Vios server need to change rootvg disk with another storage disk
Please provide steps without alt clone and reboot vios
Vios server need to change rootvg disk with another storage disk
Please provide steps without alt clone and reboot vios
For fun in 2026, let''s quickly compare three LLMs replies to @vaibhav_suravkar question (prompt) above:
Here are the steps to migrate rootvg to a new disk on VIOS without alt_clone and without a reboot, using alt_disk_copy or the manual mirrorvg method:
hdisk0) and new disk (e.g. hdisk1)lspv
lspv | grep rootvg
lspv hdisk1
Should show None for VG. If not, clear it:
chpv -C hdisk1
extendvg rootvg hdisk1
mirrorvg -S rootvg hdisk1
-S runs synchronization in background. Monitor sync progress:
lsvg rootvg | grep -i stale
lsvg -l rootvg
Wait until 0 stale PPs before proceeding.
lsvg rootvg
Confirm STALE PPs: 0
bootlist -m normal hdisk1
bootlist -m normal -o
unmirrorvg rootvg hdisk0
reducevg rootvg hdisk0
chpv -C hdisk0
lspv
lsvg rootvg
lsvg -l rootvg
bootlist -m normal -o
mirrorvg -S flag is critical -- without -S it blocks until sync completesoem_setup_env if you need root-level LVM commands:oem_setup_env
Observe:
You want to migrate the active rootvg on IBM PowerVM VIOS from one SAN/storage-backed disk to another disk, without using alt_disk_copy / alt clone and without rebooting the VIOS.
This is essentially an online rootvg mirror migration using LVM mirroring.
Orient:
On VIOS/AIX, rootvg can be mirrored live across disks. After synchronization, the old disk can be removed from rootvg and boot records rewritten to the new disk.
Typical flow:
rootvgNo reboot required if done correctly.
Assumptions:
hdisk0hdisk1Decide:
Safest non-reboot method is standard live LVM mirroring using:
extendvgmirrorvgsyncvgbosbootbootlistunmirrorvgreducevgAct:
ioscli lspv
Example:
hdisk0 00f6c1... rootvg active
hdisk1 none None
chdev -l hdisk1 -a pv=yes
If previously used:
chpv -C hdisk1
extendvg rootvg hdisk1
mirrorvg -S rootvg hdisk1
Flags:
-S avoids syncing stale partitions immediately during command executionThis may take time.
syncvg -v rootvg
Check status:
lsvg -l rootvg
lsvg -p rootvg
You want all LPs == PPs and no stale partitions.
On VIOS:
bosboot -ad /dev/hdisk1
Then update bootlist:
bootlist -m normal hdisk1
Verify:
bootlist -m normal -o
unmirrorvg rootvg hdisk0
Wait for completion.
reducevg rootvg hdisk0
lsvg -p rootvg
Expected:
rootvg:
PV_NAME PV STATE TOTAL PPs FREE PPs FREE DISTRIBUTION
hdisk1 active ...
After verification:
rmdev -dl hdisk0
Notes:
migratepv for full rootvg migration on VIOS boot disks unless you specifically understand boot LV implications.mirrorvg handles boot logical volumes correctly.savebase
before and after migration.
Use an online rootvg mirror migration (no alt_disk_copy, no immediate reboot).
oem_setup_env
lsvg -p rootvg
lspv
extendvg rootvg hdiskNEW
mirrorvg rootvg hdiskNEW
syncvg -v rootvg
lsvg -l rootvg
lsvg -p rootvg
bosboot -ad /dev/hdiskNEW
bootlist -m normal hdiskNEW
bootlist -m normal -o
unmirrorvg rootvg hdiskOLD
reducevg rootvg hdiskOLD
chpv -C hdiskOLD
lsvg -p rootvg
lspv
bootlist -m normal -o
Notes:
The task is non-trivial: a live, no-reboot rootvg migration on IBM PowerVM VIOS without alt_clone. This is a domain where accuracy, ordering, and omissions carry real operational risk.
Strengths: Clean structure, correct core sequence, includes the oem_setup_env reminder (critical on VIOS where padmin is the default shell), and correctly flags the -S behavior note.
Weaknesses:
bosboot -ad /dev/hdisk1 before bootlist. This is a significant omission -- bootlist alone does not write the boot record to the new disk. On AIX/VIOS, bosboot is required to make a disk actually bootable, not just listed.-S is backwards: -S actually means "synchronize" the stale partitions -- it does NOT skip sync. The note saying "without -S it blocks" reflects confusion about the flag semantics.syncvg as an explicit verification step.savebase mention.Grade: B- -- Good scaffold, but the missing bosboot and the -S mischaracterization are operationally significant errors.
Strengths: Most complete response. Includes bosboot, syncvg, savebase, rmdev, and the dual-VIOS caution. The OODA-style framing (Observe/Orient/Decide/Act) is unusual but actually works well for a structured runbook. The migratepv warning is a legitimate and useful callout. Flags LP == PP verification explicitly.
Weaknesses:
chdev -l hdisk1 -a pv=yes for disk verification is non-standard and potentially wrong for this context -- lspv hdisk1 is the correct check.ioscli lspv prefix is technically correct on VIOS but inconsistent with the rest of the commands shown.Grade: A- -- Most operationally complete. The bosboot inclusion and savebase tip push it above the others.
Strengths: Most concise. Correct sequence. Includes bosboot, syncvg, and the most honest and technically accurate closing note: "The change is only fully proven at next reboot, so schedule a maintenance reboot later." This is the most accurate characterization of the actual risk posture -- the migration is live-complete but boot validation is deferred.
Weaknesses:
savebase mention.oem_setup_env detail beyond the initial invocation.Grade: B+ -- Correct, concise, and honest about residual risk. Best closing caveat of the three.
| Criterion | Claude 4.6 | ChatGPT 5.5 | Codex 5.3 |
|---|---|---|---|
| bosboot included | No | Yes | Yes |
| syncvg explicit | No | Yes | Yes |
| oem_setup_env | Yes | No | Yes |
| savebase | No | Yes | No |
| -S flag accuracy | Wrong | Correct | N/A |
| Boot risk caveat | Weak | Moderate | Best |
| Overall grade | B- | A- | B+ |
ChatGPT 5.5 produces the most complete runbook. Codex is the most honest about residual risk. Claude's omission of bosboot and the -S mischaracterization are the most consequential errors in a production context. For an actual VIOS migration, you would composite the ChatGPT completeness with the Codex risk caveat -- and verify bosboot semantics in the IBM docs before executing.
All three pick the correct general method: online rootvg mirror migration. IBM’s VIOS-specific rootvg mirroring doc points to mirrorios for VIOS 3.1/4.1, not raw AIX-only habits. (IBM) The pasted comparison contains Claude, ChatGPT Instant, and Codex answers.
Best answer: ChatGPT 5.5 Instant.
It includes the critical missing step: bosboot -ad /dev/hdisk1, plus bootlist, syncvg, savebase, and warns against migratepv.
Second: Codex.
Clean, short, mostly correct. Also includes bosboot. Weakness: less VIOS-specific and uses blocking mirrorvg unless sync time is acceptable.
Worst: Claude.
It omits bosboot, which is a serious flaw for rootvg boot migration. Setting bootlist alone does not necessarily make the new disk bootable.
Use the ChatGPT Instant procedure as the base, but improve it with VIOS-native preference:
For VIOS, prefer checking whether mirrorios is appropriate first. If doing raw AIX LVM from oem_setup_env, the essential sequence is:
extendvg rootvg hdiskNEW
mirrorvg -S rootvg hdiskNEW
syncvg -v rootvg
bosboot -ad /dev/hdiskNEW
bootlist -m normal hdiskNEW
savebase
unmirrorvg rootvg hdiskOLD
reducevg rootvg hdiskOLD
Corrected ranking:
bosboot.Main defect across all three: none clearly says, “This is not truly validated until the next reboot.” Codex at least mentions that.
Quick assessment: all three are broadly pointing to the right pattern (online rootvg mirror migration), but
they are not equal in operational safety.
Ranking
ChatGPT 5.5 Instant
Codex gpt-5.3-codex medium
Claude Sonnet 4.6
Why
specific caution notes (savebase, dual-VIOS sequencing).
VIOS nuance and fewer guardrails.
around mirrorvg -S.
Key technical gaps across responses
bosboot on the new disk is a critical step for boot reliability and should not be omitted.
Better to explicitly verify no stale partitions and confirm rootvg layout before unmirroring.
All should clearly separate what runs in padmin/ioscli vs oem_setup_env.
Best practical answer
Use the ChatGPT flow, but keep Codex’s explicit “validate at next scheduled reboot” note. That combination is the safest operationally.
This example clearly shows a benefit of using multiple LLMs, suggesting and evaluating various solutions, to increase user confidence in LLM assigned tasks.
This also mirrors my work on complex LLM tasks, many documented here: