-
-
PERC Raid Controllers, Power Spikes and UPS’s with low batteries. A pitfall best avoided.
-
So, for the last few days I’ve been hearing a cart alarm go off at random intervals. I thought it was just a car down on the street, as we are several floors up and it sounded faint.
So like any other time anyone hears a car alarm go off, I ignored it.
But then it happened again at about 7:30pm, when there are never any cars parked out the front of the building.
The only other guy here had music on, so I thought it might be part of the Techno Soundtrack to Working Life album he was listening to and asked him to turn it off.After a minute of walking around and stopping to listen and looking like a goose I realised it was coming from inside the server room.
After another minute of trying to locate this alarm (the high frequency chirping makes it very hard to pinpoint) I had worked out it was coming from one of the UPS units we had.I went away and read the manual, as by this stage the chirping had stopped again.
The manual said that this specific birdsong indicated that the battery was pretty much dead. It recommended hitting the test button to see if there actually was any charge. It assured me that if the test failed the mains power would still be up, so no machines would go down.
This was true in a way.The next day when I heard the chirping again, I went in and hit the test button. None of the machines went down, but the Dell suddenly went into read only mode, and half the disk array started flashing bright orange. Having seen this before with the tape drive incident I thought that there must have been a voltage spike or something when the UPS switched over to mains power, and the PERC just shat itself.
I had to reboot and go into the PERC bios and force all the drives back online again. Luckily this was during lunch, so no one noticed.
But what shits me is that you go to the expense of setting up a RAID controller with mirrored drives, dual power supplies and a UPS and as soon as you get the tiniest fluctuation in voltage all the drives report as dead and you have to power it down to force them online. What’s the bloody point of all that redundancy if it all hinges around one shitty card?
A friend who is also in the IT industry suggested that Dell deliberately makes the PERC crap in order to get you to buy their more expensive SAN gear.
I’m starting to think that’s probably the case.
-
-
Copyright © 2008 Mike Brown. Site Design by John Kung.