keeping server parts stocked on site

[ScorpioIlya]

Well as some of you know, I work for a very small financial company...and do some IT stuff part time...we have a server that cost me $1100 from newegg....it includes 2 20 gig system drives, of which one is pretty much disabled in case the other fails, and 2 raptor drives in a hardware raid 1 config......my question is.......should I keep some parts onsite for quick replacement such as mobo, psu, and a third raptor drive to rebuld the array?...if so what % of the total cost of the server should be kept on site in replacement parts, and which parts do you guys recommend I keep...my idea is as I said:

PSU: maybe 50-60 bucks for a decent one.
WD Raptor: 110 bucks
Motherboard 865G chipset <100 bucks....so the total comes to 270 or so..which is roughly 25%..........and that would prevent me from having to pay exorbinant prices for the same parts from the local shops...where the PSU's go for 80-100 for good ones, raptor drives are like 200, and motherboards are like $150....

[TripleRLtd]

Well as some of you know, I work for a very small financial company...and do some IT stuff part time...we have a server that cost me $1100 from newegg....it includes 2 20 gig system drives, of which one is pretty much disabled in case the other fails, and 2 raptor drives in a hardware raid 1 config......my question is.......should I keep some parts onsite for quick replacement such as mobo, psu, and a third raptor drive to rebuld the array?...if so what % of the total cost of the server should be kept on site in replacement parts, and which parts do you guys recommend I keep...my idea is as I said:

PSU: maybe 50-60 bucks for a decent one.
WD Raptor: 110 bucks
Motherboard 865G chipset <100 bucks....so the total comes to 270 or so..which is roughly 25%..........and that would prevent me from having to pay exorbinant prices for the same parts from the local shops...where the PSU's go for 80-100 for good ones, raptor drives are like 200, and motherboards are like $150....

Good idea to keep ALL necessary parts on site for quick replacement...gesh, even a backup server. BTW, didn't the new server come with a warranty?

[confus-ed]

In your position, looking at your budgets, I'd say the most sensible thing is to have another machine capable of 'promotion' to this role handy & forget about spares, though a spare drive is always a blessing

You want to worry more about how feasable it is to get any replacement going again & how long that'd take, no matter what position you take on spares/redundancy ...

[FatalException0E]

If your 'small' company is anything like mine, what computers that are there are busy doing something. So if one goes, you can't have another one move in to take its place. Power supplies are easy to keep around, don't cost too much, don't change in price, and frequently go with no warning at all.

Sounds like you got a spare system drive sitting there. Is it staged and ready to run with nothing more than cable swapping? Your data is in a RAID 1 array, so if one drive goes, you should be able to run long enough to get a spare. (ship overnight in that case!)

I would keep a power supply and maybe a mobo sitting around. With technology changing, especially with mobo designs/chipsets changing like the tides, when that board decides to take a permanent vacation, it may be real hard to find a replacement on the market.

For instance, when I had to replace my dual AMD board, my SCSI card could no longer run in a 64bit slot, because I had the 'wrong' 64bit 66mhz card

[ScorpioIlya]

Good idea to keep ALL necessary parts on site for quick replacement...gesh, even a backup server. BTW, didn't the new server come with a warranty?

I built it..that's why it cost $1100....not $2500

[Escape_Driver]

Those are the only parts that I send up to my far north Servers. Any thing else can be taken out of a workstation for a quick fix... They have to have those parts or they have to be flown in at GREAT expence !!!

[gazzak]

Anything with moving parts inside always have a spare for. PSU too, stop there!

[slgrieb]

Hey, TripleR! You should know what warranties are worth these days.

[rgharper]

my question is.......should I keep some parts onsite for quick replacement such as mobo, psu, and a third raptor drive to rebuld the array?...

Every server at my shop has a 100% ready replacement on the shelf. I standardized on two basic models, so that means two complete servers ready to roll, RAM already installed. Since we use two different SCSI controllers and several different drive sizes, I don't mount those in the case but there's enough spares of each to completely replace the largest array in use.

I don't sweat the rest. If I need a floppy, CD, power supply, etc. I remove it from one of the ready-servers and order the replacement via overnight express. In the event of an absolute disaster give me twenty minutes to assemble and enough time to load from the latest offsite backup tape and I'm ready to roll again.

[TripleRLtd]

Every server at my shop has a 100% ready replacement on the shelf. I standardized on two basic models, so that means two complete servers ready to roll, RAM already installed. Since we use two different SCSI controllers and several different drive sizes, I don't mount those in the case but there's enough spares of each to completely replace the largest array in use.

I don't sweat the rest. If I need a floppy, CD, power supply, etc. I remove it from one of the ready-servers and order the replacement via overnight express. In the event of an absolute disaster give me twenty minutes to assemble and enough time to load from the latest offsite backup tape and I'm ready to roll again.

Way to go rg. And, slgrieb...yeah, I sadly do know what ya mean...the paper they are written on?
Anywho, the way Scorp talked about

we have a server that cost me $1100 from newegg

, it sounded like a full blown PC/Server with full warranty...not just parts.

[silencio]

I'd always have a spare drive on hand. If any of the other peripherials are oddballs (something you can't find locally or have onsite by the next AM), I'd keep those on hand too.

[Shalafi]

PSU: maybe 50-60 bucks for a decent one.
WD Raptor: 110 bucks
Motherboard 865G chipset <100 bucks....so the total comes to 270 or so..which is roughly 25%..........and that would prevent me from having to pay exorbinant prices for the same parts from the local shops...where the PSU's go for 80-100 for good ones, raptor drives are like 200, and motherboards are like $150....

May just be me, but I like to keep spare CPU and case fans around too. The suggestion of having spares of anything with moving parts is something I beleive in wholeheartedly. Great suggestions all around. Will have to adopt some of these other practices myself.

[kato2274]

Well as some of you know, I work for a very small financial company...and do some IT stuff part time...we have a server that cost me $1100 from newegg....it includes 2 20 gig system drives, of which one is pretty much disabled in case the other fails, and 2 raptor drives in a hardware raid 1 config......my question is.......should I keep some parts onsite for quick replacement such as mobo, psu, and a third raptor drive to rebuld the array?...if so what % of the total cost of the server should be kept on site in replacement parts, and which parts do you guys recommend I keep...my idea is as I said:

PSU: maybe 50-60 bucks for a decent one.
WD Raptor: 110 bucks
Motherboard 865G chipset <100 bucks....so the total comes to 270 or so..which is roughly 25%..........and that would prevent me from having to pay exorbinant prices for the same parts from the local shops...where the PSU's go for 80-100 for good ones, raptor drives are like 200, and motherboards are like $150....

Is your company is only running this one server? because if that's the case I'd suggest you not only stock spare parts/machines, but get another server up on the network as an additional domain controller. I'm of the opinion that in a production enviroment any domain no matter how small it is (even if there are only 10 client machines) you should have 2 domain controllers. split the dhcp scope, and let them both handle dns. If your network can be supported by one server and you have 2, you have built in redundancy right there. one domain controller goes down and aside from possibly restoring shared files from a backup you'd hardly notice a hiccup. though in the case of shares . . . .again I'd reccomend a seperate file server with a spare machine ready to replace it. of course you could also implement dfs so even the shares wouldn't be lost if a server went down.

I pretty much agree with what everyone said, just wanted to empasize that in a production enviroment you should have at least two domain controllers no matter how small the domain.

[silencio]

Is your company is only running this one server? because if that's the case I'd suggest you not only stock spare parts/machines, but get another server up on the network as an additional domain controller. I'm of the opinion that in a production enviroment any domain no matter how small it is (even if there are only 10 client machines) you should have 2 domain controllers. split the dhcp scope, and let them both handle dns. If your network can be supported by one server and you have 2, you have built in redundancy right there. one domain controller goes down and aside from possibly restoring shared files from a backup you'd hardly notice a hiccup. though in the case of shares . . . .again I'd reccomend a seperate file server with a spare machine ready to replace it. of course you could also implement dfs so even the shares wouldn't be lost if a server went down.

I pretty much agree with what everyone said, just wanted to empasize that in a production enviroment you should have at least two domain controllers no matter how small the domain.

Good advice. IMO, processor/machine speeds have reached a point where you can get a few good years or more out of a machine so it makes sense to add redundancy. Some offices are just cheap though...

[ScorpioIlya]

okay you guys are talking about 10 machines.....

Domain controllers? DHCP scope..........failover servers....are you guys freaking kidding me...........this is a financial institution with a client-based database....meaning all the query-related tasks are done by the client......internal.DNS?? COME ON PEOPLE.....its plain and simple

10 machines running win2k professional
1 machine running win2k server
all 10 assigned static Ip's
win2k server running vpn server....for remote login
linksys consumer grade router
24 port d-link switch 10/100
offsite backup onto cd's of a 50 mB datase, on a weekly basis....
internal mail server, that collects from our outside web hosting, and enables us to send huge attachments to each other and scan stuff to send to each other (the only reason for it) (it runs on a win2k box)

THAT IS ALL........okay....no PDC, NO internal DNS, no DHCP...its all plain and simple, the server runs a raid 1, with 2 more system drives.its a p4 2.4c.....desktop mobo.......that's IT......why woudl I want them to spend a buttload of money? all those things you mentioned add layers of complexity that aren't needed.

I don't even work there full time, and most of the time I spend doing other things (translation russian-> english, and internet research). They just don't need alot of IT stuff....