
that
artifact
available
for
download
on
github,
the
other
artifacts,
which
are
the
mystery
server
container
image,
which
includes
the
UI
and
the
other
individual
adapter
images
and
those
those
get
built
and
pushed
to
docker
hub
and
so
yeah.
It
seems
like
if.
B
So
well
well,
I'm
sensitive
to
the
notion.
We've
only
got
ten
more
minutes.
I
do
want
to
make
sure
we
get
to
you
tit
discussion.
I.
Think
we've
got
a
little
bit
of
I.
Think
we're
in
general
alignment
and
agreement
about
the
notion
that
the
testing
is
needed
here.
Cyprus
seems
like
it's
the
right
choice,
but
that
before
we
merge
that
in
we'll
want
to
ensure
that
we're
not
you
know,
bloating
the
size
of
those
container
images
and
there
might
be
some
pre-work
in
other
areas
to
do
before.
B
F
They
just
just
as
last
and
final
comments
about
that
I
think.
The
best
practice
here
is
to
you
have
the
unis
testing,
the
and
integration
that
is
end
to
end.
If
you
have
up
websites
70%
of
the
toss,
you
completion
by
unit
testing
go
for
a
units
testing
if
50%
of
the
tasks
are
going
to
the
unit.
Testing
and
50
integration
go
for
the
tool
that
have
boats
in
it
and
if
you
have
30%
that
as
a
fulfillment,
the
unit
testing
is
70%.
That
end-to-end
testing
then
go
for
the
tool.
F
That's
I
have
unit
end-to-end
integration,
testing,
I
think
if
we
have
the
requirement
to
include
it,
go
for
it.
If
we
have
accomplished
that
task
by
just
writing
the
unit
test,
then
I
think
we
should
go
for
the
tool
that
have
you
noticed
and
that's
for
a
lightweight
so
I
think
in
future.
We
can
talk
more
on
that,
but
I
think
go
for
is
cush
can
work
on.
F
There
are
sandbox
environments
and
what
the
impact
is,
but
I
think
the
best
practice
is
that
if
the
accomplishment
done
the
unit
testing
most
of
the
tasks
you
have
achieved
using
unit
testing
go
for
the
tool
that
has
unit
testing,
enable
in
it
that's
lightweight,
that's
perfectly
250
or
260
can
be
of
size.
That's.
C
J
D
This
demo
session
I
will
introduce
you
about
the
Kaos
action,
Kaos
engineering,
using
an
action,
github
actions
and
how
we
can
use
different
chaos,
experiments
on
our
application
part
in
our
workflow.
So
this
is
one
of
the
sample
workflow
ml.
So
it's
just
on
the
question
pool
across
phases
over
master
branch
and
we
have
created
the
other.
These
are
the
steps
we
followed
here.
D
We
have
created
a
three
node
cluster
in
the
first
step,
then
creating
one
deployment
application
over
which
we
will
perform
the
chaos
and
the
third
step
is
the
litmus
chaos
will
perform
chaos
experiment
over
here.
So
this
is
not
efficient.
This
will
be
the
part
of
1.4
right
now
so,
and
these
were
the
in
vomit
variable
will
be
passing.
D
D
So
if
you
want
to
change
like
you
want
to
modify
some
of
that
pyramid
itself,
you
can
create
a
custom
image
and
can
use
that,
and
these
are
the
other
environment
variables
which
we'll
be
using
for
the
curve
like
how
many
times
scale
it
will
continue
and
what
we
will
do
intervals
between
the
chaos.
And
then,
if
you
want
to
delete
the
litmus
just
after
the
chaos
completion,
then
you
can
just
keep
it
as
a
true
and
then
after
and
deleting
the
cluster.
D
D
I
D
I
Yeah
thanks,
who
did
I
think
just
I.
Think
a
lot
of
folks
who
are
in
this
meeting
have
already
attended
the
session
that
we
had
on
litmus
earlier
weeks
earlier.
So
don't
want
to
really
go
into
the
specifics
of
litmus
as
a
chaos'
toll
and
how
it
facilitates
e
to
be
etcetera,
just
wanted
to
bring
out
an
aspect
which
we
did
not
discuss
then.
I
So
we
did
have
this
discussion
about
how
developers
can
write
their
tests
and
they
can
create
their
case
experiments
and
how
they
can
draw
it
in
their
staging
or
production
systems.
But
the
paradigm
is
people
are
now
shifting
left
and
running
chaos
tests
in
CA
as
externally
a
TV
I
think
there
was
a
very
good
discussion
before
this
about
the
cypress
tool
and
I
also
learned
a
lot
about
UI
testing
and
all
the
considerations
that
are
being
made.
I
I
So
this
is
basically
one
abstraction
over
the
experiments
here
itself,
so
you
have
your
own
actions
or
get
lab
remote
templates
that
you
can
readily
include
inside
of
your
existing
workflows,
and
at
this
point
we
are
asking
the
developers
write
their
own
exit,
chicks,
but
probably
there
can
be
a
way
to
generate
these
action
templates
as
well.
There's
DK
that
we
are
coming
up
with
is
going
to
be
allowing
you
to
create
experiments,
but
maybe
that
can
be
extended
to
create
these
kind
of
actions
as
well.
I
So
what
that
can
do
is
it
can
help
you
construct
CIE
systems
like
this,
where
you
have
your
integration
or
your
tests
that
are
already
performed
in
this?
You
want
to
do
some
e
to
e,
then
you
can
actually
construct
multiple
chaos
tests
as
part
of
one
or
multiple
stages.
These
are
all
tests
in,
or
chaos
tests
inside
the
open,
EBS
pipelines.
I
So
you
can
see
that
all
these
are
actually
chaos,
jobs
that
are
being
run
out
of
good
luck
templates.
So
we
just
wanted
to
sort
of
simplify
this.
That's
where
we
are
trying
to
come
up
with
the
actions
and
more
templates.
For
example,
there
is
this
peer
that
we
submitted
into
the
cross
cloud
CI
for
Corvinus
tests.
I
So
you
can
see
here
that
this
involves
creating
clusters,
a
kind
cluster
was
created,
and
then
you
apply
a
lot
of
MLS
play
the
litmus
infrastructure
EML,
and
then
you
play
the
experiment,
prerequisites
and
then
you
go
ahead
and
add
the
chaos
engine
specific
to
core
dns,
and
then
you
also
have
some
logic
to
find
out
that
your
excrements
ass
failed
in
select
that
might
want
to
simplify
all
this
and
that's
where
these
actually,
these
actions
are
coming.
So
this
is
a
sample
reporting.
This
is
the
one
that
would.
It
is
actually
demoing.
I
At
this
point.
The
cluster
creation
here
is
because
we
wanted
to
demo
it.
You
might
already
have
that
as
part
of
some
requirement
yourselves,
we
can
just
use
the
cube,
config
and
propagate
that
to
the
Kaos
action.
You
could
pass
some
variables,
and
these
are
all
this
alpha
at
this
stage.
We
might
actually
go
ahead
and
change
it
going
forward
and
make
it
tighter.
So
you
pass
the
experiment
to
nibbles,
and
this
will
help
you
run
all
the
experiments
that
you
need
to
run
and
then
you
go
ahead
and
clean
up.
I
There
is
an
equivalent
for
this
in
gate
lab
as
well.
There
are
some
remote
templates
similar
to
these
actions,
which
you
can
include
like
this
and
then
exchange
those
templates
in
your
jumps.
So
this
is
what
we
are
attempting
to
do
and
trying
to
create
some
actions
which
are
specific
to
the
projects
that
are
there
in
the
CNCs
community
right
now,
they're
all
very
generic
experiments,
let's
say
just
want
to
start
off
today.
Maybe
some
Miculek
component
can
be
installed
and
trade
or
some
chaos.
You
can
reuse
this
going
ahead.
D
The
classification
has
been
completed
and
our
chaos
experiment
is
now
executing
on.
The
engine
is
deployment
which
we
have
created,
and
here
is
the
watch
over
the
name
space
of
the
pods.
So
right
now
we
saw
that
the
runner
pod
comes
in
and
it
is
going
to
create
a
pod
delete,
experiment
pod,
which
contains
the
logic
of
finding
the
target
container
and
executing
the
chaos
whatever
it
is
so
yeah.
So
it
will
run
like
that,
and
then
we
have
other
chaos.
Experiments
as
well.
Like
sure.
D
We'll
be
quick,
we
have
other
experiments,
can
include
it
in
the
action,
so
the
experiment
part
comes
up.
We
can
have
a
resource
kill
like
resource
experiments
like
CPU
hog
memory
have
over
nodes
and
parts.
So
mostly,
we
will
cover
all
the
experiments.
What
are
in
the
litmus
interactions
as
well?
We
can
use
that
also
to
integrate
with
it.
Yeah.
D
B
What
what
is
the
statistic?
Thank
you
for
presenting
this
is
highly
relevant,
pretty
intriguing
because
it
we've
been
and
I
think
exploring
the
use
of
time
and
explore
I
mean
like
it,
the
this
working
group
and
they.
Obviously,
this
is
the
third
time
anyway,
it's
a
pretty
young
area
for
us
a
threat,
lift
your
resume.
Thank
you.
We
just
there
are
miles
to
go
in
terms
of
in
terms
of
giving
a
good
pipeline
and
mastery
itself
as
a
tool
is
well
has
any
number
of
very
significant
dependencies.
B
The
extent
that
it
will
end
up
service
mush
and
then
attempt
to
manipulate
it
and
do
a
bunch
of
things
to
it,
there's
all
kinds
of
potential
for
test
failure
for
things
to
go
wrong
right,
because
there's
a
lot
of
things
that
are
changing
in
its
be.
You
know,
good
service
questions
that
it's
ending
up
there
changing
and
doing
their
own
thing.
So
so
there's
a
lot
of
work
to
be
done
here.
It's.
B
Indian
test
is,
is
I,
think
perfect.
You
are
identifying
the
part
of
the
value,
whether
it
was
nasty
chaos
or
just
a
regular
non
chaotic
death
is
the
kind,
including
all
those
scripts,
and
all
that
logic,
you
know
your
main
workflow
or,
like
anything
like
you,
want
to
compartmentalize
some
of
this
one
iterate
sort
of
version.
If
you
all
it
was
individual
and
the
end
test.
B
B
Actually
specific
to
mystery
and
service
not
use
is
that.
Well,
I
think
it's
both
interesting
anything.
It
clearly,
but
I
think
we've
been
in
the
past,
discussing
about
mysteries,
need
to
facilitate
chaos
on
certain
messes
and
and
see
how
workload
respond
and
help
users
understand
that
better
and
and
that
whole
set
of
conversations,
and
this
one
is
a
bit
more
about
like
how
is
witness
care
of
useful
to.
B
B
But
you
know
that
they
don't
entirely
overlap,
but
they're,
but
there's
even
even
some
overlap
between
them.
I
think
the
more
simple,
the
and
you're
thinking
tests
are,
maybe
the
more
that
those
might
overlap
with
I
guess
that
naturally
do
things
like
it's
simply
provisioning
the
match
or
in
applying
a
certain
to
Vegas.
If
you
want
to
test
to
see,
if
that
keeps
babe
is
we've
been
applied,
my
same
chaos
experimented
that
a
user
might
want
to
run
is
to
remove
part
of
that.
I
Yeah
I
think
we
like
to
be
most
hard
things
that
are
non
chaotically,
so
I,
just
surely
the
open
it
is
not
CIA
used
is
so.
There
are
different
stages
there,
the
functional
stage
that
I
showed
you
are
all
non
chaotic
tests,
they're
mostly
provisioning
stuff,
mostly
sanity
kind
of
experiments,
you
can
say,
and
but
they
are
also
being
cubed
as
little
as
jobs.
So
you
could
sort
of
extend
this
framework
to
do
that
as
well,
and
that's
something
that
we
are
also
seeing.
I
Having
said
that,
the
there
is
definitely
an
overlap.
The
idea
is
sometimes
what
you
do
in
your
CIA
is
what
you
may
even
hand
out
to
your
Assessor
is
to
do
on
production
sometimes,
so
there
is
an
intersection
bit.
There
are
some
cases
which
you
would
like
to
see
being
done
in
a
different
environment,
but
this
can
be
done.
This
can
be
used
for
in
a
non
chaotic
context
test
in.
I
Yes,
right
now,
I
think
we
started
out
by
creating
some
remote
templates
for
get
left
and
then
we've
also
created
some
actions.
Now
for
all
the
I
think
all
the
existing
experiments
are
covered
for
the
club
and
yet
have
there
are
a
few
experiments
remaining
that
is
planned
for
other
sea
ice
as
well.
Maybe
we're
even
looking
to
the
community
to
sort
of
which,
in
their
share
their
ideas,
I
think
yeah
circle,
CA
orbs.
There
are
a
few
options
that
way,
which
we
can
include.
I
J
I
J
I
They
can
be
used
on
any
kubernetes
platform.
I
think
litmus
can
be
run
against
any
capabilities
platform,
the
runners,
so
these
are
actually
the
rappers
or
the
jobs
that
will
run
as
run
so
get
lab
run
or,
if
you're
talking
specifically
about
the
guitar
execution
model,
they
could
be
a
communities
runner
or
a
talker
machine
or
a
shell
based
runner
which
are
executing
these
scripts.
I
J
G
B
I
J
B
This
is
great
eagle
a
little
bit
because,
because
there's
some
interesting
possibilities
here,
we
should
we
should
probably
return
and
see
what
we
can
hit
next
week,
Karthik
and
Rajan,
and
if
you
guys
are
I,
think
that
there's
an
openness
and
a
willingness
here,
it
might
be
that
it
might
be
that
we
should
take
some
of
what
we've
done
with
kind
and
kind
of
stopped
there
and
take
a
next
step
with
a
potential.
It
sounds
like
you
guys
this
might
we?
I
B
I
Yeah
so
I
think
we,
the
get
into
actions
are
not
really
being
used
anywhere
yet,
and
so
we
are,
you
could
say
in
alpha,
so
what
what
it
has
gone
just
now
is
some
on
some
local
reports
that
we
are
maintaining.
We
will
be
pushing
this
internet
bus
and
hopefully
start
seeing
some
adoption
of
it
get
lab.
The
more
templates
have
slightly
more
adoption
because
we're
doing
that
in
open,
EBS,
CA
I
think
github
actions
that,
in
principle
very
similar,
it's
just
that
it
has
maybe
a
slightly
different
operating
model.
B
You
know
let
us
know
when
it's
there
and
more
than
that
I
mean
even
in
advance
of
that
there's
a
sandbox
repository
that,
but
you
guys
could
engage
in
and
do
some
tests
on.
If
you
like,
it
might
be
that
you
guys
would
want
to
do
a
PR
or
dude.
You
know
put
it
in
there
because
well,
I,
guess
to
be
frank,
because,
as
you
can
see,
the
we're
still
trying
to
get
our
sea
legs
in
terms
of
what
we're
doing
with
CI,
so
yeah.