Battery really needed for SmartArray with Write-Caching enabled?

[24x8]

I wanted to get some feedback on this, to be sure I'm not missing something. We have P400i SmartArrays in DL360 and DL380 servers, all of which have dual power supplies connected to redundant power infrastructure, so the likelihood of an unexpected power failure is very, very slim. Given that, is it really even necessary to have a battery installed on the SmartArrays (with write-caching enabled)? Assuming that there is no meaninful chance of a power failure, can I just remove the battery, or does it serve some other function I'm not aware of?

Also, we run with the "Physical Drive Write Cache" option enabled. As I understand it, this option also essentially relies on "no unexpected power outages", since the drives themselves don't benefit from the battery (and don't have their own), so this seems to further support the idea of running without the controller battery.

The thing I can't quite figure out is why HP decided a battery was needed in the first place. The only thing I can come up with is that not all server models that can support the SmartArray controller necessarily have redundant power supplies and/or that not all installations might have redundant power infrastructure.

Note that the purpose of removing the batteries is simply to cut down on maintenance--there are a substantial nubmer of servers of varying ages, and we typically have to replace at least one or two batteries per month. Not a huge deal in the grand scheme of things, but if we're not getting any real value out of having the batteries in the first place, why bother?

[Dennis Handly]

>can I just remove the battery, or does it serve some other function I'm not aware of?

I'm not sure there are many setups like yours.  Also the write cache is required when changing RAID level.

I.e. if you take the battery out, you won't have a write cache nor other features.

[24x8]

Hmmm. I would assume that almost anyone with servers at a modern data center would have redundant A/B power as a baseline standard.

In any case, the controller configuration has a setting that enables write caching when the battery is discharged or missing. Does enabling this setting not "restore" all funtionality regardless of the battery state?

[Dennis Handly]

>Does enabling this setting not "restore" all funtionality regardless of the battery state?

I wouldn't think so.  I've seen other topics where it was stated without the battery, certain functionality isn't available, like RAID 5.

[waaronb]

When I recently placed an older server in a smaller datacenter (shared cabinet) for a customer, I was surprised to find that many/most only had a single outlet unless you paid more for another.  I did find one that had redundant supplies standard and that's the one I went with.   I'm guessing most small colo needs are for cheap servers or something.  The idea of RPS seemed foreign to many of the hosts.

Anyway, regarding the battery, yeah, you can enable write caching even if the battery fails, but as mentioned, other things need a battery or they won't work, like expansion, level migration, etc.  I assume HP does that because they really want to ensure there's a battery caching stuff during such heavy work.  It must make a lot of use of write caching.

I don't know that I'd feel comfortable running a write-cache without battery backup on a production level server.  Assuming the default 75% write cache (or is it 80% now?), you could have a lot of unwritten data in your Smart Array cache.  Sure, there are redundant supplies, diverse power feeds, but there are those extreme times...

We had a cabinet in NY a couple years back when Hurricane Sandy hit.  The datacenter (Internap NYM008) lost utility power and they went to battery, then the generators started up.  That worked great for the first couple of days, but then their generator fuel lines got clogged and the entire building shut down.  No generator, UPS only lasted a few minutes once the gens died... you get the idea.  First one circuit failed, then the other a few seconds later as the batteries gave up the ghost.

[24x8]

Thanks for the feedback. And yes, of course all of those fringe conditions could happen, but it does seem a little odd to me that HP would make this decision for me. I'm a big boy and should be allowed to use any feature of the SmartArray if I want to. I can see having default settings in place for "safety", but not being able to override them is just silly. Maybe HP thinks its customers aren't able to fully understand the risks and make a calculated decision? Or maybe they just want to sell $50 batteries every few years?

[Dennis Handly]

>Or maybe they just want to sell $50 batteries every few years?

I assume the FBWC is cheaper and lasts longer?

[24x8]

Not sure, but based on what I've been able to find out, the P400 only supports the BBWC. I had originally tried to find a "supercap" to replace the NiMh battery, but I can't find anything that would indicate that the P400 can support that either.