OK, maybe this is a stupid question, but....

I see some of you guys talking about sending your servers off to some colo somewhere. I'm asuming that some/most of these are from hosting companies. What do you guys do if/when a drive fails at 3am? I'm just trying to determine the point at which it's "safe" to go colo as opposed to a managed or dedicated server.

Originally posted by rbro
OK, maybe this is a stupid question, but....

I see some of you guys talking about sending your servers off to some colo somewhere. I'm asuming that some/most of these are from hosting companies. What do you guys do if/when a drive fails at 3am? I'm just trying to determine the point at which it's "safe" to go colo as opposed to a managed or dedicated server.

Depending upon the services offered by your colo facility, You'll either keep a spare drive(s) in their hands or they will sell you a replacement. Either way, You'll be paying them labor charges per hour for them doing the repair for you.

I recommend you look into http://www.wholesalecolo.com

If moving in a colo box and harddrive failure is a major issue, (or potentially a issue), i'd suggest you spend the extra $200 or so and put in a raid card and a extra hdd. 2x40gb raid 1 should cost under $300 for both drives and the card, and is fairly easy to configure (make sure you get promise, or highpoint controllers, because if you run *nix like most of us, they're typically the easiest to configure). Only bad thing about that is if you're like us, we chose highpoint (they're generally better known for linux compatability), but the highpoint redhat drivers wouldn't allow you to boot from the raid device, so we moved the majority of such systems to FreeBSD 4.5 (without issue, or need for special drivers i might note).

Hope that helps, if 1 drive in a raid 1 configuration fails, the other will keep on chugging without any issues.

OK, so I guess my next logical question would be what about other emergencies? Are there colo's that essentially have spare parts on hand for any forseeable situation? Will a colo know if a drive has failed at 3 am? Is it more cost effective to go colo (factoring in average potential emergency support/hardware replacement situations) as opposed to managed dedicated?

Well, it's more important that you build a redundant system, put in RAID and 2 drives, physically spaced (not one on top of the other). Most colocation facilities will NOT have spare parts for any forseeable emergency. There are just too many combinations of hardware that people can use for that to be reasonably expected, they may have some idle ram, some idle drives, but more then that i wouldn't hold your breath for. Most facilities if they dont have it, if they're dedicated to their clients will run out to walmart or whatever it takes and buy you a new part (if it's relatively common) at your expense though.

I have to second Myles. RAID card will benefit you in time of hard drive failure. You may think you don't need it, until your harddrive crash :)

In 2001, my IBM drive crashed 3 times, and if I didn't have RAID in there, I will lose a lot of customers. I would highly recommend the RAID system, even if your colo is only 10 minutes away. :)

thanks rey,

Rbro, if you decide to go raid, theres another thing i forgot to mention, if you're doing a non-rackmount setup (a lot more places are taking towers nowdays) don't forget to add redundant cooling fans, and put some through the motherboard, some right from the power supply. If you need any help picking out or configuring a raid setup (or purchasing) let me know as i've recently purchased and setup a few raid configurations.

For redundancy, aside from a redundant power supply, i can't think of anything else, and i really wouldn't recommend a redundant power supply (because of sheer cost).

Originally posted by porcupine
Well, it's more important that you build a redundant system, put in RAID and 2 drives, physically spaced (not one on top of the other).

porcupine, a totally redundant solution is not possible (unless you have a fail over server cluster). How about motherboards, CPU, RAM, power suppy? The easiest to handle is HD where you can have RAID 1, but the other hardwares can fail as easily as the hard drive.

The more I see into this, the more I think choosing a data center near you that allows a physical access 24/7 is the most important thing to consider when choosing a colo place. If your colo provider can't fix problem at reasonable time, then hopefuly you can.

cheers,
:beer:

I never said fully redundant, just as redundant as you can reasonable make it, raid, and lots of cooling (and a fan directed at the heatsync, in case the heatsync fan dies) is about all you can do for most configurations, but every little bit helps.

Originally posted by dektong

The more I see into this, the more I think choosing a data center near you that allows a physical access 24/7 is the most important thing to consider when choosing a colo place. If your colo provider can't fix problem at reasonable time, then hopefuly you can.


Great idea if you live in a big (and well connected city). There actually are folks that live outside of the concrete jungles. ;)

Regards,

Jeff