Package: ntop
Version: 1.2a7-6
Eg:
Host Act -Rcvd- Sent TCP UDP ICMP
anarres B 1.0 GB 27.1 MB 1.0 GB 2.7 MB 0
davenant B 36.6 MB 1.0 GB 32.5 MB 4.1 MB 0
[press p]
Host Act -Rcvd- Sent TCP UDP ICMP
anarres B 93.9% 2.5% 1.0 GB 2.7 MB 0
davenant B 3.3% 96.2% 32.7 MB 4.1 MB 0
[press p]
Host Act -Rcvd- Sent TCP UDP ICMP
anarres B 19.6 Kbps 868.0 Kbps 1.0 GB 2.7 MB 0
davenant B 887.3 Kbps 26.4 Kbps 32.9 MB 4.1 MB 0
Notice how in the throughput display davenant seems now to be
receiving much more than it's sending, whereas in the other two
displays it shows up as sending more. (I happen to know the traffic
pattern here, and davenant is mainly sending.)
I think the attached patch should fix this. I've tested it and it
seems to fix the problem.
However, on my early potato system ntop seems very unreliable. It
* coredumps occasionally (I haven't figured out how to repeat this,
but it seems more likely to do it at or near startup, particularly if
you press `p' or ^L near startup);
* messes up its screen layout with `loading such-and-such' messages,
at startup.
* can't cope with the widths the fields can take in throughput mode.
If you're interested I'd be happy to investigate the coredump problem
too, or provide more information (but I won't bother if ntop is
considered deprecated or you don't think it's worthwhile).
Ian.
--- /u/ian/junk/.davenant/d/ntop-1.2a7/report.c Sun Aug 6 22:13:26 2000
+++ report.c Sun Aug 6 22:02:49 2000
@@ -1674,8 +1674,8 @@
else /* Throughput */ {
sprintf(buf, "%-26.26s%c %9s %9s %9s %9s %8s",
tmpName, formatStatus(el),
- formatThroughput(el->actualSentThpt),
formatThroughput(el->actualRcvdThpt),
+ formatThroughput(el->actualSentThpt),
formatBytes(c, (short)webMode),
formatBytes(d, (short)webMode),
formatBytes(e, (short)webMode));
