Human factors in availability of systems

Usually only 20% of the causes of failures are technology failures. In 80% of the cases, human errors are the reason. For instance, a system administrator accidentally pulls a wrong cable or enters an incorrect command. Users sometimes delete inportant (system) files.

Of course it helps to have highly qualified and trained personnel, with a healthy sense of responsibility. Errors are human, however, and there is not a cure for it. End users can introduce downtime by misuse of the system. When a user for instance starts the generation of 10 very large reports at the same time, the performance of the system could suffer in such a degree that in fact the system is unavailable to other users.

Also when a user forgets a password (and maybe tries an incorrect password for more than 5 times) he is locked out and the system is unavailable for him. If that person has a very reponsible job, for instance approving some steps in a business process, being locked-out could mean that a business process is unavailable to other users as well.

Most unavailability issues however are the result of actions from system managers. Some typical actions (or the lack thereof) are:

• Performing a test in the production environment (hopefully by accident - testing in production is of course not recommended at all)
• Switching off a wrong component (not the defective server that needs repair, but the one still operating)
• Swapping a good working disk in a RAID set instead of the defective one
• Restoring the wrong back-up tape to production • Accidentally removing files (mail folders, configuration files, database files)
• Incorrect changes to configuration files (for instance the routing table of a network router, or a change in the Windows registry)
• Tripping over cables, creating a broken or disconnected cable • Incorrect labling of cables, later leading to errors when a change is performed
• Stopping an incorrect virtual machine (the one in production instead of the one in the test environment)
• Making a typo in a system command (in UNIX: sudo rm -rf / *.back instead of sudo rm -rf /*.back where one space too many leads to a complete erasure of a hard disk - did you notice the difference?)
• Insufficient testing, for instance the fall-back procedure to mover operations from the primary datacenter to the secondary was never tested, and failed when it was most needed
• A system manager or architect made a mistake in the design of the infrastructure, leading to downtime (we thought the Windows cluster was designed in a good way, but when one of the cluster nodes failed, we found that the complete cluster went down)

Many of these mistakes can be avoided by using proper system menegement procedures, like have having a standard template for creating new servers, using formal deployment strategies with the appropriate tools, using adminstrative accounts only when absolutely needed, etc.

When in some UNIX environments the user works under a administrative account (root), automatically he gets the following message:

We assume you have received the usual lecture from the local System Administrator. 

It usually boils down to these three things: 

#1) Respect the privacy of others. 

#2) Think before you type. 

#3) With great power comes great responsibility. 

I think this message makes people aware, leading to fewer mistakes.

This entry was posted on Dinsdag 28 Juni 2011