Re: On UPS monitoring

Beef wrote:
On Thu, 16 Mar 2006 09:32:02 +0000, Jean-David Beyer wrote:

Beef wrote:
On Wed, 15 Mar 2006 12:19:29 +0000, Jean-David Beyer wrote:

Beef wrote:
Our neighborhood had another power outage today. PSE&G seems to have
trouble with this... Every two weeks or so, there is a brown or black
out.

Usually, the UPS takes care of the very short outages (often under three
seconds), but today the power was off for about half an hour.

My UPS is an APC Back-UPS ES 350; when fully charged, this should power
my system for 43 minutes, if I can believe the log file:

[snip]

But today, this didn't wuite happen.

The power went out, I saw a message that shutdown was initialized, then
nothing much happened.

By hand, I quickly quit processes that were using the external discs and
unmounted them. I logged out of terminals.

I tried to start a console session as root, but couldn't, so that part of
things (blobking new log-ins) was working.

But the UPS gave out before the shutdown was complete.

Is this a problem with apcupsd badly reporting the charge left, or with
shutdown now going through the steps quickly enough? How can I find out?


I still use the PowerChutePlus-4.5.3-2 program because the others do not
work well for me in Red Hat Enterprise Linux 3 and CentOS 4.2.

In my experience, if the battery in your UPS is over about three years old,
you should seriously consider replacing it, or severely cutting down the
expected run time.

I bought this one about three months ago, so I wouldn't expect the battery
to be bad already.

On a system such as RHEL 3, it takes a few minutes (though less than 5) to
get the system shut down, so you better tell it to shut down when you have,
say, 10 minutes left.

Yes, I would hope that the ups daemon would start the shutdown with plenty
of battery time left.
Hoping will not do the job for you. It is configurable, so you can set it
where you want. The default may be too small.

I know that I eed to do something active, not just sit and hope & pray ;-)

I suppose the only way to test would be to look at the remaining battery
time reported, then pull the plug myself and time what happens...

But what worries me, is that the shutdown initiated by the ups daemon
didn't get done quickly enough, when if I issue the command, or click on
Gnome's "logout" icon in the panel and choose "shutdown", the procedure
gets through in less than three minutes.

I timed this way of shutting down (Gnome panel "logout" icon): 30 seconds.

If it starts a shutdown in 30 seconds after pulling the plug, the shutdown
is almost certainly being started because of low battery and not because the
time since the power failed option (which is what should cause a shutdown).
And if your battery gets too low in 30 seconds, your UPS is way too small.

There is definitely something wrong with the shutdown initiated by
apcupsd. Even allowing for the fact that there are extra processes and
discs to unmount, it should take less the a minute.

I do not know about that. On my machine, there are a lot of processes running:

10:25:50 up 11:37, 3 users, load average: 4.18, 4.24, 4.19
126 processes: 119 sleeping, 6 running, 1 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait idle
total 24.0% 360.4% 14.8% 0.0% 0.0% 0.0% 0.0%
cpu00 10.8% 80.8% 8.2% 0.0% 0.0% 0.0% 0.1%
cpu01 6.8% 88.7% 4.4% 0.0% 0.0% 0.0% 0.0%
cpu02 5.7% 91.9% 2.1% 0.0% 0.0% 0.0% 0.1%
cpu03 0.7% 99.0% 0.1% 0.0% 0.0% 0.0% 0.0%
Mem: 8208852k av, 7953508k used, 255344k free, 0k shrd, 346292k buff
1763024k actv, 4564716k in_d, 149164k in_c
Swap: 8193076k av, 0k used, 8193076k free 6655192k cached


PID PPID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM CTIME LC COMMAND
5779 5181 boinc 39 19 185M 185M 4844 R N 93.5 2.3 661:46 0
hadam3_um_4.07_
5557 5179 boinc 39 19 57596 56M 4672 R N 92.4 0.7 661:11 1
sulphur_um_4.23
5783 5180 boinc 39 19 70076 68M 4224 R N 91.4 0.8 659:58 3
hadcm3transum_5
5781 5178 boinc 39 19 70520 68M 4224 R N 88.9 0.8 660:17 3
hadcm3transum_5
6556 5927 root 16 0 50680 41M 7672 R 14.7 0.5 46:11 1 X

Now most of these are daemons, or started by daemons. So each one has an
entry in /etc/rc.d/rc0.d or /etc/rc.d/rc6.d to shut it down. When a shutdown
starts, all these must be executed. Then the shutdown procedure sends a
SIGHUP (I think it is) to all the remaining processes. Then it waits 5
seconds (IIRC) and then sends them a SIGKILL to get them off in case they
were too rude to quit when given fair warning. I know my IBM DB2 takes a
while to shut down, but it runs more than 10 processes and must flush its
buffers to disk (and there are a lot of very large buffers in that).

I have my system set to shutdown 3300 seconds (55 minutes) after power fails
(but I have, nominally, 66 minutes of capacity. It is also set to shutdown
30 seconds after detecting a low battery condition, which is when there is
only 5 minutes reserve in the battery.

It presently reports it is running at 23.9% of capacity. UPS is APC
Smart-UPS 2200 which is about the size of a mini-tower.

BTW, I'm in Bergen County, NJ. Power was out a couple of streets away
early this morning. Took out the traffic signals as well, but our street
was OK. Trees everywhere; every town has a sign "Tree City USA". Yet all
the cables are on poles. Crazy. Why not bury the things underground?

Here is some of the recent events on my main UPS:

100401 02/27/06 07:03:13 Scheduled UPS self-test passed
200007 03/01/06 02:32:01 UPS on battery: Large momentary spike 124.8 V
100300 03/01/06 02:32:01 Normal power restored: UPS on line
200006 03/03/06 06:27:13 UPS on battery: Deep momentary sag 087.1 V
100300 03/03/06 06:27:13 Normal power restored: UPS on line
100401 03/06/06 07:03:13 Scheduled UPS self-test passed
200000 03/09/06 17:53:20 UPS on battery
100300 03/09/06 17:53:20 Normal power restored: UPS on line
200000 03/10/06 07:05:35 UPS on battery
100300 03/10/06 07:05:35 Normal power restored: UPS on line
200000 03/12/06 07:48:56 UPS on battery
100300 03/12/06 07:48:56 Normal power restored: UPS on line
100401 03/13/06 07:03:09 Scheduled UPS self-test passed
200007 03/13/06 09:50:30 UPS on battery: Large momentary spike 122.8 V
100300 03/13/06 09:50:31 Normal power restored: UPS on line

Burying things underground is very capital-intensive, and offers little in
rate-of-return for the power company shareholders. There may also be legal
difficulties since the easement for the overhead lines may not give the
power company the right to bury things in the property. I, for one, would
cheerfully grant the power company an easement for that, provided they pay
the entire cost. And it might save them the trouble of paying people
overtime to come out late at night during a sleet storm to restore the
overhead wires. But I am pretty sure a power company would have the data to
see if burying the wires would be cost effective and, if so, they would be
doing it already.

--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ PGP-Key: 9A2FC99A Registered Machine 241939.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 10:20:01 up 11:31, 3 users, load average: 4.34, 4.26, 4.18
.

Relevant Pages

  • Re: UPS backup
    ... the system will shutdown WHEN the battery reaches a critical ... The communication is fine and I can see the power load, if the UPS system is working on battery or not... ... Network UPS Tools upsmon 2.2.0- ...
    (linux.redhat)
  • Re: UPS waveform question for SUN servers
    ... O.K. The pure sine wave is better for systems with input ... The SB-1K and SB-2K use the same power supply. ... this UPS has caused more trouble than ... it has alleviated - with the battery sometimes discharging for no ...
    (comp.sys.sun.hardware)
  • Re: back-up power supply
    ... :I was thinking to get a back-up power supply for my PC and saw a book ... The least expensive type of UPS is 'passive'. ... the battery goes back out again when the power returns. ... One issue with passive UPSs is that the electricity is not flowing ...
    (comp.security.misc)
  • Re: ups installation question
    ... They protect connected devices from sudden power ... UPS AKA battery back-up: ... Most have surge protection et al, ... The issue of UPS vs. auto battery life is more a limitation of the ...
    (alt.sys.pc-clone.dell)
  • Re: UPS and surge protection for LCD TV
    ... and surge protectors for large screen projection TV's? ... features as well as standard UPS units for computer protection. ... That depends on how much power everything you are going to hook to it ... Battery cost vary greatly depending on size. ...
    (alt.tv.tech.hdtv)