Worked example: All commands assume lpar01 as the client LPAR with hdisk0 as the in-place migration target and hdisk1 as the unassigned rollback disk. ISO 1 / ISO 2 from ESS are loaded into the HMC virtual media library on hmc01. Sub in your own values throughout.
When to choose this method over nimadm: no NIM master available (and not feasible to build one) · LPAR is on an isolated network with no path to a NIM master · single, one-off migration where NIM setup overhead is not justified. For repeated or fleet migrations, nimadm is strongly preferred — see RB-PATCH-AIX-002.
Console operator required for entire window. The BOS installer needs interactive input at multiple menu screens — do not plan to leave this unattended. The system is offline for 90 — 180 minutes during the install itself.
Out of scope: new installations / "preservation" installs (different BOS menu selections; data-loss implications) · PowerHA / HACMP-clustered nodes (cluster must be quiesced; migrate one node at a time per IBM's PowerHA migration procedure). For 7.1 → 7.3 the procedure is identical — substitute 7.3 media throughout.
alt_disk_copy rollback.Third-party compatibility matters more here than for an SP. Verify against AIX 7.2 TL5:
| Component | What to check |
|---|---|
| TSM / IBM Spectrum Protect client | IBM compatibility matrix for AIX 7.2 TL5 SPx supported client versions. |
| PowerHA / HACMP | Hard version dependency. Do not migrate clustered nodes ad-hoc. |
| Oracle, DB2, SAP | Each has its own AIX 7.2 certification matrix. |
| Kernel extensions | genkex shows the loaded list — anything 3rd-party must be re-validated. |
| Java, OpenSSL, OpenSSH | Ship at newer baselines on 7.2. |
| Monitoring / AV agents | ITM, Nagios, Tanium, etc. |
Three options, in order of practicality. Most modern environments use Option A (HMC virtual optical). Options B and C exist for edge cases.
dd. Fiddly — not all firmware levels handle multi-DVD installs cleanly from USB. Prefer A.Two paths: HMC GUI (slower for large ISOs but accessible) or HMC CLI via viosvrcmd (faster, scriptable).
Via HMC GUI:
Via HMC CLI:
Mid-install swap (when BOS prompts for ISO 2):
Have this command ready to paste before the install starts — at the swap prompt the install is paused and waiting on you.
Shared DVD across LPARs: If the DVD is shared between LPARs via VIOS, ensure it's currently allocated to lpar01 and not held by another partition. A held DVD is the most common reason an Option B install hangs at the BOS device-list screen.
USB stick must be at least 16 GB. AIX 7.2 install ISO is hybrid — bootable as either DVD or USB.
On a Linux workstation:
On AIX 7.x (a system other than the one being migrated):
Triple-check the target device. A wrong of= destroys the disk it points at, no questions asked. lsblk / lsdev -Cc disk | grep usb first, every time.
Caveat: Then physically plug the USB stick into a USB port on the frame allocated to the target LPAR. Not all firmware levels handle multi-DVD installs cleanly from USB — prefer Option A when available.
Wherever the ISO is staged (HMC, DVD burn source, USB write source), verify the SHA256 against the value on the ESS download page.
Why this matters: A corrupted install ISO causes failures partway through the install that look like hardware faults. By the time you see them you've burned an hour of the maintenance window. Catch corruption at the doorstep, not 60% through phase 7.
Capture everything: Wrap the whole pre-flight in script /tmp/preflight_$(date +%Y%m%d).log. You will not have shell access to the running 7.1 system during the install — anything you forget to capture here is gone.
Stop condition: If lppchk -v reports issues, remediate before migration. The migration installer carries broken state forward and amplifies it.
Anything matched is worth confirming with the application team before migration. perfagent.tools in particular has historically caused post-migration noise.
| Filesystem | Min free | Why |
|---|---|---|
| / | 100 MB | Bootloader, ODM |
| /usr | 3 GB | 7.2 fileset footprint is materially larger than 7.1 |
| /var | 1 GB | Migration logs |
| /tmp | 1 GB | installp scratch |
| /opt | 500 MB | Optional product trees |
✓ Verify: hdisk1 shows None in VG column · size ≥ hdisk0 · all LVs in lsvg -l rootvg are open/syncd.
The migration installer preserves /etc but does three-way merges on some files. Capture pre-migration originals to diff against later.
Why these specific files: The 7.2 installer adds new lines (new services, new inittab entries for 7.2 features) and may overwrite local customisations. Diffing pre vs post in step 32 is how you catch what was lost.
Record details you may need to bring up networking from the console if the ODM fails to migrate cleanly.
Print these or store off-LPAR. You will not have access to the running 7.1 system during the install. If the migrated 7.2 comes up with no network, your only path back is the vterm console — and only with these values in hand.
Worst-case use: If both disks become unbootable post-migration, you'll boot the 7.1 install media in maintenance mode and restore this mksysb (RB.3). It's slow (1 — 3 hr) but it works.
Mandatory for boot-media migration. Without an alt_disk_copy clone your only rollback is mksysb restore — which itself requires booting from media in maintenance mode and takes 1 — 3 hours. With alt_disk_copy, rollback is a bootlist change and a reboot.
| Flag | Purpose |
|---|---|
| -B | Do NOT change bootlist — we want to control that explicitly |
| -d | Destination disk |
Confirm bootlist is on hdisk0, NOT hdisk1, before booting install media. The migration installer will offer hdisk0 as the target — if you accidentally pick hdisk1 (the clone), you'll overwrite your rollback. There is no undo.
Use -F not -Fr. We want the LPAR to power off, not auto-reboot off the existing bootlist. -Fr would reboot you straight back into the live 7.1 system before you have a chance to enter SMS.
Wait for the LPAR state in HMC to show Not Activated.
Via HMC GUI:
Via HMC CLI:
You'll see firmware POST output, then the SMS Main Menu:
List all Devices enumerates everything firmware can see. After a short scan you'll get a numbered list including hard disks, tape drives, network adapters, and the virtual optical (or DVD/USB).
What the install media looks like in the device list:
| Entry | Source |
|---|---|
| USB CD-ROM | Virtual optical via VIOS (Option A) |
| SATA CD-ROM | Physical DVD (Option B) |
| USB Mass Storage | USB stick (Option C) |
| IBM 9.5mm DVD-RAM Drive | Frame DVD drive direct |
Select the install media entry, then:
The LPAR will load the BOS installer kernel from the media. Expect 60 — 120 seconds of boot output, then the BOS installer's first prompt:
Press 1 + Enter to claim the vterm as the system console.
Critical menu choice ahead. The default option is "New and Complete Overwrite" which will destroy your rootvg. Do not pick "Start Install Now with Default Settings". The path is: 2 → 1 → 3 → select hdisk0 → 0 → 0.
DO NOT pick 1. The default is "New and Complete Overwrite" which destroys your rootvg. Always pick 2 to get to the menu where you can switch to Migration Install.
Currently shows the default Method as "New and Complete Overwrite". We're going to change that next.
Migration Install upgrades BOS in place and preserves application data and configuration. This is the only correct choice for a 7.1 → 7.2 upgrade preserving the running system.
hdisk1 is your alt_disk_copy rollback. If you select hdisk1 here you will destroy your rollback. The disk to select is the one currently labeled rootvg in the VG Status column — that's hdisk0.
After selection a chevron >>> appears next to hdisk0 indicating it's selected. Press 0 (zero) and Enter to continue.
You'll be returned to the Installation and Settings menu, now showing:
Optional — select 4 (More Options) to review:
| Option | Recommended |
|---|---|
| Trusted Computing Base | No (migration default) |
| Backup Configuration | No |
| Remote Services | No (typically) |
| 64-bit Kernel | Yes (default on modern AIX) |
When satisfied, press 0 (Install with current settings).
Last sanity check: The list of filesystems should match what you saw in df -g at pre-flight. If something looks wrong (a filesystem missing, an extra one listed), abort now (Ctrl-C / power off via HMC) — recovery from a bad start is easier than recovery from a wrong-disk overwrite.
Phase takes 60 — 120 minutes depending on rootvg size and IO speed. The vterm is your only window into progress — keep it open and watched.
Mid-install media swap (around 70 — 80%): If the install spans both DVDs, you'll be prompted:
For HMC virtual optical, switch to your HMC session and:
Then return to the vterm and press Enter. For physical DVD: open the drive, swap discs, close, press Enter.
When the install completes, the LPAR reboots automatically. If the bootlist is set correctly (firmware will set it to the migrated disk), the system comes up on AIX 7.2.
If it boots back into SMS or back off the install media: manually set the bootlist from SMS:
The first boot off the migrated 7.2 rootvg may:
TERM prompt before login — enter vt100 or xterm and continue. Subsequent boots will not prompt.errpt during initial config — review but expect some noise.Don't panic: Any of the above on the first boot is expected. Repeated occurrences on the second boot are not.
In most cases, networking comes up cleanly from the migrated config. If it doesn't (no IP on the expected interface), use the captured config from step 14 to recreate it from the console.
Then troubleshoot why the migration didn't preserve the network config — usually a stale ODM entry referencing a 7.1-only attribute, or an interface name change. Don't leave the LPAR running with hand-rolled config indefinitely.
✓ Acceptance: oslevel at target · lppchk clean · ML found · no inoperative subsystems · bootlist on hdisk0 · hdisk1 still shows altinst_rootvg (rollback intact) · kernel extensions reloaded with no surprising drops.
Expect changes — but review them. The migration installer will have added new lines (new services, new inittab entries for 7.2 features). Check for any local customisations that were lost in the merge and re-apply them. Common loss: custom /etc/security/limits stanzas for application users.
Hand off to application owners for service validation. Treat the change as in-progress until they sign off. Keep the rollback warm — don't run any of the cleanup steps in Phase 9 yet.
Trigger conditions: LPAR fails to boot off the migrated rootvg · lppchk -v reports broken filesets post-migration · new error classes in errpt linked to the migration · application owners report a regression that cannot be remediated in the maintenance window.
Fast path — the reason alt_disk_copy was mandatory in step 16.
System boots from hdisk1 (pre-migration 7.1 clone). Then follow RB.1 verification.
If both disks are unbootable:
This is slow (1 — 3 hours of restore plus reboot). It works, but it's why alt_disk_copy is the rollback you actually want to use. Keep your 7.1 install media accessible until burn-in is complete.
Preserve for IBM support before nuking the failed disk:
/var/adm/ras/devinst.log/var/adm/ras/bos.log/var/adm/ras/bosinst.dataOnly after the migrated system has run cleanly and application owners have signed off.
Mirroring is the high-value cleanup option: mirror the new 7.2 rootvg onto hdisk1 so you have redundancy on both disks again. Skip if your storage layer already provides redundancy.
Update the change ticket with completion timestamp and close. You earned it.
Boot from install media:
Set permanent boot device order:
Boot from a specific hard disk one-time:
| Symptom | Cause / Fix |
|---|---|
| Default option is "New and Complete Overwrite" | Always pick "2. Change/Show Installation Settings", then change Method to Migration. Never "Start Install Now with Default Settings". |
| Migration option not offered | Target disk does not contain a recognised AIX rootvg, OR target version on media is older than installed version. |
| Install hangs at media swap | Virtual optical not actually swapped on VIOS, OR DVD not seated. Check loadopt status. |
| Reboot loops back to install media | Bootlist not set to hard disk. From SMS, set hdisk0 as 1st boot device. |
| First boot drops to console "Set TERM" prompt | Normal on first boot post-migration. Type vt100 or xterm. |
| First boot has no network | ODM did not migrate cleanly. Use captured config from step 14 to manually rebuild. |
lppchk -v reports broken filesets | A migration prerequisite failed silently. Try installp -C then re-run any specific failed filesets from a 7.2 lpp_source. |
Do not follow this blindly — values (level numbers, hostnames, hdisk names, dates, managed-system serial) will change. This is a sequence reference, not a copy-paste recipe.
| Command | Purpose | Notes |
|---|---|---|
| viosvrcmd ... "mkvopt -name ..." | Upload ISO into VIOS media library | Phase 2 step 01 |
| viosvrcmd ... "loadopt -vtd vtopt0 -disk ..." | Load ISO into LPAR's virtual optical | Phase 2 step 02 |
| viosvrcmd ... "unloadopt -vtd vtopt0" | Eject virtual optical | Mid-install media swap |
| dd if=<iso> of=/dev/sdc bs=1M | Write ISO to USB stick | Option C — triple-check target device |
| sha256sum / csum -h SHA256 | Verify ISO integrity before attaching | Phase 2 step 05 |
| alt_disk_copy -d hdisk1 | Mandatory primary rollback clone | Phase 4 step 16 — non-negotiable |
| bootlist -m normal -o | Confirm bootlist points at hdisk0 pre-boot | Critical safety check after alt_disk_copy |
| shutdown -F now | Power off (NOT reboot) the LPAR | Phase 5 step 17 — -F not -Fr |
| chsysstate -m ... -b sms | Activate LPAR in SMS mode via HMC CLI | Phase 5 step 18 |
| SMS · 5 → 1 → 7 | List all devices · select install media | Phase 5 step 20 |
| BOS · 2 → 1 → 3 | Change settings → System Settings → Migration | Phase 6 — not 1 (default overwrite!) |
| BOS · select hdisk0 | Pick the disk currently labelled rootvg | Phase 6 step 26 — never hdisk1 |
| oslevel -s (post-reboot) | Confirm migration succeeded | Expect 7200-05-11-2546 |
| lppchk -v | Verify fileset consistency | MUST be clean pre and post |
| diff /tmp/etc_pre/* /etc/* | Catch lost local customisations | Phase 7 step 34 |
| diff /tmp/genkex_pre.txt <(genkex) | Confirm kernel extensions reloaded | Silent failure mode |
| bootlist -m normal hdisk1 | Roll back to alt_disk_copy clone | RB.1 — fast path |
| SMS · 5 → 1 → 5 → hdisk1 | Boot from clone if migrated disk dead | RB.2 |
| BOS · 3 → 4 (from 7.1 media) | Worst-case mksysb restore | RB.3 — slow but works |
| alt_rootvg_op -X altinst_rootvg | Free hdisk1 after burn-in | Phase 9 — only after sign-off |
alt_disk_copy is non-negotiable. Without it, your only rollback is mksysb restore — which itself requires booting 7.1 media in maintenance mode and takes 1 — 3 hours. With alt_disk_copy, rollback is a bootlist swap and a reboot.hdisk0 (not hdisk1) before you boot the install media. The BOS installer offers hdisk0 as the migration target — accidentally selecting hdisk1 destroys your rollback. There is no undo.shutdown -F (not -Fr) before SMS-mode activation. -Fr would auto-reboot off the existing bootlist before you can enter SMS./etc originals (inittab, inetd.conf, services, rc.tcpip, limits, filesystems) at pre-flight and diff them post-migration. The installer does three-way merges and can drop local customisations silently — limits stanzas for application users are a frequent loss.errpt entries. Expected on the first boot, not on the second.altinst_rootvg on hdisk1 needs to stay warm.