UPS failure

[jmak]

Morning all

 

Walked in this morning to find all around panic as "the internet's down". Looked inserver cupboard to find server and UPS with no signs of life. I bypassed the UPS and the server sprang to life - so far nothing unusual. But I then tested the power supply to the UPS and found it to be live. I then plugged the UPS into the mains again and it too sprang to life. The strange thing is that it was still fully charged. Also, there was no controlled shutdown on the server - the events log shows previous system shutdown was unexpected. I've confirmed with the caretaker (who lives on site) that there have been no power cuts over the weekend.

 

Any ideas what it could be or what I can do to find out? Or is it just that someone really hates me?

 

Thanks as ever.

[glennda]

Is it APC? - doesn't surprise me in the slightest if it is - had a few fail in similar ways but mainly when they die they die.

[Michael]

I had a very similar problem in the past few weeks with an APC UPS. It took me a while to work it out, but every two weeks the UPS does a self-test. It was failing on this self-test and forcefully powering off the server. No logs to give any clue as to why, it's as if it never happened.

 

In the end replacing it with a brand new UPS with new data and power cabling (just to be double sure) did the trick. No problems since.

[jmak]

Is it APC?

 

It's badged as a Dell UPS 1000W - but I don't suppose they make it. I'll do a bit more digging.

[Michael]

Does your Dell UPS come with software - a bit like APC's PowerChute? Within PowerChute you can manually perform a self-test and see whether or not the server goes off. I suspect the Dell software will have a similar function.

[jmak]

Hi

 

It comes with "Dell UPS Management Software" - which I knew about, but have never investigated In the past it has managed a controlled shut down and leaves me nice messages telling me what happened.

 

I now see that it has lots of options for testing and emailing results to me - although I gues it won't email me if it just dies..........

 

There are a few self test aborted messages in its logs, but 5 at random intervals over the last 18 months - all of which it aborted.

 

Now to work out an appropriate test plan.

 

Thanks

[Michael]

It sounds very similar to the problem I had. The battery is unable to take over correctly in the event mains power is lost, yet all the LEDs display everything as 'normal' status.

[jmak]

Thanks Michael - the odd thing is that there was no power loss in the school, so the whole problem was caused by the UPS......... No idea why it turned off.

[SYNACK]

Thanks Michael - the odd thing is that there was no power loss in the school, so the whole problem was caused by the UPS......... No idea why it turned off.

 

It probably encountered 'slightly' dodgey power conditions, couldn't cope and shut off completely. We had an hp one that was total rubbish for that. That is why we only get APC stuff now as it can at least do its job.

[jmak]

I'm thinking of getting a redundant PSU for the server to plug into the mains direct.........

[SYNACK]

I'm thinking of getting a redundant PSU for the server to plug into the mains direct.........

 

I'd go with a decent UPS, if your directly in the mains and you get a big surge it won't just take out the secondary supply, it'll cook the whole lot as you loose all the protection. We keep the rubbish UPSs on secondary supplies if they are still about and use a decent one for the primary. It can increase runtime too.

[Michael]

I'd agree with the above, buy a replacement APC UPS. The fact that the UPS has aborted many self-tests is an indication the battery just cannot take the load. I'd suggest you temporarily disconnect the server from the UPS and plug it in the mains to avoid any other problems and get a new UPS on order ASAP.

[jmak]

Thanks guys.

It should be well within its capabilities - it's rated at 1kW Output, 10A per socket and we only have the server connected. The maximum draw of the server should be 450W, but it shouldn't get near that as it will essentially be at idle if there's a power outage, because nothing else in the school will be running and idle is supposed to be around 120W. I've seen it running at about 15%of max load when I've been in the server room, which would match with that.

 

Trying to work out the capacity of the batteries - they don't make it obvious.

 

In the mean time......

 

Do I put the UPS back in line, or leave it disconnected? It doesn't seem to offer much protection in the event of a power outage, but it might act as a decent surge protector...........

 

Thanks

[Michael]

If you regularly have surges put it back in, but if not connect the server to the mains.

 

Buy something like an APC Smart UPS 1000 or 1500. This is suitable for most servers with a good battery backup time.