[Nut-upsuser] occassional problem wtih upslog and apc ups units via snmp ... [NA] status
clepple at gmail.com
Fri Aug 14 13:27:54 UTC 2015
On Aug 14, 2015, at 8:53 AM, Bob Brown <bbrown at harpercollege.edu> wrote:
> I see this in syslog:
> Aug 12 03:57:22 orion upsd: Data for UPS [a102-apc-10] is stale - check driver
I don't have a SNMP UPS to test with, and this might be different for NUT 2.4.1, but the current defaults seem to make stale data more likely in your case.
In http://www.networkupstools.org/docs/man/snmp-ups.html#_extra_arguments the default for "pollfreq" is 30 seconds, and in http://www.networkupstools.org/docs/man/upsd.conf.html the "MAXAGE" default is 15 seconds. So you might want to either increase MAXAGE, or decrease pollfreq.
I would recommend making MAXAGE at least pollfreq (possibly 1.5x larger), because decreasing pollfreq might run up against the amount of time that it takes to poll all of the SNMP variables from the UPS. There are some retries and timeouts in snmp-ups which make this difficult to guess without seeing the debug logs. You might check the list archives to see if anyone has suggestions on what pollfreq and MAXAGE values worked for them.
> (one for each UPS).
> Then I see a bunch of messages like this:
> Aug 12 03:58:02 orion upsd: write() failed for ::1: Broken pipe
This just means that the client (upslog, probably) disconnected before upsd expected the session to end. The upsd-to-driver connection is a Unix-domain socket (and the driver-to-UPS connection is UDP).
> Aug 12 03:58:02 orion snmpd: Connection from UDP: [127.0.0.1]:50940
Might be unrelated, unless I misunderstood your NUT setup (upslog -> upsd -> snmp-ups -> UPS). The snmpd daemon and the UPS comms card are SNMP servers ("agents" in SNMP parlance, IIRC), and snmp-ups is a SNMP client.
clepple at gmail
More information about the Nut-upsuser