Hi all,
I’m setting up smokeping to try and gather some latency statistics on my ISP to different parts of the world. Does there exist a list or any generic recommendations of different targets to config within smokeping? Google and searching the NANOG mailing list have failed me, and I don’t want to just spam random IP addresses with ICMP requests if there’s a more official/accepted method for doing this.
Thanks!
Brian
Something I found that is helpful once you've gathered a list of targets is
the following command for generating config to paste:
traceroute -w 3 [IPaddress] | grep -v "*" | grep -v "traceroute" | sed -e
's/(//g' -e 's/)//g' | awk '{ gsub(/\./,"_",$2); print "++++ "$2"\nmenu =
"$3"\ntitle = "$2" - "$3"\nhost = "$3"\n"}'
That generates a valid output for configs to ping each hop along the way to
your destination, which can be super useful. Not all of them allow ICMP but
a decent amount do.
Very cool - thanks for sharing that.
It is also super stupid, because routers reply to ICMP echo requests
with a very low priority: this introduces jitter which makes these
measurements unreliable.
If you are not monitoring a server then you are wasting your time.
(Also, it would be nice to have the owner permission before deciding to
permanently send a lot of ICMPs to a device.)
Something I found that is helpful once you've gathered a list of targets is
the following command for generating config to paste:traceroute -w 3 [IPaddress] | grep -v "*" | grep -v "traceroute" | sed -e
's/(//g' -e 's/)//g' | awk '{ gsub(/\./,"_",$2); print "++++ "$2"\nmenu =
"$3"\ntitle = "$2" - "$3"\nhost = "$3"\n"}'That generates a valid output for configs to ping each hop along the way to
your destination, which can be super useful. Not all of them allow ICMP but
a decent amount do.
curious... why is pinging along the path interesting?
nodes on the network which process packets for routing purposes tend
to handle local destination packets vastly differently from transit
packets... additionally, there's no guarantee that the path to the
device you are pinging is congruent with the path to the far end you
traceroute to, so I'm not even sure you'd be testing the path
correctly anyway.
Hi all,
I’m setting up smokeping to try and gather some latency statistics on my
ISP to different parts of the world. Does there exist a list or any
generic recommendations of different targets to config within smokeping?
one thing I'd recommend is NOT publishing a list of 'ping these for
useful results', there are enough people that already ping /8.8.8.8/
or /4.2.2.2/ or /favorite website/ that there are definitely cases of
intermediate paths parts limiting traffic for monitoring.
Isn't the question to ask here: "Why don't you monitor things you care
about access to?"
For instance, if your mailserver is smtp.example.com perhaps testing
access over time to that is important?
8.8.8.8 is pretty popular. There's lots of them. (Anycast)
But, it's not nice to send remote networks unwanted traffic (it's a DNS server, not a light house), and I know GOOG receives enough ICMP at some of those anycast nodes to be "problematic".
I'd say keep your smokeping targets to devices you or your connectivity provider(s) own/are paying for rather than abusing random 3rd parties just to satisfy your curiosity.
Hello Brian,
you might want to consider joining the nlnog ring (https://ring.nlnog.net/).
You can request access to a full mesh smokeping for all hosts too.
Besides the ring host RIPE Atlas anchors (
https://atlas.ripe.net/anchors/list/) might be another option for hosts to
add to your list.
The NLNOG RING servers would be good targets and they expected to get
pinged.
https://ring.nlnog.net/participants/
Doh! Not sure why I didn’t think about Atlas prior to posting my question - that’s perfect.
I have an Atlas node on my network too. After I put it in and played with it for a week I started a big project at work and put it on the “to play with later” list and never got back to it. 
Thanks!
Brian
I have a couple of different SmokePing implementations related to our SaaS platform.
I have monitors inside our network looking at several of our largest customers.
I have monitors in some IaaS providers looking back at us.
I do include a couple of common addresses previously mentioned since I need some sort of confirmation to show the sites can actually get out.
My Alaska-focused public SmokePing instance:
http://akmon.techsolvency.com/smokeping/
... has links to other public SmokePings. Many of them have a general
"is the Internet healthy" target group.
I looked through those groups and selected a cross-section for my own group:
http://akmon.techsolvency.com/smokeping/smokeping.cgi?target=Other
Some of them are subject to CDN or anycast; other are not.
If anyone has a public SmokePing instance, send me a link and I will add.
Royce
Most of these "OMG, think of the target!!!!!111" posts are unwarranted. The OP asked for lists of IPs that the community agrees can be safely monitored. If it can be safely monitored , obviously the host is aware and agrees to it.
Yes, if a particular hop along the way has a higher latency than ones behind it, it's just an overloaded control plane.... but that network should be looking to upgrade that router anyway. (Cue the OMG, it's forwarding just fine e-mails... don't.)