I am trying to write a configuration for a network with a lot of
redundant connections. It doesn't work at the moment, because nagios
complains about the circular parent/child dependencies. If the cycles
were around the edge of the network, I could fake it with dummy hosts.
For instance:
A B
| |
CD
| |
E F
Here A and B are connected to the rest of the network, C has parent A,
D has parent B, E parent C, F parent D.
E is normally reachable via C and A, and F is via D and B.
If A breaks down, E gets connectivity via C, D, and B.
If B breaks down, F gets connectivity via D, C, and A.
Nagios won't let me make C the parent of D and vice versa.
I can work round this by inventing C-via-D and D-via-C,
so E has parents C and C-via-D, C-via-D has parent D.
Likewise for D-via-C.
However! This doesn't work when the host is a switch in the core of
the network, or when there are multiple redundancies. So I need to
change Nagios itself. Here's what I think needs to be done.
1. Remove the cycle detection for parents in the pre-flight checks.
2. In check_host, remember which hosts have been visited when going
through parents and children.
3. Likewise, avoid looping in the CGIs.
4. Allow for upward links in the tree-style status maps.
5. Some extra work might be needed in cgi/outages.c over and above
allowing for loops, but I doubt it would be much.
Have I missed anywhere? Are there any obvious flaws in this plan?
Peter.
This post was automatically imported from historical nagios-devel mailing list archives
Original poster: [email protected]