Why not use the system MIBs ?
- Because I found some broken MIBs on some distros. I prefer to use a bit more disk space, and have a consistent and tested set of MIBs instead of spending time telling Cisco how messy they are.
- This is not strictly necessary, but I have found that on some platform (notably Ubuntu), SMI has built-in default compiled SMIPATH, making it to go to its own PATH to find and load MIBs. It then produce all sorts of strange problems like segfaults and SMI lint errors.
- Renaming /etc/smi.conf to /etc/smi.conf_orig might provide a hint or even fix issues.
I'm getting ERROR or WARNING fn=InterfaceAPI/get : devicename : SNMP get failed : long() argument must be a string or a number, not 'NoneType'
Be sure to install the latest version of Agent-Jones. I have added some sanity code to cover Cisco creativity in dot1dBasePortIfIndex.
I'm getting "EntryPoint object has no attribute resolve". Update setuptools.
- PEP 479 seems to break ProxyIter/iteritems() in snimpy/manager.py.
- This patch is needed, fixed in snimpy 0.8.13
I'm getting a lot of unknown vendors when using showmac=1
You may want to update the netaddr pip module using
pip install --upgrade netaddr
However, I have found some gear vendors using MAC prefixes out of nowhere, unknown from the official IEEE/OUI database. I'm not sure if those vendors shall be blamed or if I misunderstood the OUI standard.
I'm not yet sure about this case, it looks like it's an IOS bug for stacked switches, or very slow and big Cisco stacks. I have added a distinct timeout for those operations.
Caused by Avaya phones. See the Release notes and install the latest version, it shows the VLAN for which each MAC is seen.
Because the underlying snimpy library does not support them.
This patch might be needed depending on your combination of snimpy and PySNMP versions.