►
From YouTube: Node Operator Roundtable - May 17th, 2023
Description
Join us as we explore revolutionary advancements in blockchain technology, discussing automated peer selection, dynamic transaction handling, immediate analytics, and effective node synchronization for improved system efficiency and transparency.
00:00:00 Start
00:00:36 Leap software Development update
00:02:53 EVM node operations
00:07:28 EVM RAM usage
00:11:24 P2P Improvements
00:31:20 Resyncing nodes from snapshots
01:03:34 End
The EOS Network Foundation coordinates financial and non-financial support to encourage the growth and development of the EOS Network. We’re harnessing the power of decentralization to chart a coordinated future for the EOS Network as a force for positive global change.
// EOS Network Foundation //
https://www.eosnetwork.com
https://twitter.com/EosNFoundation
https://www.linkedin.com/company/eos-network-foundation
https://t.me/EOSNetworkFoundation
https://discord.gg/eos-network
info@eosnetwork.com
#EOS #Web3
A
Welcome
everybody
another
round
table
it's
May
17th
today
on
the
agenda,
we're
going
to
do
a
quick
update
on
or
looking
to
have
actually
this
what
he's
dug
into
on
a
evm
nodes
for
Eos
and
we'll
talk
on
peer-to-peer
improvements,
but
before
we
do
that
I'll
hand
it
over
to
Brian
to
talk
about
some
update,
not
enough
State
updates
about
leap,
4.0
or
4.1
or
whatever
it
is
I'll
give
it
to
you.
You
know
what
you
thought.
B
Yeah,
absolutely
so
leap
4.0.1
will
be
a
patch
release,
delete
fixing
a
few
bugs.
We
there's
two
outstanding
issues,
one
of
which
is
confirmed
in
PR,
the
other
one
I'm
I'm
working
on
getting
the
status
update
on,
but
yeah,
so
so
very
close
to
completion
on
on
that
new
date
set
for
it
just
yet,
but
it's
coming
it's
coming
around
the
mountain
and
and
then
on
the
CDT
side,
we're
working
on
4.0,
rc2
and
then
the
thing
I
have
the
most.
B
So
it's
good.
We
have
a
patch
release
to
1.1.
Coming
out.
There
was
a
conflict
with
boat,
camel,
goon
and
I
have
a
link
which
I'll
share
here
in
the
chat.
If
anybody's
interested
in
more
details
on
this,
they
can
look
in
and
look
at
this
issue
to
see
you
know
what
what
that's
about,
but
the
patch
rate
release
will
resolve
the
name,
conflicts
that
we
have
there
on
Linux
with
ocamel
Dune.
B
B
B
D
D
So
what
I
learned
is
actually
it's
not
before
I
said
it
was
like
one
node
that
would
handle
reading
transactions,
and
you
know
giving
back
responding
to
all
the
read-only
kind
of
calls,
as
well
as
doing
you
know,
handling
the
East
Central
transaction
to
actually
grab
that
metamask
or
a
good
transaction
into
a,
and
you
feel
like
an
normal
antelope
transaction
and
push
that
on
scene
and
pay
for
the
costs
and
so
on
and
so
forth.
So
I
predictable.
D
What
it's
done
is
the
Json
RPC
that
the
metamask
users
connect
to
is
actually
a
proxy
already
and
it
proxies
all
requests
that,
except
to
so
outside
sending
the
raw
transaction
and
getting
gas
it
routes
all
these
two,
what
they
call
the
evm
RPC.
So
there's
like
three
parts
to
it.
There's
a
dbm
RPC
node
there
is
the
evm
node,
which
is
execution
node,
and
there
is
an
evmtx
wrapper,
which
is
basically
what
we
were
calling
the
miner
Burana
Miner
just
answer
that
first
question
throughout
a
minor
is
quite
straightforward:
it's
a
stateless
node!
D
All
it
needs
to
do
is
to
be
able
to
fetch
one
row,
basically,
the
one
with
the
gas
from
the
evm
contract,
as
well
as
be
able
to
take
that
rlp
transaction
and
pay
for
CPU
and
net
and
push
it
to
the
blockchain.
So,
in
terms
of
running
lab
that
uses
almost
no
resources,
all
you
need
to
do
is
have
access
to
an
API
endpoint
that
can
be
even
remote
and
very
easy
to
do.
The
evm,
RPC
node
is
also
very
light.
D
It's
just
really
a
basic
Json,
RNC
server,
the
swiftness
C
plus
plus,
but
but
it's
usually
coupled
with
this
evm
node
and
that's
the
one
that
is
a
fork
of
like
silkworm,
just
like
a
a
client
for
ethereum,
written
in
C,
plus,
plus
and
but
modified
to
basically
use
the
stream
Bishop
blocks
and
be
able
to
form
its
own
database.
To
respond
to.
You
know
that
transaction
by
Ash
and,
like
all
these
other
call.
D
Calls
that
you
would
make
just
RTC
requests
when
you're
interacting
with
smart
contracts,
not
just
for
using
metamath.
So
the
heavy
part
here
is
really
just
that
evm
node
and
that's
the
part
that
would
really
need
scaling,
because
others
are
I,
think
anybody
can
run
it
very
easily
and
another
core
difference.
D
What
it's
doing
is
actually
socializing
the
the
ram
costs
as
far
as
I
can
tell,
and
then
from
the
fees
that
get
incurred
by
these
evm
users.
Part
of
the
FC
goes
to
in
back
the
the
minor
so
to
speak,
and
another
part
is
to
be
used
to
pay
for
that.
Rand
additional
cost
that
that
vvn
incurs
and
then
potentially
the
last
whatever
remains,
could
be
burnt
or
moved
to
Rex
or
or
whatever
else
I
was
just
wondering.
B
E
You
know
the
ownership
of
all
that
Ram,
then,
is
just
owned
by
the
the
contract,
which
makes
it
you
know
simpler
to
to
manage
a
long-term.
You
know
in
terms
of
like,
if,
if
you
you
know
at
some
point,
want
to
remove
bad
or
or
a
compacted
in
some
way,
or
you
know,
do
some
some
changes
along
that
line
to
certainly
supplies
of
that
plus
I
think
it's
supplies.
These
are
experience
and
that
you
know
you
only
need
to
worry
about
a
CPU
in
that,
and
you
don't
have
to
worry
about.
E
D
Not
it's
definitely
simpler
and
easier
and
more
convenient
for
the
user,
better
experience,
but
doesn't
that
open
an
attack
Vector
where
somebody
can
just
abuse
by
you
know
creating
things
that
keep
using
more
RAM,
since
this
is
only
the
smart
contract
that
pays
for
it,
and
then
the
cost
is
socialized
to
every
evm
user.
With
this
transaction.
E
D
D
Yeah
but
the
gas
cost
is
paid
by
everybody,
so
it
means
like,
if
I'm,
one
user
that
is
going
and
making
many
contracts
that
are,
you
know,
abusing
RAM
or
using
a
lot
of
ram.
The
gas
price
will
go
up
for
everybody,
while
I'm
thinking,
you
know
the
same
as
everybody
else,
but
I'm
not
like
you
know
paying
for
my
Ram
for
the
speakers
Mr,
whatever
additional
RAM
I'm
making
the.
If
you
have
news
that
course
is
socialized
to
everybody,
because
that
is
is
get
gas
call
that's
coming
to
a
Smart
contract.
D
It's
calling
a
conflict
table
with
no
parameters
of
that
user
or
anything
related
to
that
user.
So,
just
like
it's
just
to
have
some
type
of
mitigation
of
I,
don't
know
how
maybe
in
a
smart
fun
fact
and
limit
how
much
you
know.
I,
don't
know
like
a
user
can
push
based
on
his
history
or
it's
just
like
something
to
something.
I
thought
about
that.
I
thought
might
be
important.
B
Yeah
yeah,
it's
definitely
interesting,
as
I've
mentioned
before
you
know,
I'm
I
haven't
been
to
to
date
heavily
involved
in
the
evm
project,
so
I'm
I'm
also
coming
up
to
speed
yeah,
certainly
an
interesting
topic
with
regards
to
how
the
Ram
works
and
I'll
look
into
that
and
share
with
you.
If
you
want
to
direct
message
me
on
on
Zoom
here,
your
your
telegram
handle
I'll
I'll
follow
up
with
you.
B
Awesome,
okay,
so
the
the
other
topic
I
had
in
mind
today.
This
is
not
a.
This
is
not
a
proposal.
This
is
sort
of
information
gathering,
so
something
that
the
antelope
Coalition
has
expressed
interest
in
is
improvements
to
P,
to
P,
there's
kind
of
two
pieces
of
this
there's
pure
pure
Discovery,
and
then
there's
sort
of
a
broader
segment
of
of
improvements
just
to
kind
of
scope.
This
conversation
we're
not
talking
about
Discovery
improvements
in
the
questions
that
I
want
to
kind
of
throw
out
to
the
group.
F
B
Get
feedback
on
here
we're
talking
about
peer-to-peer
improvements
more
broadly
in
the
event
that
your
feedback
is
about.
Discovery
don't
hesitate
to
share
it,
but
but
I
will
try
to
I
will
try
to
stem
any
in-depth
conversation
in
that
specific
area,
simply
because
it's
not
the
focus
of
what
I'm
trying
to
learn.
The
goal
of
segments
like
this
is
really
just
to
get
Community
feedback.
From
specifically
the
node
operator
perspective
on
a
certain
Topic.
In
this
case,
peer-to-peer.
E
B
And
just
for
you
know
to
understand
what
we're
going
to
do
with
this
feedback.
Next
steps
is,
is
developing
a
proposal
for
specifically
what
improvements
we
will
make
and
how
right
what
basically
refining
what
the
goals
of
the
improvements
are.
There
are
some
stated
in
the
current
P2P
improvements
RFP,
but
we
want
to
sort
of
refine
this
and
be
sure
that
those
are
in
fact
the
most
important
goals
to
to
to
address,
and
then,
of
course
you
know
another
part
of
The
Proposal
is
the
solution
aspect.
B
Some
of
them
won't
necessarily
sound
like
feedback
questions,
but
I'm
trying
to
get
like
experiential
feedback
for
these
questions
so
yeah
first,
first
I
just
want
to
ask
you
know
to
the
group.
Anybody
who
has
experience
with
this,
which
I
I
would
expect
would
be
most
or
all
node
operators
in
what
struggles
or
pains
that
you
currently
are
experiencing
with
hearing
that
aren't,
aren't
related
to
sinking
or
discovery.
C
With
it
first
came
up
right
when
we
it
was
on
the
wish
list.
It
was
a
wish
list
and
I'll
put
my
head
up
and
said:
well,
it'll
be
great
if
we
had
automatic
Discovery
and
it
would
select
the
most
appropriate
peer
building,
some
kind
of
state
cable,
they
knew
what
the
latency
was,
and
you
know
what
was
available
on
the
other
side.
C
You
know
whether
they
had
well
the
full
block
slogan
Etc,
the
other
kind
of
layer
to
it
was
was
you
know,
selecting
the
most
performance
peer
and
sending
transactions
to
the
most
performant
peer
and
I
think
it.
We
got
a
little
bit
made
last
time
where
you
know
one
of
them
is
just
sinking
and
the
other
one
was.
Actually.
C
You
know
flying
the
right
direction
for
transactions,
which
was
you
know,
a
lot
of
the
stuff
that
we
try
to
solve
manually
on
wax,
where
they've
got
really
really
busy
at
some
point,
when
we
started
doing
nodes
that
were
only
blocks,
nodes
and
nodes
that
were
only
transaction
nodes
to
ensure
and
in
those
ones
were
appeared
to
like
trainees,
yeah
so
I
mean
that's,
that's
the
kind
of
other
aspect
to
it.
You.
B
C
Well,
you
want
to
ensure
that
transactions
get
to
the
right
place
with
the
least
amount
of
latency,
and
then
you
also
want
to
ensure
that
you
are
sinking
I
know.
You
say
this
is
not
sinking
of
it.
You
want
to
ensure
that
you
are
sinking
from
the
the
latest
lowest
light
needs
peer
that
you
have
available
so.
C
C
Yeah
I,
don't
know
it's
it's
all
about
just
ensuring
that
the
blood
producer,
that's
finding
that
signing
those
transactions
gets
that
quickest
way
possible
and,
as
as
operators,
tend
to
create
our
own
little
back
framework
very
very
manually
at
this
point,
in
order
to
ensure
that
we
have
the
lowest
kind
of
latent
lengths
and
the
software
itself
doesn't
automatically
select
the
the
best
pure.
So
we
kind
of
manually
force
it
to
do
that
kind
of
stuff.
So.
B
C
Well,
it's
being
that,
it's
been
a
little
bit:
hey,
Marco
and
Matthew
could
probably
jump
in
here.
We
we
were
hovering
it
on
on
the
wax
Network,
because
people
were
exploiting
some
contracts
or
getting
an
extra.
You
know
exceptional
amounts
of
traffic,
so
it
kind
of
hits
the
Tipping
Point.
That
was
a
few
months
ago.
50
months,
we
haven't
had
anything
like
that
all
late,
but
are
you
looking
for
like
a
user
story
as
to
what
what
did
the
user
experience.
C
What
was
happening
a
lot?
It
was
because
things
were
so
jammed
up
and
kind
of
just
getting
wherever
transactions
were
getting
queued
and
expiring,
so
they
would
hit
the
chain
they
would
get
an
API
that
says
no
problem,
and
then
they
go
into
the
unapplied
queue.
They'd
get
stacked
out
through
a
couple
of
rounds
and
time
out.
C
You
know
that
that
age-old
issue
of
well
the
API
said
it's
fine,
it's
because
it
kind
of
was
yeah,
but
once
it
got
down
the
chain
and
and
started
meshing
into
the
the
sea,
it
really
started
Domino
affecting
the
user
experience
because
they
thought
it
would
work.
It
just
disappeared.
So
and
then
we
even
had
layer
twos
that
freaked
out
because
they
weren't
you
know,
handling,
Forks
and
and
bad
data
is,
is
important
anyway.
B
C
You
don't
know
which
order
they're
going
in
this
I
think
I
think
what
we
ended
up
doing
was
lubricating
the
back
end
manually
to
make
sure
everything
was
manually
lowest
lakings
with
the
best
peers
with
our
neighbors
in
order
to
mitigate
that
issue,
so
that
issue
ultimately
ended
up
being
the
way
a
specific
smart
contract
was
used,
that
it
gave
us
a
exceptional
amount
of
traffic,
so
it
just
re-edits
it.
You
know
that
wasn't
specifically
the
issue
we
just.
C
We
solved
it
by
being
a
little
bit
more
efficient
on
the
back
end
manual
and
kind
of
got
told
it.
What
would
happen
is
you'd
get
a
BP
unintentionally.
That
would
peer
in
one
of
these
not
filtered
and
kind
of
optimized
peers
and
all
of
a
sudden
I
mean
if
I'm,
allowing
a
PRN
they're
shoving
it
into
my
throat.
C
So
you
know
it's
I,
guess
from
a
peer-to-peer
aspect,
that's
kind
of
we
call
it
the
an
effective
virus
when
they
connect
oh
shit's,
about
to
start
flying
both
ways
and
and
yeah
to
build
this
kind
of
no,
no,
no,
no,
no,
let
it
let
it
be
after
subjective
billing.
After
you
know,
Blacklist
after
whatever
you
get
one
person
that
happens
to
get
one
of
those,
no
sneaks
back
in
this
whole
kind
of
private
Network
would
re-ingest
all
the
nasties
and
kind
of
circumvent
the
back
side
of
it.
That
makes
sense.
E
Yeah
I
was
taking
notes
what
I
just
wrote
down
there
was
so
I
I
heard.
You
know
you
got
a
number
of
peers
and
you're
getting
transactions
from
those
peers,
so
I
think
we
definitely
want
to
be
able
to
track
and
report
number
of
transactions
per
peer,
but
I'm
wondering
if
we,
if
you
could
see
a
desire
for
more
than
that
in
terms
of
like
being
able
to
configure
and
or
limit
a
transactions
from
Individual
peers.
E
C
Really
kind
of
contain
to
the
unapplied
Q
kind
of
it
and
and
these
types
of
scenarios,
so
it's
it's
not
really
the
transactions
that
are
getting
relayed
in
a
block.
It's
really
whatever
that
mesh
is
of
here.
I've
got
some
trash
to
pour
into
the
the
55
gallon
bucket,
and
then
it
kind
of
just
starts
moving
as
this
coagulated
and
and
whatever
order
it
puts
them
in,
and
some
people
were
able
to
figure
out
how
to
kind
of
monkey
with
that.
C
That
so
I
think
that,
in
your
mind,
or
at
least
in
the
code,
Maybe
tracking
them
and
looking
at
them
are
two
completely
aspects.
One
is
the
number
of
transactions
that
would
be
coming
in
that
are
like
in
a
block.
The
other
is
this
unapplied
key,
and
that's
where
it
really.
This
pure
peer
mechanism
seems
to
kind
of
easily
jam
up,
because
it's
that
kind
of
central
choking.
C
You
know
there's
there
seems
to
be
a
lot
of
things
that
rely.
I
mean
I
mean
if
you
think
about
it,
multiple
nodes
around,
like
hundreds,
thousands
of
nodes
around
the
world,
everybody
trying
to
sync
up
as
efficiently
as
possible.
Well,
that's
what
we
as
the
users
that
would
rewind
for
The
Operators!
That's
what
we
want.
It's
not
necessarily
certain
that
the
software
itself
is
trying
to
organize
itself
into
the
most
Optimum
way.
There
needs
to
be
some
kind
of
state
between
the
nodes
that
you
connect
to.
C
C
Well,
multiple
piers
and
they're,
trying
to
get
out
of
my
house
all
coming
just
with
one
RP
address
and
they're
hitting
nodes
and
there's
multiple
nodes
out
during
my
testing
and
editing
nodes
that
only
allow
two
RP
addresses
by
default
for
connection.
My
nodes
don't
know
this,
so
they
try
and
connect
to
that
same
node.
That's
only
allowing
tune.
Let's
say:
I've
got
like
four
or
five
nodes
running
here
on
the
same
network.
All
coming
from
the
same
IP
address
it's
in
the
same
period
it'll
actually
store
my
node
in
my
house.
C
My
my
cast
node
does
not
know
that
it's
because
I'm
coming
from
an
IP
address,
so
my
node
will
actually
just
sit
there,
not
sync
anymore
until
you
know,
and
Kevin
I'm
not
sure
what
it
is
that
makes
it
go
to
the
next
Pier
to
see
if
I
can
use
it.
You
know,
obviously,
sometimes
it's
hitting
multiple
peers
that
you
know
because
I've
configured
everything
the
same
and
it's
getting
multiple
duplicate
addresses
trying
to
connect
to
me.
C
C
Firewall
on
premise:
that
is
a
lesson.
I've
learned
the
hard
way,
an
idea,
I've
dedicated
external
peers
that
only
have
one
connection
to
them.
If
you
try
to
yeah
I
mean
I
got
flagged
all
the
time.
Of
course,
it's
much
better
at
trying
to
go
to
other
peers
now,
instead
of
going
in
a
coma
but
sounds
like
you
still.
A
C
C
Kevin
used
the.
What
is
that
not
the
host
name,
the
agent
name,
because
we
not
everybody,
does
it
not
everybody
realizes
exactly
what
it
is,
but
once
I
realize
what
it
does.
You
know
and
kind
of
tells
me
who
it
is
or
tells
me
the
incorrect
actress
if
I
put
it
in
wrong
but
versus
the
actual
Network
layer,
IP
that
is
coming
from.
C
If
in
the
config
you
denote
that
it's
a
different
agent
name,
you'd
know:
oh
nope
you're
trying
you
are
the
same
connection
but
you're
actually
saying
you're
a
different
agent
name
in
theory,
the
same
node
shouldn't
be
able
to
respond
to
different
agent
names,
but
can
it
uniquely
enforce
the
agent
name
is
well
I
mean
that's
just
the
menu
I
mean
some
people.
Have
it
all.
C
I
mean
they're
using
it,
it
is
coming
from
the
same.
Ip
I
mean
from
a
from
an
IP
firewall
level.
Yeah
you
got
multiple
connections
problem
is,
is
not
o
f,
isn't
intelligent
enough
to
say
well,
two
nodes
may
be
connecting
from
the
same
IP,
but
it
on
your
nodes.
You
properly
identify
a
unique
agent
name,
which
is
what
it's
using
to
tell
the
peer-to-peer.
C
It
tells
it
that
anyway,
that's
how
you
know
it's
who
it
is,
and
people
have
it
wrong
all
the
time.
But
that's
why
I
ask
is
that
if
it's
Unique
wrong
or
not
that
one
and
if
we
and
if
we
do
use
that,
then
we
need
to
ensure
that
everybody's
well
aware
of
making
a
State
node
by
node
agent
there
is
that
not
something
you
would
do
normally,
no
I'm
gonna
jump
your
sphere,
your
main
net
agent,
Nev.
B
So
we're
we're
at
we're
a
little
past.
Half
over
I
want
to
cover
a
bit
more
ground.
Interesting
conversation.
You
know,
I've
got
some
notes.
Kevin's
got
some
notes,
but
I
want
to
cover
some
more
ground.
Some
are
there
any
other
sort
of
struggles
or
pains
around
peering
unrelated
to
what
we've
just
been
discussing.
B
C
C
We
keep
snapshots
from
all
all
about
trains
that
we
support
and
there's
always
something
going
on,
whether
there's
more
memory
more
disk,
whatever
you
know,
wherever,
wherever
the
states
database
had
to
get
replaced,
whether
that's
because
the
state
database
is
running
inside
Ram
or
it's
because
there
was
an
upgrade
that
required.
You
know
version
three
to
version
four,
whatever
that
needed
a
a
snapshot.
You
know
this
there's
a
lot
a
lot
of
it
to
give
you
that
contact
already
oh,
go
ahead.
C
Mike
I
was
gonna
say
that
camp
that
fast,
that
he
keeps
mentioning,
is
it's
kind
of
a
hot
rod.
Hack
That
note
OS
is
brutal
on
the
state
database,
just
especially
on
the
bigger
drives.
It's
massive
whatever
so
span.
I
think
is
the
first
one
to
kind
of
do
it,
but
you
basically
load
it
up
as
a
temp
at
fast
Drive
in
Ram
and
let
it
beat
the
snot
out
of
ram
faster
and
obviously
what's
abusive
on
your
drives.
The
end
result
is
that
it
doesn't
persist
on.
A
A
C
And
the
quickest
way
to
rebuild
the
state
database
is
with
the
snapshot.
So
it's
not
kind
of
a
normal
thing
that
node
os
does
it's
normal
for
the
big
boy
operators,
Big
Boy
chain
operators,
it's
a
half
to
Upright
their
machines
where
that
state
database
is
in
Ram,
which
then
results
in
yeah
anytime.
You
do
anything
maintenance,
normal,
you
kind
of
got
to
use
a
snapshot
to
relaunch
it.
C
Similar,
like
Matthew,
probably
uses
snapshots
a
lot
when
you
use
deepmind,
you've
kind
of
got
to
start
it
at
a
certain
point,
and
that's
that
snapshot
process
that
basically
says
here's
where
I
want
you
to
be,
and
March
forward
go
forth
again:
yeah
even
picks
up
one
you'll.
You
know
it
picks
up
on
your
blocks,
log,
where
you
liftoffs
and
Carries
On
and
State
history
like
say
it
it.
B
B
B
C
Matthew's
got
a
Matthews
is
usually
you
guys,
go
well,
yeah
I'm,
pretty
it's
pretty
solid
yeah,
unless
you
just
really
have
it.
Jacked
up
Kevin's
seen
me
break
it
in
unique
way
for
the
most
part,
as
long
as
your
log
files
aren't
corrupted,
really
really
corrupt
yeah,
it's
pretty
solid,
but
like
say
if
it,
if
you've
got
some
bad
data,
it
can
get
and
even
in
the
the
block,
slug
Tool,
what's
called
leap
utility
now
that
actually
fixes
up
a
lot
of
those
and
manual
Corruptions.
C
Nothing
says
in
a
paper
but
yeah
yeah,
Matthew
I,
don't
know
why
he's
not
talking
today
he's
got
a
tricky
little
script
that
actually
you
can
run
and
it
asks
you
which
snapshots
you
want
to
use
or
anything
it's
pretty
cool,
so
piggybacks
on
that
I
have
all
the
big
snapshot.
Providers
tend
to
have
kind
of
adapted,
probably
but
Matthew's
League,
to
have
on
their
snapshot
site
like
a
latest.zst
that
people
have
kind
of
started.
C
Building
these
processes
that
say
sphere
run
snapshots,
PS,
USA,
Nations,
whatever
go,
and
look
here
and
pull
the
latest
snapshot
latest.gfd.
Well,
it
extract
it
restore
it.
So
there
has
kind
of
been
a
pseudo
standard
that
we
use
DST,
because
it's
better
than
r
as
Matthews
made
us
accept
I'm
on
these
big
large
files.
These
things
are
getting
to
be
10,
12
gig
snapshot
files
for
some
of
the
larger
chains
and
they
don't
get.
C
Old
snapshots
I,
don't
trust
those
Canadians
all
right
yeah,
we
all
provide
them.
We
all
provide
them
as
a
service
in
the
community,
so
you
can
actually
go
to
our
snapshot
or
tools
below
this
and
part
of
the
idea.
Is
you
need
a
verification?
Everybody
would
have
to
trust
that
Mason's
snapshots
are
or
the
real
accuracy
of
the
chain.
When
you
have
multiple
providers
that
allows
you
well
look.
Not
everybody
thinks
that
they're
on
Matthew's
chain
yeah,
we
all
have
given
our
own
version
of
it,
which
obviously
should.
C
C
I've
got
something
to
add
from
from
a
creation
standpoint.
When
you
run
a
snapshot,
the
node
stops
right.
So
it
also
Hampshire
the
state
freezes,
correct,
yeah,
the
the
note
it
doesn't
crash
it
just
does
not
progress
anymore.
It
hangs
while
the
snapshot
is
taken
for
our
services
when
we
create
the
snapshot,
we've
got
separate
nodes
for
that.
So
that's
a
that's
a
bit
of
a
like
pain
like
I
suppose
this
is
what
you
got
to
do.
I
guess,
but
I
think
that
we
would
run.
You
know
snapshots
on
every
note.
B
C
B
C
B
C
It
kind
of
is
because
it
wax
has
this
massive
State
running
in
memory
and
we
run
i9s
which
are
an
ACC
and
they
they
run
fine,
but
about
every
week
or
so
one
of
them's
just
like
now
now
and
just
Falls
over,
and
it
needs
to
kind
of
be
kick-started.
So
but
again
it's
a
hardware
kind
of
level
issue
I
mean
it
runs
perfectly
fine
as
long
as
it
ain't
running
a
wax
node,
but
because
state
is
just
monopolizing
every
chip,
it's
got,
it
finds
the
bad
ones.
C
So
node
OS
is
very
picky
about
memory
and
wonkiness
and
Brian.
It's
not
we're
not
restarting
because
of
a
notice
issue
here,
all
right,
we're
starting
because
mainly
it's
from
a
maintenance
thing
for
whatever
reason,
stuff
Force
move
biases
are
upgraded
whatever,
whatever
at
the
turn
the
machine
open
on
again
and
because
everything
was
running
in
the
in
Tampa
Fest
in
Ram,
it
needed
a
snapshot
to
fire
it
up
again.
F
C
C
Well
I:
don't
even
have
my
Snapshot
really
kind
of
available
until
about
15-20
minutes
into
it
and
then
I
download
this
10
gig
file
extract
it
and
then
kick
off
the
restore
or
the
snapshot
recovery
that
takes
15,
maybe
30
minutes
and
then,
if
there's
kind
of
anything
else,
either
to
repair
replay
or
resync,
which
on
waxed
I
mean
you
get
an
hour
behind.
It,
takes
you
about
30
minutes,
yeah,
maybe
15
minutes
to
resync
icing
on
the
cake.
Your
API
is
responding
while
you're
doing
that
and
serving
out
stale
data.
C
B
C
The
snapshot
is
probably
maybe
20
gigs
and
then
I
think
we
all
the
our
ball
of
zip
them
up
and
it
becomes
about
15
gigs,
but
again
I
mean
that's
when
it's
10
gigs
a
pop
and
and
it
just
it.
It
can
take
you
a
solid
hour
best
case
scenario
and
here's
an
interesting
thing.
It
took
me
14
hours
to
recover
a
wax
State
history
node
last
week,
I
hit
seven
up
and
was
like
holy
hell.
C
What
is
this
and
ultimately,
when
it
starts
happening
to
repair
any
of
those
like
State
history,
log
files
get
a
day
almost
I
mean
if
Florida
in
it
with
the
99
ripping
the
out
of
it.
It
just
took
back
I
know
that
if,
if
you
want
to
start
a
state
history,
node
that
has
no
pre-existing
statistical
blocks
from
a
snapshot,
in
other
words
starts
a
light
node.
That
thing
can
toil
for
hours
and
hours
and
hours
while
it
rebuilds
itself,
and
that's
not
even
going
to
go
back
to
the
beginning
right.
E
C
E
C
D
Do
it
take
your
way
for
now,
there's
a
lot
of
disk
I
know,
that's
really
like
I.
Remember
correctly,
that
was
the
bottleneck.
It's
like
this
guy
always
needs
to
write
this
initial
State
history
files
and
it's
just
like
killing
you,
you
know
like
I,
usually
run
it
on
like
grade
zero,
nvmm
and
vme
arrays,
or
something
and
you're
still
like
you're
you're
maxing
that
out
so
I
guess.
That's
still
the
bottleneck.
C
And
and
that's
where
Ross
gets
the
temp
at
best
I
mean
memory
is
faster
than
Raid
0
nvme
kinda,
but
it
was
still
I
mean
yeah
the
speed
of
line.
We
would
have
the
limit.
We
were
trying
ssds
away
even
just
on
the
EOS
Network,
just
running
that
state
database
in
three
years
and
you
throw
them
in
the
bit
yeah
mid
hour
and
consumers
evos,
but
for
the
for
the
record
everybody
to
match
the
out
of
my
nvmes
for
the
state
DB
about
you
know
in
Vietnam.
C
Do
you
know
I
can
I
checked
the
smart
wear
out
on
the
drives
and
since
I
do
the
Championships?
It's
just
the
io
goes
up
and
the
actual
disk
I
o
goes
down
and
the
the
the
where
art
just
lost
like
it's,
not
even
touching
the
disc,
but
the
sacrifice
he
makes
is
that
anytime
he's
never
short
stories
from
a
snapshot.
Every
single
I
don't
know
I'm
fine
like
right
now,
even
when
I
have
to
start
a
note
at
other
snapshot,
I
feel
weird
like
what's
going
on,
I'm
so
used
to
it
now
so.
B
It
sounds
like
there's,
okay,
so
there's
two
very
different
cases:
right,
there's
restarting
a
state
history,
node
and
there's
restarting
a
non-state
history
node.
That
second
case
is
the
more
common
case
and
it's
much
faster.
Let's
focus
on
that
one.
First,
in
that
case
you
said
maybe
15
minutes
or
perhaps
a
little
longer
for
a
wax
node.
B
C
Twofold
one
you're
getting
behind
those
15
minutes
two
when
it
does
come
online
and
tries
to
start
scrambling
to
catch
up,
you're,
feeding,
stale
data,
if
you
haven't,
had
some
extra
help
checks
or
change
the
API
borders.
Oh
that's
the
other
you
can
get
behind
and
I
can
tell
on
all
my
monitoring
system
when
everybody
does
that,
because
all
of
a
sudden
I
see
something
from
10
minutes
ago,
where
the
hell
that
come
from
yep
somebody's
apiat
and
we're
just
it's
lesser
of
two
evils:
I'm
like
yeah.
C
C
You're,
one
of
the
things
we
talked
about
I
think
is
a
feature
request.
Was
that
ready
and
it
returning
like
a
like
a
ready
in
point
or
something
of
that
nature
to
where
you
could
intelligently
say
if
it's
returning
a
200
assume
it's
it's
ready
to
accept
connection.
That
said
well,
what
drives
that
threshold?
In
my
opinion
and
Kevin
I,
think
narrowed
it
down.
The
node
reacts
differently
when
it
knows
it's
behind
lib
versus
trying
to
catch
up
to
the
head
block
it
sinks
in
bulk.
It
does
things
differently.
My
opinion.
C
B
C
A
C
D
C
D
C
Just
denoting
or
automate
it
so
we
used
to
have
this
problem
all
the
time
nodes
wouldn't
resync,
because
all
the
peers
would
connect
and
whatever
so
we
wrote,
console
services
and
console
template
that
rewrites
nginx
automatically
based
on
whether
the
node
is
in
sync
or
not
so
nginx
will
send
peer-to-peer
traffic
or
API
traffic
only
if
the
node
is
in
sync
right:
they've
modified
their
proxy
yeah,
because
otherwise
I
was
dealing
the
same
thing
with
Michael.
It's
like
manually,
put
it
in
now.
I
was
like
this
is
me
so
so
the.
C
Sorry,
I'm,
sorry,
the
ships
are
already
compressed.
It's
easy,
yeah
I
mean
the
files
themselves,
other
than
being
massive.
Big
I
mean
they're
blush.
It's
blockers
gives
you
about
30
percent
I
use
lz4
on
using
ZFS,
and
that
gives
you
about
30
State.
History
is
less
so
you
say
Matthew,
it's
already
complete
staff
in
this
chain,
yeah
I
thought
State.
History
was
already
impressed
right,
yeah.
E
D
C
A
C
E
Well,
it
was
coming
up
because
you
know
we're
talking
about.
You
know
things
to
optimize
thinking,
right
and,
and
we
were
talking
about
well.
What
are
we
really
talking
about?
Are
we
I,
don't
think
we're
talking
about?
We
need
to
optimize
thinking
from
Genesis
right.
What
we're
wanting
to
optimize
it
sounds
like
is
sinking
from
a
start
of
a
snapshot
right
up
to
up
to
hit
right.
C
That
are
two
different
issues:
I
mean
one's
less
prevalent,
but
a
lot
of
people
don't
want
to
start
from
a
snapshot.
They
will
I
mean
like
exchanges
or
you
know,
validation,
heavy
Awards.
They
want
to
have
the
old
chain
from
block
one
history
provider
right.
So
all
right,
but
that's
different
than
95
of
the
nodes
or
aren't
even.
B
E
C
Natural
tendency
is,
and
the
way
instructions
intend
to
be
is
get
the
Genesis
block,
hold
some
peers
and
sync
it
all
in.
But
as
we
have
all
painfully
learned,
you
don't
do
that
number.
One!
Good
luck!
Finding
the
peers
number
two!
It's
gonna
take
months
and
months
and
months
and
months
and
months,
so
I
sent
him
to
Ross's
site
over
bill.
It's
just
block
log
of
a
couple
terabytes
and
then
get
that
jump
well,.
D
D
Like
what
aside
from
Genesis
or
even
like
a
snapshot
like
they
might
but
I,
think
the
real
use
case
for
replay
is,
if
you
didn't,
have
a
history
solution,
but
you
have
all
the
data
and
you
want
to
replay
it
to
extract
that
history
solution
even
for
SNAP
for
using
replay
to
a
certain
part,
normally
you'll
just
trim,
and
then
you
can
just
continue
normal
syncing.
To
that
point,
like
you,
don't
really
need
a
replay
as
much
because
there's
nothing
to
replay
after
you
trimmed.
E
C
That's
the
jet,
try
you
can't
yeah.
In
theory,
it's
still
looking
to
ending
every
block
and
transaction,
but
from
what
I've
found
in
the
conversations
and
I
don't
have
too
many
of
them,
but
they
any
external
Trust
of
source
other
than
peer-to-peer
block
by
block
validation
of
the
transactions
is
nope.
We
don't
do
that.
We
start
at
the
beginning
of
the
chain.
We
sink
the
whole
chain.
We
we
know
we
have
personally
validated
and
there's
no
way
is
how
they
didn't
say
it
now.
C
Yes,
you're
right,
it's
the
same
thing:
here's
the
log
it's
going
to
validate
it,
I,
don't
even
know.
Does
it
validate
I
guess
there
is,
but
but
it
would
be
because
when
you,
when
you're
syncing
with
a
peer,
you
basically
just
copying
from
their
blood,
look
right
and
it
validates
but
you're
validating
every
transaction.
If
you're
doing
a
full
validation.
D
Right
so
there's
replay
like
it's
the
same
thing,
but
it's
really.
It
comes
down
to
conceptually
like
people.
The
feelings
from
you
know
coming
from
different
crypto
and
stuff
they're
less
like
no.
No,
we
gotta
start
from
Genesis
We're,
Not
Gonna
trust
someone
else
under
the
Earth.
It's
already
the
same
you're,
just
learning
back
extra
Network
bandwidth
and
you're,
but
you're
still
doing
the
validation
and
making
sure
it
builds
on
top
of
each
other
yeah.
That's
that's!.
E
F
That's
that's
the
thing
is
that
with
like
Bitcoin,
when
you,
when
you
resync
the
or
when
you
sync
with
node
you're,
not
just
pulling
for
one
from
one
other
node
from
one
peer
you're
pulling
blocks
from
you
know
a
thousand
other
peers
over
the
over
the
three
day
period
or
whatever
that
you're
downloading
it.
So
you
know
it's
that
redundancy
that's
built
in
that
kind
of
has
solidified
itself
into
ethos
there,
especially
among
exchanges
and
that
sort
of
thing.
B
It's
got
a
lot
of
notes
and
and
new
information
I
didn't
have
previously
next
steps,
as
I
mentioned
before
is
but
developing
you
know
first
draft
proposal
for
peer-to-peer
improvements
as
we
go
through
that
process.
It's
very
likely
that
we
will
encounter
new
unanswered
questions
that
we
didn't
think
to
ask.
B
D
Maybe
just
a
very
small
request,
since
we
were
talking
about
snapshot,
if
we
can
have
the
snapshots
from
node,
EOS
output
be
similar,
like
us,
nations
want
like
in
the
naming
where,
instead
of
just
having
that
you
know
the
hash
of
the
block
where,
yes,
you
can
extract
the
block
number
from
it.
Instead,
you
can
have
the
hash
and
the
block
number
already
there.
It's
just
much
easier
when,
when
matching.
C
So
yes,
I
mean
even
just
simply
the
block
number
itself,
even
if
it's
not
a
date
time,
I
live
off
daytime,
but
the
block
number.
That's
not.
What
is
that?
Where
the
hell
is
it
yeah
and
I
promise
you're
getting
it
off
by
five
minutes.
Pisses
you
off
damn
missing
gaps
in
your
state
history.
Then
you
gotta
go!
Do
it
all
over
again,
because
you
were
five
minutes
after
when
it
actually
was
needed.