HCDas> hello,
HCDas> is there any way to create a losspattern which only fires when there is
HCDas> loss on the last hops? Actually we are getting alot of false positive
HCDas> mails, because there is packetloss in the middle of the trace which
doesnt
HCDas> bother anyone.
HCDas> thanks beforehand!
I'm not sure what you're talking about.
There's two situations I can envision that might match what you're saying.
[But, as you'll see, it still doesn't make much sense.]
Option 1) There's packet loss in the "middle" of your connection somewhere -
but there's nothing you can do about it since it's not in a network you have
any control over. [I still expect you'd want to _know_ about it - so
complaining about notices seems odd. On a side note: I don't send notices
directly from smokeping. I use Nagios for *reporting* of smokeping events. It's
not as fine grained as smokeping in the alert patterns, but it is a LOT better
squelching, setting alert hours, and groups etc. Since I use Nagios for other
monitoring tasks, it's a no brainer to use the smokeping plug-in and avoid
using smokeping for alerts all together.]
Option 2) There's actually no packet loss, but if you do a trace-route one of
the hops in the middle won't respond to you. As I understand this, this happens
when a router is configured to give ICMP traffic low priority, and it's very
heavily loaded. Thus, virtually all traffic IS being handled, but responses to
low priority traffic like ICMP gets sporadically dropped. But this second case
shouldn't show up in a standard FPING test. Further, it's a warning sign that
something IS dropping traffic, even if it's "only" ICMP. [And, thus, I don't
think it's a false-alarm. It's just an early warning that some traffic is being
lost. And you'll probably find that latency sensitive, or packet-loss sensitive
traffic IS affected - but you personally may not have a much or any of those
types of traffic so you don't get complaints. My experience is that regular
users probably won't complain until packet loss reaches >15%. But things like
VPN tunnels and the like will run very badly in a 15% loss case. Browsing the
web probably won't be noticeable enough to generate a complaint for must users.]
So, the question you pose is quite confusing.
I do have a setup in smokeping where certain triggers cause an MTR [Matts trace
route - which is much like Window's Pathping.] This gives me the stats to
evaluate exactly where the loss is, and it's nature so I can exert pressure on
whomever needs to address the issue. And it prevents the vendor from weaseling
out by claiming the loss is happening elsewhere outside their network. [Most
effective on first or last hops, since we usually pay these vendors directly.]
So, clarify your situation - because I'm not at all sure I understand, or have
offered anything of value.
-Greg
_______________________________________________
smokeping-users mailing list
[email protected]
https://lists.oetiker.ch/cgi-bin/listinfo/smokeping-users
