On 4/11/2019 11:07 PM, Christian Herzog wrote:
FPING should be the path to fping (/usr/sibn/fping?) not xymonping. you're right. The frankensteined Debian package I built for 4.4 alpha didn't set that correctly. I now compared all config values to our production setup, but FPING was the only notable difference.
I think it might be worth it to officially retire xymonping in the 4.x branch, as it'll need to be updated for IPv6 as well and fping development seems to be healthy and reliable on platforms. Or at the very least, gate it for IPv6 building in the short term.
Anyway, the status is the same as yesterday: for a v6 only host, conn is now green, but http and ssh stay red (DNS error)
current xymonnet debug log: https://people.phys.ethz.ch/~daduke/xymonnet-log
@JC: how can I debug this further? What other (IPv6 related) test would be interesting for you?
Looking at xymonnet.c and contest.c, I'm actually suspecting here that we're not building NET6 connections at all in xymonnet tests, which means there's more connecting work left to do than expected. I'm not sure if the original plan was to migrate to xymonnet2 before IPv6. The factored-out connectivity routines in lib/tcplib.c are being used for IPv6 connectivity in sending *messages* back and forth between xymon servers and clients, but not for network testing.
For the short term, testing xymonclient messaging and xymond record management (using conn tests as sources) might be the most useful. xymonnet/TCP polling will take some additional patching.
Regards,
-jc