►
From YouTube: 2017-JUN-15 :: Ceph Performance Weekly
Description
Weekly collaboration call of all community members working on Ceph performance.
http://ceph.com/performance
For full notes and video recording archive visit:
http://pad.ceph.com/p/performance_weekly
A
B
B
B
I
just
won
this
in
testing
right
now
that
refused
a
scrub
job
with
the
higher
priority
sounds
good
when
it
in
QA
there's
about
grind
error,
so
I'm
you're
splitting
that,
apart
to
see
based
at
PR,
the
other
one
that
was
causing
a
problem
but
presumably
that'll,
go
in
I'm
still
waiting
on
sort
of
some
consensus
on
the
up
light
up
commit
thing,
but
it's
going
to
be
post
luminous,
there's
another
related
PR
that
just
reopened
today.
That
was
changing
the
order
of
the
callbacks
in
file
store
and
we
had
before.
B
Like
that's,
it's
really
for
performance,
stuff
you're.
The
only
other
thing
with
worth
mentioning
is
that
we
had
discussion
last
week
about
boost
or
tuning
cash
sizes,
and
so
on
out
of
that
there's
code
that
went
in
that
lets.
You
configure
the
rocks
to
be
cache
sizes
in
terms
of
the
overall
blues
or
cast
size.
B
So
it's
just
a
ratio
that
you
can
figure
Marcus
sailing
around
at
those
ratios
to
figure
out
what
values
actually
makes
sense,
but
because
they're
so
there's
just
one
tunable,
that's
the
amount
of
memory
that
boosts
or
takes
to
set
I'm.
Just
good
I
think
you
still
plan
on
bumping
that
up
to
three
or
four
gigs
I'm
believing
them
ant.
C
Okay,
so
the
the
gist
of
it
is
that
we
ran
through
a
bunch
of
tests
and
relays.
We
found
out
wait
a
bug,
so
I
know
that
that's
are
valid.
Unfortunately
at
least
know
the
ones
that
Adam
I,
where
we
were
trying
to
dedicate
a
lot
of
the
cash
to
to
rocks
levy
rather
than
the
fees
it
was
for
cash,
so
that
scam,
where
it
is
right
now
is
now
we
got
that
pic.
So
we
need
to
figure
out.
C
Do
we
see
significant
benefits
by
having
Roxie
be
store
things
in
its
cash
that
are
already
encoded
and
keep
the
encoding
overhead,
or
are
we
better
off
letting
blue
store
store
things
in
its
own
cache
before
encoding,
but
then
potentially
it's
bigger
and
some
cash
fluff
stuff
and
there's
all
she
also
on
third
hand:
a
compressed
cash
from
rocks
TV
that
we
could
use.
That
would
further
compress
things,
but
then
now
we
do
in
the
encoding
overhead.
Then
we
also
have
the
week
at
the
the
compression
overhead.
C
So
it's
just
kind
of
weighing
out
where
the
the
right
balance
points
are
and
all
of
this
so
that's
happening
day
now,
but
please
be
still
in
know.
C
B
C
C
The
the
LDPR
from
o
to
2015
and
it's
sort
of
working
that
it's
not
working
that
well
yet
I'm,
not
sure.
If
is
actually
going
to
be
very
fast
in
the
end,
but
I
thought
it
didn't
so
much.
He
got
really
a
good
good
evaluation.
Last
time
we
tried
to
do
it
so
at
the
very
least,
maybe
I
can
see
it,
how
bad
it
really
is
and
whether
or
not
it
is
worth
pursuing.
So
there's
that
as
sort
of
a
side
project
so
we'll
see
how
how
quickly
any
part
or
something
gets
made.
That's
it.