ararat synapse TPingsend for multiple IP addresses
lazarus / 38-Networking and Web Programming

----------------------------------------------------------------------
sporex · 2021-01-04 17:08

Hi there,

I am using the function below to ping a sequence of IP addresses every
5 sec in the following way:
Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i repeat    
        result:=real_ping(IP1…..)             
result:=real_ping(IP2…..)              result:=real_ping(IP3…..)     
       sleep(5000)until false This is the function:
Code: Pascal  [Select][+][-]window.onload = function(){var x1 =
document.getElementById("main_content_section"); if (x1) { var x =
document.getElementsByClassName("geshi");for (var i = 0; i function
real_ping(IP:string; TTL,TMOUT:integer):boolean;var s:string;
 r:boolean;begin     RealPingSend := TPINGSend.Create;   
 Result:=false;     RealPingSend.TTL:=TTL;   
 RealPingSend.Timeout:=TMOUT;     RealPingSend.PacketSize:=16;   
 try     r:=RealPingSend.Ping(IP);     s:=RealPingSend.ReplyFrom;   
 if r and (s=IP) then Result:=true else writeln('ERROR : ',IP,' :
',s,' : ',RealPingSend.ReplyErrorDesc);     finally     
 RealPingSend.Destroy;       //FreeAndNil(RealPingSend);     
 //RealPingSend.Free;     end;     //sleep(30);end; 
In case of error the function prints the pinged IP, ReplyFrom IP and
Error Description.

What is troubling me is this:
- occasionally the function prints the preceeding IP in the sequence
as "ReplyFrom"
- example: while pinging IP3 I get sometimes error result: ERROR IP3 :
IP2 :
How is this possible as I call RealPingSend.Destroy in the finally
clause?
I also tried RealPingSend.Free; FreeAndNil(RealPingSend);
The result is the same in all cases.

All this on Linux Debian 9 (LAZ 1.2.4+dfsg2-1, FPC 2.6.4) and of
course I am running the app as root.
Thank you in advance for any help.

----------------------------------------------------------------------
MarkMLl · 2021-01-04 17:13

A delayed or duplicate response? Suggest you monitor the connection
with wireshark.

MarkMLl

----------------------------------------------------------------------
sporex · 2021-01-04 17:23

Quote from: MarkMLl on January 04, 2021, 06:13:37 pmA delayed or
duplicate response? Suggest you monitor the connection with wireshark.

Yes of course it is a delayed response. But as I destroy the Object
after every attempt I thought that the property ReplyFrom should be
cleared before the next attempt to a new address?

----------------------------------------------------------------------
MarkMLl · 2021-01-04 17:41

Remember that ICMP is connectionless.

MarkMLl

----------------------------------------------------------------------
sporex · 2021-01-04 17:44

It seems that TPingSend has also other hitches under Linux. I.e.
unpredictable outages.
Even if I kill the app I cannot get a reply when I launch the Linux
Ping on the OS (only after a reboot)
So I will go for a TProcess that invokes the Linux Ping.
The only problem is that the Linux Ping does not accept a timeout of
less than 1 sec.

----------------------------------------------------------------------
sporex · 2021-01-04 17:49

Quote from: MarkMLl on January 04, 2021, 06:41:52 pmRemember that ICMP
is connectionless.

MarkMLl

OK, does that mean that if I create a new TPingSend the delayed
response will be accepted by the new object?
I begin to understand.

----------------------------------------------------------------------
MarkMLl · 2021-01-04 18:08

With the caveat that I normally do this sort of thing directly rather
than going via Ararat etc., but basically much of the unix-style
networking works by a bit of software coming along and saying to the
kernel "hey, I'm interested in protocol xxxx, send me anything that
arrives".

ICMP echo messages generally have a sequence number in them, and
that's the only identification other than the IP address of the
participating systems: it's entirely possible for echo requests
(pings) sent by different processes to get mixed up (I've seen this
when testing firewalls etc.) although this mostly happens with
broadcasts/floods.

So if you're in control of the sequence number starting it with a
random number will help, as will keeping a window of possible response
numbers. I can't remember what the unix ping utility does (but the
source is intelligible), and I don't know whether it's possible to do
anything with the "code" (as distinct from "type") or if this being
zero is mandated by ICMP.

MarkMLl