IBM AIX Administrator common mistakes
IBM AIX Administrator · Understand the traps and how to avoid them.
Editing /etc/inittab and expecting the change to take effect immediately, or forgetting to run telinit -q to re-read the file after adding a respawn entry.
Candidates confuse inittab with a live process table; init reads inittab at boot and on telinit -q, so edits sit dormant and the candidate then misdiagnoses the missing service as a broken script.
What to do instead
After editing /etc/inittab, run telinit -q and verify with lsitab and ps -ef that the respawned process exists; use mkitab/chitab/rmitab rather than raw editing so syntax errors are caught.
Using the wrong boot device or skipping SMS when asked to boot a system into maintenance mode from an installation media, then reporting that the media is defective.
On POWER systems the boot list order and the F5 key path in the SMS menu are procedural steps candidates have read about but never performed, and the firmware silently falls through to the next device in the bootlist.
What to do instead
Enter SMS (F1 on the console or via HMC's boot to system management services), set the boot list to the CD/DVD or NIM device, exit and boot; confirm the device in the boot list with bootinfo -b before rebooting.
Treating a filesystem as full when df shows space free, without checking inode usage or deleted-but-open files.
df -k reports blocks; candidates forget df -i for inodes and forget that a process holding a deleted file keeps the blocks allocated, so they resize the LV unnecessarily or declare the disk failed.
What to do instead
Run df -i to check inode exhaustion and use fuser -dV or lsof to find processes holding deleted files; restart the holder to release blocks. Only then consider chfs -a size=+N.
Growing a filesystem with chfs -a size=+N without first confirming the logical volume has free physical partitions or the volume group has free PPs.
The command fails with a cryptic error and candidates assume chfs is broken rather than recognizing that LV size and VG free space are separate constraints.
What to do instead
Check lslv for MAX LPs and PPs, and lsvg -l for free PPs; extend the LV with extendlv first if the LV is at its maximum, then run chfs.
Creating a volume group on a disk that still carries a PVID or an old VGDA, and ignoring the resulting warning or failure.
Disks pulled from another system retain signatures; candidates skip the cleanup step and then see the disk appear as a member of a foreign VG or fail mkvg with a device-busy style error.
What to do instead
Run lspv to check for a stale PVID, then chdev -l hdiskn -a pv=clear (or reducevg the foreign VG) before mkvg; verify with lspv that the PVID field is 'none' first.
Configuring NIM to install a client but not verifying that the lpp_source and spot are at compatible AIX 7.2 levels, or that the client's network is defined on the NIM master.
NIM setup is multi-object (network, lpp_source, spot, mksysb, client); candidates define some objects and then blame the network when the install aborts early.
What to do instead
Use lsnim -l to confirm each resource's location and Rstate=ready, run nim -o check on the lpp_source and spot, and confirm the client's network object matches the subnet the client boots on.
Changing a network interface's IP with chdev or smitty while the interface is up, then assuming the change is live.
Many adapter attributes require the interface to be detached or the change is deferred; candidates then test with ping and conclude the change failed or the network is down.
What to do instead
Use chdev -l en0 -a ... only after ifconfig en0 detach (or use smitty tcpip's 'Minimum Configuration' path), then re-attach with ifconfig en0 up and verify with netstat -in and lsattr -El en0.
Assuming a mksysb image captures everything, then discovering user data filesystems or non-root VGs are missing on restore.
By default mksysb backs up the rootvg filesystems listed in /etc/filesystems; candidates do not check the exclude list or add non-rootvg data, so a recovery test fails.
What to do instead
Review /etc/exclude.rootvg and the mksysb flags; for data VGs use savevg or a separate backup, and always perform a test restore to a spare LPAR or NIM client before relying on the image.
Diagnosing a hung or failed process by killing it with kill -9 immediately, losing the evidence needed to find the root cause.
Candidates under exam time pressure reach for the strongest signal; kill -9 prevents the process from flushing state and discards core-dump and trace information.
What to do instead
Try kill (SIGTERM) first, inspect with ps -ef, procsig, or procstack while it is alive, and check errpt and the application log; escalate to kill -9 only after capturing diagnostics.
Ignoring errpt and the error log when troubleshooting, and instead reading only application logs.
AIX records hardware and many system errors in the error log with a unique identifier; candidates unfamiliar with errpt -a or errpt -d H miss the actual failing device.
What to do instead
Run errpt -a (or errpt -d H,S) and interpret the error identifier, class, and resource name; use diag -d for hardware errors and clear entries with errclear only after the cause is addressed.