UPS Auto shutdown

[Don Spare]

I have 2 L-class servers and a Netserver LH300 in a rack protected by a PowerTrust II A6584A 9KVA UPS. I have the UPS Manager software on each of the servers. The display shows all 3 servers are communicating. I have the output of the UPS connected to the appropriate plug on the console serial line.

This past Saturday I attempted a test. I shutdown all the databases (just in case) then pulled the plug on the UPS. It immediately swithced to battery mode but nothing else happened. I waited 45 mins before aborting the test. At that time the battery life indicator was showing 50% remaining.

Why did the UPS not send a notification? Or if it did, why did the servers not shutdown as I expected? Maybe I have the wrong parameter set. I was expecting the "ONBattery" event to trigger this. Is that correct?

What else can I look at?

[Pete Randall]

What have you got in your /etc/ups_conf file?


Pete

[Sandip Ghosh]

You also need to install EMS for monitoring the Unix Servers so that it can email/page you about the faliure.

Sandip

[Don Spare]

My /etc/ups_conf file contains:

# @(#) $Revision: 70.4 $
shutdown_delay_mins:1
shutdown_timeout_mins:5
upstty:/dev/tty0p1 # Port 1 (second) on terminal mux 0 for UPS rs-232


How do I tell if I have EMS installed?

[Pete Randall]

OK, you should have initiated shutdown after 1 minute, and done a hard shutdown after 5 minutes (in case the graceful shutdown hadn't completed).
Check /var/adm/syslog/syslog.log for ups_mond messages:
grep ups /var/adm/syslog/syslog.log

Pete

[Sandip Ghosh]

I think your UPS connection has not been enabled. First check whether the serial cable has been connected to the UPS and the Server. Open the inittab file and see that whether the line with UPS is a # infront of it. If the # is there , remove it. Give init q to restart the getty. then do a ps -ef|grep upsd. See the UPS daemon is running or not.

For installing the EMS, install latest Online Diagnostics. OR you can give swlist -l fileset|grep -i ems.

Sandip

[Don Spare]

Okay. Here's the latest:

No messages in syslog.log.

inittab was commented - I removed it. I will not be able to issue the init q until after work hours. This is on the production server and my boss gets nervous about these things being done when there is the possibility that customers may not be able to access the server/DB. I'm going to go back thru the UPS Manager manual to see it I missed this part.

I'll update this after I have a chance to test these suggestions.

Thanks.

[Pete Randall]

Don,

I was thinking maybe it failed to start but it sounds like it never even tried to start. You should be able to see a process called ups_mond running when you do "ps -ef |grep ups". Let us know.

Pete

[Kelli Ward]

Hi Don,

You could consider simply starting the ups_mond daemon without rereading the inittab. You can then run your test on the UPS and see how it works.

By default that line is commented out so you will probably have little worries about how well it works.

Just make sure you have the conf file set up how you want it.

Good luck,
Kel

[Darren Prior]

Don,

Can you confirm whether you are using ups_mond or UPS Manager.

regards,

Darren.

[Don Spare]

Darren,
I am definitely using the UPS Manager II software.

[Darren Prior]

Hi Don,

As you're not using ups_mond, you don't need to worry about the ups_conf file at all. I'd also ensure that you DON'T have the ups_mond line in inittab enabled, as having both ups_mond and UPS Manager II accessing a UPS will cause problems.

Other things to check - if you are expecting the system connected to the UPS to alert the others via a network connection then ensure the hub used is protected by the UPS too!

Personally I know very little about UPS Manager II, but there are a number of documents in the knowledge database with troubleshooting guidance (you'll need to have your contract details linked to your ITRC user id to view these). Here's a couple of docs to get you started: KBRC00008135 (How to configure UPS Manager II for two systems and one UPS), UUPSKBRC00009437 (HP UPS Manager II - shutdown alarm timing explained)

You may also wish to post your query in the Windows forums on the ITRC if you need help with the NetServer part.

Hope this helps,

Darren.

[Bill Thorsteinson]

I have had problems communicating with the UPS
with SNMP passwords set to a non-default (secure) value.

Run throught the upsmgrclient
setup and see if you can
get the status of the UPS. If the client can't then the
ups manager software likely
can't either.

Given that you appear to have
well over an hours worth of
battery power, I would set the
delay before shutdown
significantly higher than
1 minute.

[Don Spare]

Okay. Here's where I am. I have commented the inittab file to prevent ups_mond from starting. The attached file shows the UPS Info display from UPS Manager II console. Note that UPS Time, Battery Voltage, UPS Load and Input Line indicators all show 0 (zero). The "Active Alarms" shows "communications lost". I take that as an indication that the UPS Manager software is not actually communicating with the UPS. If that is so then how can I change that status? I'm guessing that if I can get that status changed then the rest should start to work.

[Don Spare]

I've gone throught eh proxy setup and it still doesn't work. I also notice that mgupsd is not running and my attempts to make it start only get a usage message so I'm probably doing something wrong there too! The manual says that to stop the ups daemon issue the command: mgupsd -k
Taking that for a clue I tried mgupsd by itself and got the usage dump of:

run: upsd -d -n [ -i -s ]
dump shared memory: upsd -a
kill all proxy daemons: upsd -k

So I took that to mean that I need to provide some arguments to the daemon to get it to run. I assumed that to mean details from the proxy setup so I issued:

mgupsd -d/dev/tty0p0 -n1
MGProxy: UpsId: 1, Device: /dev/tty0p0
Error: cannot set terminal attributes

So now it is complaining about terminal attributes.

Where do I go from here?

[Gavin Clarke]

I guess you've fixed this now. I recently had a power cut, no evidence of a shutdown.

I have a very similar situation to the one you describe.

After rooting through the manual I found that in chapter six it mentions the amber light. Then it suggests adding a managed node and there's a picture with a tick box saying UPS Manager control.

This you can get to from
/opt/upsmgr/manager/bin/upsmgrmap

I found that by making that box darker (depressed) then running

/opt/upsmgr/manager/bin/upsmgrstop
/opt/upsmgr/manager/bin/upsmgrstart

The traffic light went green (I might have stopped /opt/upsmgr/client/bin/upsmgrclient & then restarted I'm not sure)

I haven't tested it yet, it certainly looks more hopeful than it did. I guess it's all in the manual i just found this bit a tad ambiguous.

[Don Spare]

Gavin,
It turns out I had the wrong UPS configured. I somehow had the Manufacturer and Model set to something that didn't match my actual UPS. Once I figured that out, I reconfigured so the Manufacturer/Model was Hewlett Packard/HP R09000. Then I started seeing status on the UPS - Info window that I hadn't seen before. I managed a 'pull the plug' test in December and it worked fine on the Unix boxes after I figured out where the delays were set. Note: even though I also configured the Netserver 3000 as part of this UPS group, it did not shutdown gracefully. It simply died when the battery ran out. The 2 Unix boxes shutdown gracefully after about 8 minutes of delay.
To check if you are communicating with the UPS go to the UPS Domain Map, UPS -> Info window. If you see UPS stats (colored bars on right, Active Alarms other than 'Communications Lost') then you are communicating and the rest should be a matter of configuring the separate alarm actions.

Hope this helps.

Don

[Gavin Clarke]

It does, the status bars are there. I can test it with some confidence now.