Hard Disk Drives

see also:

• Solid State Drives
• SMART – hardware monitoring for disks

Every spinning drives with pending/moved sectors or uncorrectable errors MUST be replaced.

Most drives can pull about 40k-50k power on hours. To keep things fresh, you SHOULD phase them out at 35k hours (=24h/d*365d/y*4y) or 4 years. Then they can be used as archive HDDs or cold spares, but SHOULD NOT be trusted with data you only have once.

If you feel fancy, you CAN replace disks at 20,000 hours.

• Make sure your drives are coming from different retailers so you are not getting drives that are built in the same batch.
• Don't assume your HDD will not fail in X amount of hours. Back them up and put them in a RAID to minimise downtime.
• Do a full SMART test out of the box for any drive that is going to be used in a RAID group.
• Run a high level IOMeter test against new drives for 24-48hours, then rerun the SMART test looking for any pre-fail stats.
• Do not mix and match RPMs, slower RPM drives will slow down access times to the faster drives on the same controller when heavy IO hits the slower drives.

Both your mainboard and your HDD have to support hot swapping.

Even if your mainboard generally supports hot swap, your mainboard may not support it on all ports. Similarly, if your disks themselves don't support hot swap, you can't hot swap them even in a mainboard which supports it.

Swapping disks while the PC is booted should only be done if you're absolutely certain both the mainboard and the disk support hot swapping properly, else you may damage the disk. Especially consumer-grade disks are susceptible to failures in this way.

Being on standby may simply mean they have parked their r/w arms and are still spinning. It does not mean you can hot swap the disks.

No, you should not shout in the data centre. If you do, your disk arrays might throw read errors.