►
From YouTube: 2021-05-27 Kubernetes SIG Scalability Meeting
Description
Agenda and meeting notes - https://docs.google.com/document/d/1hEpf25qifVWztaeZPFmjNiJvPo-5JX1z0LSvvVY5G2g/edit?usp=sharing_eil&ts=5d1e2a5b
A
Okay,
hello,
everyone:
this
is
six
scalability
meeting
my
27th
2021.
A
B
Yeah,
I
have
a
few
questions
on
the
large-scale
cluster
configurations
so
recently
I'm
a
low-test
like
10k
nodes,
cluster
with
400k
plus
and
we're
trying
to
like,
because
we
use
very
powerful
hardware
we're
trying
to
understand
any
like
parameter
training
did
for
these
scale
clusters,
especially
on
the
api
server
side.
A
So,
do
you
know
that,
like
the
api
server
itself
is
your
through
boot
and
not
the
other
components.
B
I
use
more
clients,
and
I
noticed
I
think
I
use
three
api
servers
and
the
qps
is
around.
I
think
six
thousand
yeah
per
second,
I
think,
even
I
add
more
clients,
I
don't
see
the
like.
This
drupal
go
higher,
so
I
thought
probably
some
like
like
issues
on
the
api
server.
I
don't
use
the
api
like
fahrenheit
and
the
priority,
and
I
changed
the
mutation
in-flight
mutation
request
to
higher
number,
but
currently
it
stayed
at
that
level.
B
Right,
so
I'm
I'm
thinking
if
any
other
things
we
can
do.
Most
of
my
requests
are
like
crud
requests,
so
I
think
it
hit
almost
keep
the
limit
yeah.
A
B
Seems,
okay,
but
the
the
slow
query
and
pending
proposal
numbers
goes
high
and
the
overall
end-to-end
latency
increased
as
well.
So
I
think
because
I
most
of
my
requests
are
right,
like
part
patch
and
create
a
I
think,
etc
could
be
a
bottleneck
as
well,
and
we
noticed
like
actually
the
performance
on
edc
3.4
has
worse
performance
than
3.3.
B
We
submit
few
patches
and
on
the
concurrent
buffer
optimization
after
that,
I
think
it
close
to
the
three
to
three
performance.
B
So
we
are
not
sure
like.
What's
the
next
steps,
yeah.
A
Yeah,
because
I
I'm
surprised
it
shouldn't
be
the
limit
like
there
are
a
couple
of
things
where,
like
api
server
could
be
limiting.
One
thing
that
comes
to
my
mind
is
like
the
throughput
of
watch
cache
in
particular.
A
B
Yeah
we
haven't
tried
like
five
api
server
yet
and
yeah.
So
that's
the
like
to
that's
in
my
to-do
list.
I
can
have
a
try,
yeah,
so
watch
cash.
Could
you
explain
the
watch
cash
limit.
A
I
don't
know
the
exact
numbers
from
the
top
of
my
head,
but
the
problem
is
that
there
are
certain
things
that
are
happening
in
serial.
I
mean
we
are
dispatching
particular
events
in
serial,
so
we
need
to
finish
dispatching
the
previous
previous
event
before
we
start
dispatching
the
next
event.
A
Right
I
mean
not
the
whole
dispatching,
but
like
not
the
whole
dispatching
in
a
sense
like
that,
all
the
logic,
but
there
are
certain
smaller
things
like
putting
it
into
cues
that
are
or
to
channels
that
are
later
than
being
processed
by
like
per
watch,
go
routines,
so
it
used
to
be
significant
problem
in
the
past.
We
did
a
couple
optimizations,
I
don't
know
two
years
ago
or
so
so
it
wasn't.
A
It
wasn't
our
immediate
bottleneck,
but
I
don't
think
we
are
especially
if
you
are
trying
to
achieve
higher
throughput.
I
I
don't
think
we
are.
We
are
really
close
from
from
the
throughput
from
the
current
limit
the
current
throughput
limit,
so
especially
if
you,
if
you
start
observing
things
like
but
startup
time
becoming
longer
and
stuff
like
that,
that
would
be
my
first
guess
that,
like
watch
cash,
throughput
is,
is
becoming
a
a
bottleneck
here.
B
Okay,
okay
sounds
good
yeah.
There.
A
B
Okay,
yeah,
let
me
try
some
like
low-hanging
fruits,
like
adding
more
api
servers.
First
yeah,
I
can
report
the
status
bank
in
the
next
meeting.
B
B
Another
question
I
see
there
was
some
discussion
on
the
compar
compression
between
api,
summary
and
atcd.
So
I
checked
the
code
since
we
use
the
some
encoder
and
just
the
marshall,
to
convert
the
object
to
the
device
and
I'm
I'm
trying
to
see
if
we
can
use
some
compression
algorithm
to
to
optimize
the
size
we
send
to
edcd.
B
Do
you
think
that's
a
like
a
reasonable
authorization.
A
Well,
I
think
it
depends
where
exactly
the
bottlenecks
are,
because
compression
also
isn't
for
free,
so
yeah
I
I
mean
it
probably
depends
it
may
like
it.
It
depends
really
on
on,
like
what
workloads
do
you
have
and
stuff
like
that,
like
how
your
usage
patterns
look
like,
it
may
potentially
be
worth
trying.
B
Okay,
yeah,
so
one
thing
another
similar
issue
is,
I
remember,
like
I
think,
two
years
ago,
there's
a
patch
to
support
the
gzip
between
the
client
and
api
server,
and
at
that
moment
I
think
we
only
have
we
only
allowed
the
compressions
on,
I
think
payload
over
16
kb.
As
I
remember,
do
you
think
it's
better
to
twin
that
number.
At
this
moment
we
can,
since
we
have
like
powerful
machines.
Probably
like
resource
is
not
a
limitation.
B
We
can
change
the
number
to
lower
lower
number
and
see
if,
like
small
requests,
can
benefit
from
that.
A
Yeah
I
mean
you
can
always
try.
We,
we
didn't
really
see
significant
benefits
when
trying
that,
for
I
mean
it
was
using
way
more
resources,
I
mean
cpu
mostly
and
like
the
the
gains
weren't
significant,
so
we
didn't
do
that,
but
yeah.
As
I
said,
it
really
depends
like
on
your
usage
patterns.
So,
okay.