0

I'm in the process of developing procedures for my company of what would happen if the primary server went offline or something like that. We have a server running right now live, but if it were to go offline I'm not sure what our expected down time would be or exactly how we'd go about getting it back up. I'm working on determining the RAID configuration/setup here: Show hard drive setup in linux

All I know about the server is that it is a HP Proliant DL580 with 8x Intel(R) Xeon(TM) MP CPU 3.00GHz (32 bit) and 16 GB of RAM.

With that many processors, is 1 power supply enough? There are currently 2 power supplies connected to the machine, if it requires 2 then I don't have any power backup if 1 were to fail, but if it likely only requires 1 then I should be good to go if 1 were to fail.

I'm using JungleDisk to backup the server configuration. Ideally I'd like to be able to take a 2nd server and just plug it into the data center and be good to go. The second server we have now is not nearly as powerful as the one we're running live, but it would get the job done.

What would you guys suggest I do here?

5
  • The DL580 only supports 4 processors, so I assume you mean you have 8 cores, correct?
    – EEAA
    Nov 30, 2010 at 16:26
  • What is the wattage of the power supply? How many hard drives do you have in the server?
    – EEAA
    Nov 30, 2010 at 16:27
  • HP, like most large vendors, really doesn't sell servers with multiple non-redundant power supplies. Unless your PSUs are broken or running out of spec, or you've constructed a Frankenstein server well outside your warranty terms, you'll have plenty of power. You should test that, however, and unplug one PSU while the server is running, plug it back in, then unplug the other and make sure that you can fail both ways. (Obviously, don't do this while the server is doing anything important.) Nov 30, 2010 at 16:59
  • The wattage I'm not sure about... 2 hard drives... when I run /proc/cpuinfo it shows 8 processors
    – Ben
    Nov 30, 2010 at 17:20
  • cpuinfo shows all "execution units" for lack of a better term, which includes all cores and accounts for hyperthreading. For each 'processor' listed the field "physical id" shows which physical processor it is on (starting at 0) and "core id" shows which core it is on that physical processor.
    – Scott Pack
    Nov 30, 2010 at 17:53

3 Answers 3

1

What you're doing is a very basic form of disaster recovery planning. There are entire books on this which don't completely answer the question.

Basically, it comes down to this: what does downtime cost to your business? Based on the likelihood of downtime, how much should you be spending to avoid it? What failure scenarios do you anticipate, and how can you protect against them?

The best way to approach DR is the same way you would approach any IT operations problem: understanding your systems top to bottom. If there's any piece where you don't understand how it works, learn it. You might have to build a virtual machine or a new system and cobble together bits and pieces of the old system until it works.

Having said that, there's all kinds of technologies that will help you avoid long-term downtime. SAN boot, bare metal recovery, virtualization, hot-spare and cold-spare parts, etc. all contribute. Some of them can and should be used together, others shouldn't. Before you can figure out what provides the best bang for your buck, you need to figure out what the bangs and the bucks are. You need a budget and a strategy.

0

PSU redundancy: So first thing is you got a powerful server and you want to know if one PSU is enough even with 8 cores. it should be enough, in the back of your server, you will see two plugs and both of them is the redundancy of the other one. Ideally, you should plug each of them in a different power socket for real redundancy.

RAID configuration: I think you should go with RAID 1. It is very rare that both drives will fail at the same time. Just be sure to monitor your server for bad parts. Do not ever rely on just one drive especially on production.

Backups: so you might want to get a better understanding on backups is and redundancy. In our case, backups are something that you can move around and does not rely on the main server to be up or not. It is something that we rely on if we want to rollback to a specific date. Since you are already doing this, its really good for you.

Redundancy is what you would like to implement to ensure high up time on your server. There are many approach to this from clustering to hot backup. It would really depend on your requirement and how your application is deployed.

Let me know if you there is anything I missed and you are looking an answer for.

Thanks

0

First off, you need to figure out how big your downtime window is, and what their expectations are. The window and their expectations will determine what technologies are used to reduce downtime.

Secondly, see how long it takes you to restore the server on the second box from the ground up. This is your baseline.

Third, figure out how you can meet the window and reduce restoration time. Also, figure out how much they want to spend. This is where you negotiate uptime versus cost. You can get crazy by buying vSphere with a SAN, or you can run some scripts every couple of hours and ship the data to a warm spare.

Fourth, run DR drills to make sure everything works.

The server PSUs should be redundant and hot-swappable. I would call HP to be sure and test it in a maintenance window after calling them.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .