I know I shouldn't be surprised, because it seems like every time I try to use one of the Heartbeat STONITH plug-ins I find it just doesn't work. Of course, with the exception of the "external" plug-in which I designed, and Scott Kleihege largely implemented. The problem is that STONITH plug-ins are fairly complicated to implement because of the way they're designed. So, getting in and fixing one tends to be pretty hard work. Implementing a new one even harder.

Anyway, the short answer is that the AP7900 just doesn't work with either the apcmastersnmp or the apcmaster STONITH modules. Read on for some more information.

I could have sworn that we got one of these units working in the past, but we must have been using a different firmware version or something. APC has fairly dramatic changes between different versions of their management modules. The OIDs that the SNMP module supports just don't seem to exist in the 7900 units (and similar I'm assuming).

So, as is usually the case, we went with the STONITH external module. In the past we had written one that uses the web interface, but the URLs that it was expecting had changed. <sigh> I finally wrote an external script that uses the Python "telnet" module to do the reboot and get status, to get things working.

The apcmaster interface should have worked similarly, but it just didn't and the debugging information wasn't at all helpful. The apcmaster module uses the expect library to telnet in, so it has quite a few moving parts and is fairly hard to debug. After my attempts with the IPMI module, I decided to just avoid the 4 hour debugging session and implement an external script like I ended up doing with IPMI.

So, if you are trying to use the APC AP7900 or AP7901, you definitely should be looking at implementing an external script rather than using one of the supplied plug-ins. I didn't try the "apcsmart" plugin because we needed multiple machines to have access to devices on each APC switch. If your needs are different, you may want to try that plug-in.
(Post Reply)