►
From YouTube: 2018-Jun-13 :: Ceph Testing Weekly
Description
Weekly collaboration call of all community members working on Ceph Testing.
http://ceph.com/testing
A
B
A
C
E
C
A
Go
out
the
action
items
from
last
time,
I
created
this
meeting.
It
sent
an
email
so
that
got
done.
I,
don't
know
if
we
have
tickets
in
red
mine.
Yet
does
anyone
did
that
happen
here?
Do
you
know,
or
did
you
see
those
Zach
or
anything,
I
guess,
I
actually,
don't
know,
let's
see
tracker
dot
set
comm
projects.
F
A
A
A
C
E
C
A
A
C
I
still
kind
of
like
that
idea
what
I'll
say
about
about
about
the
OpenStack
back
end
and
getting
patches
in
I'm,
not
sure
how
many
other
users
there
are
of
it.
I'm
not
aware
of
like
anyone
on
our
side
that
that's
using
it
so
yes
submit
patches,
let's
get
them
in
as
long
as
they're,
not
you
know
like
guaranteed
to
break
other
people
that
might
be
using
it.
I
don't
see
a
whole
lot
of
issues
eventually.
I
think
that
back-end
will
probably
die.
A
B
D
A
D
Obviously,
and
some
some
of
the
new
issues
that
and
trying
to
fix,
like
two
guides,
have
troubles
with
workers
want
to
told
so
it's
if
there
is
a
one
bug
new
that
I've
placed
I
figured
out
there
is
you
know
this
week?
It's
it's
one.
If
the
job
is
quitting
by
time
out,
then
there
is
unhandled.
Exception
occurs
in
working
its
die.
Oh
and
I've
checked
the
I've
checked
these
upstream
master,
and
there
is
no
patch
for
this.
Yet
thirds
so
I
go
not
to
create
a
PR
for
this
issue.
Okay,.
D
D
D
D
A
A
A
C
A
I'll
be
good,
okay,
the
I
I
think
so
yeah
from
last
week.
The
action
items
were
for
the
tickets
in
red
line,
which
it
looks
like
aren't
there
yet
and
like
we
just
want
to
get
some
more
PRS,
and
so
we
can
all
get
into
the
flow
of
actually
dealing
with
pr's
and
see
where
the,
where
the
weaknesses
are
happening.
Yeah
in
that
in
that
practice,
so
we
had
one
was.
A
A
C
F
D
D
Just
getting
any
new
notes,
instead
roles
using
notes
in
the
Emerald
country
I'm
just
wondering:
why
do
you
need
extra
node
configuration
if
we
can
reuse
roles
like
extend
this
format
to
be
more
flexible?
So
if
you
have
an
array,
it
will
take
roles.
If
we
have
a
dictionary,
it
takes
some
known,
complicated
configuration
to
read.
D
The
current
status
is
I
mean.
Implementation
of
this
part
is
regarding
that
we
do
not
take
CROs.
We
we
use
nodes
right
so
and
now
and
under
record
notes.
We
treat
all
elements
over
a
as
some
structure
with
specification
of
the
nodes.
You
know
requirements
instead
of
rules
and
someone
commented
that
guys
just
don't
use
rules
if
nodes
record
exists
in
the
config
and
vice-versa.
But
my
question
is:
why
should
we
use
it
should?
D
D
D
Also
also
also
I
have
an
idea
is.
We
can
extend
this
format
to
support.
You
know
enumerations,
so
just
no
sequence,
definition
for
the
roles.
So,
instead
of
declaring
each
note
note
by
note,
we
can
just
say
that
it's
it's
number,
let's
say
hundred
of
notes,
and
we
do
not
need
to
declare
everything
every
note.
Oh
yeah.
D
C
D
C
D
C
Well,
so
if
that
sounds
like
you'd
like
to
be
able
to
to
tell
to
follow
G
that
you
want
50
nodes,
but
how
should
ii
thought
you
decide
what
to
do
with
all
of
them?
You
want
a
way
for
it
to
just
make
a
decision
about,
like
you
know,
alright
it
this
I'm
being
asked
to
use
50
nodes,
so
I'm
gonna
set
up.
C
Think
it's
an
interesting
idea,
but
I
also
think
that
it.
It
feels
like
a
bit
of
a
separate
one,
because
I
think
most
of
the
work
is
going
to
be
gonna,
be
in
in
coming
up
with
a
way
for
thorgy.
You
just
decide
what
proportions
to
use
when
building
a
cluster,
if
it's
just
given
a
block
of
nodes
without
rules.
C
Yeah,
because
so
so,
if
you're,
not
if
you're
not
enumerate
all
of
the
all
of
the
nodes
that
you
want
and
you're
just
saying
give
me
give
me
50,
no
you're,
not
telling
it
because
right
now
for
each
each
node
in
the
in
the
rolls
stanza
there
you're
saying
alright
I
want
one
node
that
runs,
you
know
to
OSDs
and
am
on
and
a
manager
and
then
I
want
another
node.
That
runs,
you
know
cetera.
D
D
C
D
C
B
Because
so
logic
is
actually
based
on
Rawls
I
think.
The
idea
is
that
you
named
the
roll,
you
name
the
note,
and
then
you
ask
to
do
something
of
that
note.
So
I
think
the
only
way
when
we
may
have
to
have
something
without
rolls
if
there
is
a
use
case
when
you
say
I
need
to
do
something
on
that
note.
But
I
don't
need
to
know
what
its
name
yeah.
D
So
I
understand
that
technology
have
to
support
some
operations
and,
on
these
roles,
I'm
saying
about
only
assignment
the
roll
names,
the
test
itself
should
should
to
define
what
what
what
is
gonna
do
with
rolls
to
test
logic
could
actually
be
aware
about.
The
different
amount
of
you
know:
machines
with
particular
rules,
yeah.
C
C
For
example,
you
can,
you
can
just
define
a
new
roll
to
use
for
the
tests
that
you
want
to
run
yeah
so
so,
for
example,
I
have
a
it's
still
in
progress,
but
I
have
a
some
tests
for
set
metrics,
which
is
our
are
monitoring
geese
force
F
for
now
at
least,
and
the
tooth
ology
task
that
I
use
I
think
I
think
the
node
that
it
runs
that
it
deploys.
The
dashboard
on
is
just
called
F.
Metrics
I
could
be
wrong,
but
something
like
that
works.
C
E
A
A
F
Okay,
well,
in
that
case
last
summer,
radish
slaw
started
doing
some
work
on
running
to
thought
asks.
Did
we
relate
to
the
rgw
outside
of
mm
II
start
just
like
in
a
dev
environment?
This.
F
F
I
guess
the
first
thing
I
have
to
say
is
you
know,
kind
of?
Is
there
any
further
work?
You
guys
would
be
interested
in
doing
with
that
I'll
shoot
my
instructions
in
that
PR
as
well,
for
the
small
use
case
that
it
works
for
so
far,
I
think
it
only
works
for
the
rgw
task
or
will
only
works
for
testing
s3
tests
and
testing
the
register.
W
admin
suite
and
the
other
thing
I
wanted
to
ask
was
my
outreach.
F
F
I
wouldn't
mind
taking
some
time
out
of
my
time
and
working
on
that
I,
don't
know
over
the
next
three
or
four
weeks
or
something
like
that.
If
you
guys
think
that
would
be
beneficial
or
a
few
think,
we
would
be
able
to
kinda
help
standardize
on
a
way
to
write
these
tags
or
somewhat
standard
at
standardize
on
a
way
to
write
these
things,
especially
going
forward.
I.
A
B
C
C
C
F
F
A
That
I
think
maybe
in
DEFCON
4
earlier
this
year,
and
just
at
the
time
they
said
they
why
John
did
I
think
it's
based
on
John's
work
from
three
or
four
years
ago.
He
did
for
the
file
system
tests
and
we
found
that
very
useful
in
the
file
system,
because
it
means
we
can
run.
You
know
CEFCU
a
tests
not
in
the
tooth
ology,
which
makes
it
way
easier
to
develop
them,
though
I
would
definitely
vote
for
I
mean
ice.
A
Assuming
this
code
is
good,
which
I
which
I
imagine
is
since
people
are
using
it
then
yeah,
that's
something
we
should
merge
and
I
would
definitely
encourage
people
to
write
against
those
API
is
if
they
can,
because
like
to
theology,
is
our
test
framework
and
we're
always
gonna
need
like
a
scheduling
tool,
presumably
inherit
from
do
theology,
but
one
of
the
problems
that
we
come
across
is
just
like.
You
know,
people
who
aren't
the
size
of
zoos
or
that
have
difficulty
setting
up
technology
and
they
need
to
be
able
to
run
tests.
C
So
we
just
we
need
to
one
thing
that
that
I
would
I
would
want
to
see
here
is:
is
some
kind
of
Docs
somewhere
telling
people
how
to
use
this
yeah.
F
I
said:
hey:
do
me
the
rest
from
how
to
read
me
up
for
this,
but,
oh
definitely
and
we'll
all
do
is
I'm
gonna
throw
the
instructions,
though,
like
quick
instructions
I
have
into
that
PR
as
well?
If
you
want
to
play
around
with
it,
it
should
only
take
you
I,
don't
know
a
couple
minutes
of
your
time.
Yeah.
F
A
All
right,
so
I
got
a
bounce
for
another
meeting.
Now
it
sounds
like
maybe
people
are
done,
but
if
you
guys
want
to
keep
talking
and
then
go
ahead
thanks
a
bunch
guys
thanks
Greg
take.
C
You
slow
down
for
a
sec
I
just
want
to
want
to
make
sure
that
that
we
don't
have
anything
else.
We
want
to
cover
here.
I,
don't
think
I
do
I
just
want
to
leave
the
opportunity
open.
Does
anyone
have
anything
else
they
want
to
cover
in
this
meeting?
I'll.
Take
a
look
at
that
multi-os
PR,
because
I
haven't
looked
at
it
in
like
a
year
and
see
if
I
can
figure
out
where
it's
at
and
I'll
keep
an
eye
out
for.
C
But
it's
basically
it's
a
griffon,
ax
and
Prometheus
set
up
with
some
dashboards
that
monitors
the
set
manager,
Prometheus
module
and
Prometheus
node
exporter
and
I
have
some
tests
into
college
II
that
that,
like
deploy
and
run
some
integration
tests
against
set
metrics,
but
I'm
really
not
sure
how
we
want
to
test
the
step.
Dashboard
going
forward.
I
haven't
thought
a
whole
lot
about
it.
So
I'd
love
to
hear
any
thoughts
that
you.
G
G
C
B
F
C
D
C
D
C
C
Well,
let's
see
it
should
I
I
didn't
have
access
to
any
others
to
do
any
meaningful
testing.
We
had
some
internal
OpenStack
setup
that
was
just
really
quirky,
and
so,
though,
I
I
basically
just
had
OVH
cuz.
That
was
the
one,
the
only
one
that
that
I
had
access
to.
In
theory
it
should
work
for
other
ones.
There
I
think
there's
a
chance.
It
might
need
some
modification
to
support
them,
because
you
know
OpenStack
deployments
can
have
a
lot
of
variants,
but
I
think
it's
possible
I.
Think
it's
worth
trying.
No.
C
I
mean
I
think
it'd
be
worth
trying.
So
the
way,
the
way
this
thing
is
written,
you
can
configure
multiple,
multiple
openstax
in
the
same
deployment
of
two
ecology,
and
they
just
the
way
you
distinguish
between
them.
Is
they
you
you
give
them
each
a
different
name
and
that's
what
the
Machine
type
will
be.
C
So
you
could
name
it
a
Susa
cloud.
You
could
name
one
that
you
can
name
one
OVH
or
you
could
have.
You
know
longer
more
descriptive
names
there.
You
know
they
what
they
do.
Is
they
just
they
key
into
the
configuration
that
you
set.
It'll
get
you
know
like
the
tenant
information
and
credentials
and
whatnot.
So.
C
C
D
It's
probably
it's
only
in
our
branch
and
it's
not
in
the
upstream
okay,
there
Derrick's,
actually
restrictions
for
security
groups.
What
so,
when
you
means
you
deploy
technology
cluster,
oh
no
way
it
creates
security
groups
which
does
not
allow
traffic
for
any
traffic
for
target
nodes
outside
of
the
the
security
group.
D
A
C
C
So
I'll
I
will
come
up
with
with
with
something
to
get.
You
started.
What
the
the
main
thing
you'll
need
is
a
machine
or
a
VM.
That's
that
is
dedicated
to
running
tautology
itself
and
any
other
services.
And
then
all
the
test
nodes
can
be
ephemeral.
The
the
lib
cloud
back-end
does
not
currently
have
a
way
to
bootstrap
tooth
ology
inside
an
ephemeral
node.
You
know
like
the
OpenStack
one
does
it
needs
a
permanent
master
which
I'd
like
to
add
that
feature
to
it,
but
it
just
doesn't.
Have
it
right
now.
C
D
C
D
D
C
Script,
that's
interesting,
I
I
think
that
would
be
really
cool.
So
here
is
the
thorgy
roll
there's
there's
a
separate
paddles
roll
and
a
separate
pool,
pitot
roll
again.
These
might
be
a
bit
out
of
date,
I
I'm,
not
sure
though,
but
this
should
at
least
be
close
to
working,
or
maybe
they
just
fully
work.
I.
Just
can't
tell
you
guys:
I
haven't
deployed
tooth
ology
and
so
long
but
yeah.
This
is
where
it
starts
and
I
think
yeah.
The
idea
of
using
this
instead
of
shell
scripts
is
a
great
one.
D
C
D
B
B
B
D
D
B
B
D
D
B
C
B
C
C
Is
part
of
why
the
workflow
sucks
is
that
you
know
we
have
we
can't
let
we
can't
let
any
major
bugs
into
master,
so
reviews
have
to
be
really
thorough
and
and
yeah
it's.
It's
not
a
whole
lot
of
fun,
an
idea
that
I've
had
in
the
past
that
that
I
still
think
is
a
decent
one
is
if
we
had
two
different
branches.
C
C
D
You
know
one
of
the
issue
that
I'm
working
on
is
I'm
trying
to
work
on
is
that
you
know
if
we
schedules
some
two
toldja
jobs
simultaneously,
let's
say
for
the
same
branch,
then
the
the
jobs
are
mixed
up
with
with
within
them
too
and
like
separate
in
the
at
the
same
time
step
it's,
it's
not
enough.
So
one
of
the
feature
that
I
wanted
to
see
that
do
you
have
custom
job
name.
D
C
D
D
Toronto
merge
at
the
same
at
the
same
time
different
PRS
God
eat,
but
the
target
branch
is
the
same
and
the
same
sweet
it's
running,
then
it's
it
makes
it's
crazy
and
the
jobs
actually
mixed
mixed
in
with
the
same
run,
I.
C
D
D
I
have
a
patch
on
our
branch
like
adding
a
new
parameter
for
told
you
sweet,
like
that's
name
that
will
replace
with
custom
gem
number,
but
it
appears
that
puddles
has
a
design
issue
that
it
it
it
decides
it
bound.
On
the
you
know:
job
name,
court
format,
it
its
passes,
passes
job
name
to
detect
the
time
stamps,
which
is
weird.
B
C
D
B
Just
just
you
know,
quick
comment
like
custom,
job
names
and
custom
run
names.
This
is
like
long
time
dream
of
mine,
because
oh
right
now
right
right
now,
we
we
well
I
mean
once
you
kind
of
you
know,
run
that
stuff,
a
couple
of
million
times
you
get
used
to
like
whatever
timestamp
or
URLs
etc,
and
we
haven't
seen
these
described,
maybe
because
you
have
multiple
people
get
info,
Peter
and
puddles
and
pathology
at.
B
Time
but
generally
just
one
thing
and
and
I
actually
surprised
that
we
don't
hear
from
downstream
folks
screaming
for,
like
you
know,
drop
flesh
test
case
ID,
because
it's
very
important,
but
you
know
the
way
that
we
address
it
right
now
like
when
I
think
of
a
job
I
think
it
has.
A
description
is
equal
to
job
ID
and
description
is
actually
unique.
So
that's
the
issue
on
the
job
level.
B
D
D
G
D
D
C
D
D
E
D
C
C
C
Okay
yeah,
so
that
reminds
me
back
to
where
we
started
whenever
you
get
a
chance,
I'd
love
to
see
some
some
of
the
simple
requests
for
your
changes
to
the
OpenStack
back-end,
because
I
think
that
I
think
we
should
yet
we
should
try
and
and
get
you
merged
into
master
I.
Think
it'll
make
a
lot
of
the
other
things
that
we're
talking
about
easier
to
collaborate
on.
D
B
B
C
B
Quick
idea
and
I
discussed
it
with
Josh
a
little
and
just
I.
Think
I
saw
a
couple
of
yards
from
you
yesterday
yesterday
and
you
know
I
decided
to
check
with
you
so
I'm
working
on
sound
like
upgrade
tests
and
in
particular,
like
we
actually
missed
what
one
bug
that
was
caught
by
downstream
folks
and
it
was
in
point-to-point
release.
So
I
want
to
spend
a
little
bit
more
time,
actually
doing
more
thorough
job
on
mimic
point-to-point.
So
essentially,
I
want
to
do
every
point.
I
don't
want
to
do
like
you
know.
B
F
B
Know
when
I
love
them
I
love
them
all.
It's
just
like
hours
of
routes
and
I
checked
with
Josh
and
I
think
on
radius.
It's
not
a
big
deal.
I
can
just
specify
for
bench
bench,
whatever
bench
range
when
something
like
time
of
meanest
around
so
I
can
say
five
minutes.
So
I
wonder
if
similar
scenario
can
be
done
for
our
GD
rgw
type
of
workloads.
B
B
Know
I
I
will
point
you
like
today
like
like,
like
examples
too
sweet,
but
the
idea
is
general,
so
idea
is
general
I
can
when
you
think
about
it,
like
you
know,
when
you're
on
upgrade
so
I
can,
let's
say
we're
on
our
GW
suite
and
I
have
the
and
that
you
guys
I
want
to
do
like
you
know
in
depths
like
you
know,
zillions
of
scenarios
etc.
Yeah.
F
B
You
run
upgrades
the
question
is:
do
we
really
want
to
do
that
exhaustive
type
of
testing
core,
especially
what
we
want
to
do?
We
want
to
upgrade
it
from
one
version
to
another,
and
just
kind
of
you
know
very
few
of
them
and
if
we
can
do
that
and
we
can
cut
down
on
time
run
time,
that's
significant,
because
you
know
we
can
actually
run
more
tests
and
actually
do
like
more
cycles
as
opposed
to
sit
and
run,
like
particular
workloads
that
already
being
tested
by,
like
you,
know,
major
suite
itself.
F
B
Typically,
what
we
do
like
you
know
you
can
take
any
example
of
a
grace
right.
You
know
say
you
know.
I
gave
you
example
about
point,
a
point
which
is
like
you
know
like
point
to
point
right
now.
You
know
it
would
be
the
luminous.
Only
so
luminous
point
one
released
another
one
if
you,
if
you
look
at
at
Suites,
likes
a
luminous
X
so
you're
on
luminous
and
then
you
are
created
to
mimic,
for
example,
or
to
master.
B
B
Wonder
I
wonder
if
if
we
can
change
that
or
maybe
we
can
like
either,
you
might
have
some
parameters
in
your
workloads
that
allows
me
to
specify,
like
on
work
rate
range
for
Raiders,
so
like
specify
time
or
if
it
doesn't
exist,
you
can
either
consider
aging
this
functionality
or
you
can
consider
saying.
Okay,
I'll
give
you
like
a
stripped
down
version
of
my
whatever
ss3
tests,
and
you
know
if
you
call
that,
instead
of
full
full
full
workload
that
will
address
what
you
want.
F
I,
don't
think,
there's
a
specific
right
guy
I
think
you
know
it
would
be
at
least
the
group
effort
of
two
to
three
people
to
figure
out
what
those,
what
those
slim
down
subset
of
tests
between
the
upgrades
would
be
and
how
to
do
that.
But
I
think
it
would
be
a
pretty
good
idea.
If
you
just
could
you
come
to
the
rgw
stand
up,
say
on
Friday
sure
what
time
is
it
so
the
rgw
stand
up
on
Friday
is.
B
F
B
It's
not
urgent,
so
it's
not
something
that
you
know.
We
dropped
that
do
it
right
now,
so
let's
try
Friday.
If
not
I
will
try
to
come
next
time,
whatever
you
feel
it
and
we
can
talk
about
it
and
I
want
to
do
the
same
for
our
BG
and
then,
if
we
can
get
it
done,
I
think
it
will
I
think
it'll
be
positive
overall,
because
it'll
save
us
time,
yeah
a.
F
B
F
It's
funny
you
mention
the
s3
testing,
the
internet
mentor
egg
she's
she's
about
to
put
two
more
repos
of
s3
tests
in
two
different
languages,
a
little
like
at
some
more
time
or
her
name
is
Antoinette
she's
from
Switzerland
she's,
just
working
for
us
for
the
summer.
She's.
B
Out,
okay,
very
pretty
cool
I
think
that
whatever
we
discuss
in
busy
right
now,
I
think
it's
a
natural
problem,
because
I
think
that
people
they
put
tests
and
and
work
clothes
in
and
we
use
them
and,
like
you
know,
we
know,
we
kind
of
you
know,
came
to
this
realization
right
now
like
when
and
right
now
is
a
little
bit
like
you
know,
I
wouldn't
say
slow,
but
you
know
people.
You
know
like
like
thinking
about
like
stuff
like
they're
actively,
because
more.