-
Notifications
You must be signed in to change notification settings - Fork 0
IT Mistakes
John Pellman edited this page Mar 18, 2020
·
16 revisions
[[TOC|levels = 2]]
- Typically a symptom of not reading the manual for a shell command first to totally grasp the flags and options.
- A great many mistakes of this variety (and of other classes) are issues of communication. You cannot embrace knowledge/information if you do not first encode it. This occurs in other contexts as well (such as not reading e-mails carefully enough).
- This can also be a symptom of inadequate testing.
- If possible, do a dry run of a command beforehand (i.e., rsync's '-n' flag) or use 'echo' before a script to make sure that the variable has the correct values.
- Also test results afterwards- make a checklist of what you expect an output to contain and what it should not contain.
- General advice: Avoid using cd with relative paths in scripts (i.e., be wary of 'cd ..'). If a glitch occurs in anything before the 'cd' command, or a previous cd command before the relative path behaves peculiarly, you may end up in an unexpected directory.
- Inadequate planning before putting together a script.
- Inadequate logging for a script or command (e.g., should have a text file with status of various steps piped to it; send time for file sync).
- Lack of consideration for how script may operate across environments (i.e., different versions of Linux, etc).
- You cannot change ownership of a file or directory from a normal user to root easily (unless you are root). If you are changing ownership to root (or any other user) as a normal user, it will not work.
- Don't do uploads over Wifi when Ethernet is possible. Don't use Ethernet when Infiniband makes sense.
- Make sure that cords aren't plugged in in a way that is so mangled that unplugging one cord could accidentally power off another device.
- In some instances a user can be educated. Other times, it is better to apply a solution without educating the user, and then trying to educate the user later. Example: A user is agitated and under a lot of stress and needs the solution immediately.
- Inadequate communication with users (getting user needs) before beginning a purchase process.
- Performing a reboot without consulting users.
- Be mindful of the potential need to document changes in directory structure on a file system (redirecting the output of 'tree' or 'ls -R' to a text file might be a good idea here).
- Inadequate documentation of licenses before a migration to new systems.
- Inadequate planning before a migration. Not checking if and how licenses can be migrated between systems.
- Upgrading a system without a checklist.
- Upgrading a system without a rollback plan. Not having a time scale for a rollback plan.
- When transferring via Sneakernet (still occurs in rare situations): Not creating a manifest of files to be transferred before transferring them to a physical medium. Without such a manifest, there's no way of keeping track of whether or not all the files that were intended to be transferred actually were.
- Don't assume that because a bay can be taken out of a server, it is plug'n'play. You will be burned on this if you do not bear this in mind. Server hard drives are not to be treated like USB hard drives you may use with a desktop.
- dd if=[block level device or image] bs=[block size A] | pv | dd of=[block level device or image] bs=[block size B] conv=sync is a great way to corrupt an image. It's always a good idea to a) make sure you understand what a flag is doing (conv=sync) rather than trusting the documentation it came from blindly and b) make sure that the block size is the same on either side of a pipe using multiple dds.