Sometimes, a failing power supply may not cut out immediately. It may, over time, start fluctuating the power it is sending down to the various components within the server.
When a server is behaving badly, erratically, or strangely, the power supply is sometimes the last thing on the technician's mind.
We will look at things like memory, or perhaps a set of bad sectors on one of the RAID array disks, or even the server board itself before considering the PSU.
For those of us who have been around long enough to see a few of the irratic behaviour types caused by a failing PSU on either a server or workstation, the problem gets a little easier to figure out by the server's behaviour patterns.
For servers without redundant power, things are a little more difficult. We will need to find a like PSU that is matched to the server we are working on. In some cases, we may be able to get away with plugging in a Delta or like PSU with all of the right connectors.
In the case of proprietary systems though, we are looking at trying to purchase a PSU from the manufacturer and find out they are in that mysterious class of metal: unobtainium. That is, they are not made anymore, and may not even be found on auction sites and the like.
The critical aspect to no redundant power is the need to down the server while troubleshooting or changing out the PSU. This can be sticky for clients who are into their peak season or just plain busy.
A server that is behaving erratically reflects badly on us and our abilities to provide a trouble free I.T environment for our clients.
Given the cost difference between a chassis with the redundant power option (plus the extra PSU) over a standard chassis with only one PSU, we suggest redundant power as the default choice for server configurations. As with any client related situation, your mileage will vary. ;)
Here is a screen shot of the Intel SC5300LX series chassis with both PSUs installed:
And, here is a screen shot of a similar setup in the Intel SC5295/9BRP chassis:
Intel SC5300LX with Redundant Power Supply Installed
The SC5295BRP was the previous generation to the now in production SC5299BRP chassis.
Intel SC5295/9BRP with Redundant Power Supply Installed
Some of the benefits of redundant power on the server:
- The ability to pull one or the other PSU module while the server is online to eliminate the PSUs as a possible source of trouble
- The ability to move the A/C power cords, one at a time, between power sources such as when replacing or upgrading an existing UPS while the server is online
- The ability to replace a defective PSU while the server is online
- The ability to test the server monitoring software's ability to detect a faulty PSU before the server goes online at the client by pulling each PSU one at a time here in the shop
When justifying the extra server cost to our clients, we can put the matter forward about keeping the server online while working on the power modules, making changes to the A/C infrastructure, UPS units, and more.
To explain downtime costs to our clients, we use the arbitrary number of $100/hr per user on the network as far as what the theoretical costs to the company are for downtime.
For some clients that are in key industries, those costs can approach $1,000/hr per user.
On top of that, there is the cost associated with having a technician dispatched to diagnose and take care of a power related problem on the server. Again, the PSU may be the last thing on the technician's mind.
Once there are some concrete numbers in front of the client as far as the benefits versus the up front costs, we see most of our clients go with redundant power on servers. Redundant power is a little extra insurance that pays off huge in the event of a problem.
In our experience, there is no better way to outfit a server setup.
Microsoft Small Business Specialists
*All Mac on SBS posts are posted on our in-house iMac via the Safari Web browser.