Course of action summary

Step 1 - Initial size-up

• Possible cause(s) of the failure.
• Number, size, and location (which physical disk(s) contains the volume?) of the affected volume(s).
• Number, size and location (on which controller/channel?) of the affected physical drive(s), if any.
• Overall host machine condition (memory functional? RAID controller(s) functional? any degraded but mountable RAID(s) need backing up immediately). This is needed to determine if it is feasible to continue recovery on the same machine, or the affected device(s) should be attached to some other (known-good) system.
• Approximate amount (size; number of files), file types and importance of the data lost.
• Availability and state of the backup copies of that data, if any.

Based on the size-up results, we typically assign two "scores" to each case, each based on the binary scale:

• Understanding of the situation, which can be either "clear" or "unclear". It is often counterproductive to troubleshoot an "unclear" situation. Reasonable allocation of effort would be to continue gathering information until eventually everything becomes "clear".
• Damage severity (how valuable the data is? are there any backups?), either "minor" or "major". This has little effect on the recovery process (since all the procedures do not differentiate between valuable and replaceable information), but provides some timeframe/budget constraints.

Step 2 - Evaluation of available resources

• If there is a local computer expert available, you may want give him a call and ask for assistance or for a cross-check. You should specifically ask him to cross-check the information collected during previous step.
• Check if you have some "spare" computer(s) available (if the original machine is for some reason unstable, you will need to connect hard drive(s) to that "spare" machine).
• Check if there is plenty of disk space available to hold the data recovered. We found two approximations useful: a). half the size of the damaged volume, and b). twice the size of the data you're going to recover. When planning, keep in mind you will most likely need the largest of the two.

Step 3 - Setting up

Second important thing at this point is to decide what hardware reconfiguration is required. Typical scenarios are outlined below:

• Most widely used approach is to get the failed drive out of its host machine and put it into some known-working system. Additional (good) hard disk(s) might be needed depending on a free space availability.
• If there were more than one drive in the failed system and the system is still reasonably operational it may be possible (subject to free space requirements) to perform recovery without any hardware reconfiguration.
• If there is some special controller involved (e.g. RAID controller, or a SATA controller), it may be required to relocate the controller along with the damaged hard disk(s).

Whatever method is chosen, the following points should be considered:

• Ensure adequate power supply.
• Double check proper cabling, (SCSI) termination, and jumper setup.
• For IDE drives, check proper addressing mode (LBA vs. CHS/LARGE) configuration in BIOS.
• Ensure adequate cooling (ambient temperature higher than 25 deg C? any physical malfunction identified/suspected?); install additional fans if deemed necessary.

Step 4 - Recovery run

• If the machine performing recovery freezes at some point, check its memory (using some memory testing tool like GoldMemory). If memory turns out good, disk and/or bus controller(s) may be at fault (older-generation Promise controllers/drivers are known  to lock up under load - check IRQ sharing).

Step 5 - Evaluation of the results

Step 6 - Cleaning up

• Take care when rolling back any hardware changes you have made.
• Physically malfunctioned drives should not be placed back into the production environment. Replace them under warranty. There are few alternative uses of the physically damaged hard drive - we use them as a large (say, 40GB) floppy disk to transfer large amounts of data machine-to-machine, but special precautions should be taken to ensure that the backup copy of the data on the unstable drive is always readily available.