On 2019-12-17 06:18, Roger Price wrote:> > I definitely need to check that the vile daemon is present and running. > > In upsmon.conf there are no NOTIFYFLAG declarations, especially for events > ONBATT and LOWBATT. When ONBATT and LOWBATT occur nothing will happen. There > are no NOTIFYMSG declarations and in particular no messages for the ONBATT and > LOWBATT events. > > If you add NOTIFYFLAG and NOTIFYMSG declarations to upsmon.conf, you should get > a better idea of what is happening. > > Additionally, in upsmon.conf there is no NOTIFYCMD declaration which would call > upssched.conf. So the only system shutdown criterion will be on LB status.OK, so I definitely missed configuration steps and, apparently, forgot that I'd never finished. I will have to look over everything again. Also, looking at my own monitoring graphs of what happened around the outage, it looks as though runtime on my UPS is strangely non-linear with respect to load — i.e, dropping to my critical-systems-only load which is about 40% of my normal total load (which is to say, slightly less than 20% of UPS capacity vs. about 45%) does not appear to have extended runtime nearly as much as expected. It more than doubled the UPS's PROJECTED runtime according to the UPS, but not the *actual* runtime. Also, it looks from my graphs as though the UPS was claiming 10% capacity when it shut off, but 80% when the power came back on. There are some very odd things going on here that I need to get to the bottom of. -- Phil Stracchino Babylon Communications phils at caerllewys.net phil at co.ordinate.org Landline: +1.603.293.8485 Mobile: +1.603.998.6958
On Dec 17, 2019, at 3:45 PM, Phil Stracchino <phils at caerllewys.net> wrote:> > Also, looking at my own monitoring graphs of what happened around the > outage, it looks as though runtime on my UPS is strangely non-linear > with respect to load — i.e, dropping to my critical-systems-only load > which is about 40% of my normal total load (which is to say, slightly > less than 20% of UPS capacity vs. about 45%) does not appear to have > extended runtime nearly as much as expected. It more than doubled the > UPS's PROJECTED runtime according to the UPS, but not the *actual* > runtime. Also, it looks from my graphs as though the UPS was claiming > 10% capacity when it shut off, but 80% when the power came back on. >Slightly tangential to your original issue, but do the values like load (as reported by NUT) generally track the front panel display? We have a number of CyberPower issues whose root cause is a disagreement as to how the HID descriptor (and therefore, the HID reports) should be parsed and interpreted. Link to the whole mess (which includes some other non-HID CPS issues): https://github.com/networkupstools/nut/issues?q=is%3Aissue+is%3Aopen+label%3A%22CyberPower+%28CPS%29%22 That said, I don't think that differing load when on/off AC is going to fall into the HID category, but there are certainly a lot of open questions surrounding some of the reported values. As Roger suggested, doing a manual self-test on the batteries may be useful. (I don't know if we should see it or not for this model, but I don't see a variable with self-test results in the upsc output.)
On 2019-12-17 16:43, Charles Lepple wrote:> On Dec 17, 2019, at 3:45 PM, Phil Stracchino <phils at caerllewys.net> wrote: >> >> Also, looking at my own monitoring graphs of what happened around the >> outage, it looks as though runtime on my UPS is strangely non-linear >> with respect to load — i.e, dropping to my critical-systems-only load >> which is about 40% of my normal total load (which is to say, slightly >> less than 20% of UPS capacity vs. about 45%) does not appear to have >> extended runtime nearly as much as expected. It more than doubled the >> UPS's PROJECTED runtime according to the UPS, but not the *actual* >> runtime. Also, it looks from my graphs as though the UPS was claiming >> 10% capacity when it shut off, but 80% when the power came back on. >> > Slightly tangential to your original issue, but do the values like load (as reported by NUT) generally track the front panel display? We have a number of CyberPower issues whose root cause is a disagreement as to how the HID descriptor (and therefore, the HID reports) should be parsed and interpreted.That certainly sounds like something that could be contributing to the problem here. As of right now, my upswatch graph is showing 41%-42% load and 32-34 minutes estimated runtime. The front panel agrees, stating 41% load and 33.2 minutes. upsc is reporting ups.load: 41 and battery.runtime: 1920, with battery.runtime.low: 592. Battery charge is reported battery.charge: 100 with battery.charge.low: 35. All of these look pretty much as expected. -- Phil Stracchino Babylon Communications phils at caerllewys.net phil at co.ordinate.org Landline: +1.603.293.8485 Mobile: +1.603.998.6958