1 |
On 11/14/2014 10:53 AM, Thanasis wrote: |
2 |
> I have an APC SC620I, which in case of power failure, it successfully |
3 |
> initiates a shutdown to the connected (via SMART cable) PC, but if the |
4 |
> mains power returns, the UPS does not recycle the power to the PC, and |
5 |
> consequently the PC stays off. |
6 |
> |
7 |
> Regardless if the mains power returns soon after the UPS has initiated a |
8 |
> shutdown to the PC, shouldn't the UPS recycle the power anyway, so that |
9 |
> the PC comes back on as set in BIOS? |
10 |
|
11 |
I assume that your PC shuts off thus reducing the load - and, of course, |
12 |
this increases the runtime of the remaining battery so the UPS never |
13 |
actually shuts down? |
14 |
|
15 |
I don't think you can get apcupsd in any case to cycle the outlet |
16 |
groups, but you can try a couple things. |
17 |
|
18 |
First, is apcupsd even sending the signal to shutdown the UPS? |
19 |
|
20 |
/etc/init.d/apcupsd.powerfail needs to be added to the shutdown runlevel: |
21 |
|
22 |
`rc-update add apcupsd.powerfail shutdown` |
23 |
|
24 |
At the end of the shutdown this is run and it tells the UPS to power |
25 |
off. If you have multiple PCs this should be enabled on the slowest to |
26 |
shutdown to make sure it doesn't kill a machine still shutting down. |
27 |
|
28 |
> |
29 |
> Am I missing something in the configuration or the daemons that should |
30 |
> be running? |
31 |
> |
32 |
> (Running sys-power/apcupsd-3.14.8-r2) |
33 |
> |
34 |
> /etc/apcupsd/apcupsd.conf folows: |
35 |
> KILLDELAY 0 |
36 |
|
37 |
If the above fails (if the above does indeed fail, some troubleshooting |
38 |
should happen to try to figure out why it doesn't work), KILLDELAY is |
39 |
the parameter you likely seek, but it is dangerous. If you set this it |
40 |
will wait x seconds after a shutdown was requested and forcibly shut |
41 |
down the UPS power. |
42 |
|
43 |
However, in the even that power is restored between the UPS kill and the |
44 |
time it actually turns off the mains will still not be cycled. But |
45 |
theoretically this window should be pretty small. |
46 |
|
47 |
Dan |