F5 BIG-IP i7800 to BIG-IP r5900 Replacement
F5 LOAD BALANCER REPLACEMENT — FIRST-TIME GUIDE
BIG-IP i7800 → BIG-IP r5900
A click-by-click migration runbook for your first HA-pair hardware swap
The plan in one sentence: your two i7800s are a pair — you'll remove the standby one and plug the first r5900 into the team; prove the new box can carry everything alone for real; then remove the second i7800 and plug in the second r5900. Users never see the change.
Field | Detail |
Document | Click-by-click runbook for a first-timer |
Pair names used in examples | i7800-A (currently ACTIVE) · i7800-B (standby) → r5900-1, r5900-2 |
Method | Rolling member replacement — one box at a time |
Expected outage | None for users (a short controlled failover is part of testing) |
Before you start | Read Sections 1–3 fully; skim the tests in Section 8 |
Version | v1.0 |
1. Read This First — What You Are Actually Doing
Your company has two i7800 load balancers guarding the apps. They work as a team called an HA pair: one is ACTIVE (handles all traffic), the other is STANDBY (a hot copy doing nothing until needed). If the active box dies, the standby takes over in seconds — that takeover is called a failover, and it's the whole trick this migration relies on.
Instead of taking both old boxes out at once (scary, all-or-nothing), you'll swap them one at a time so one working box is always live:
Moment | ACTIVE box | STANDBY box | Who serves users? |
Today | i7800-A | i7800-B | i7800-A |
After Night 1 (Phase 1) | i7800-A | r5900-1 | i7800-A — never changed |
During the proof test | r5900-1 (for a few minutes) | i7800-A | r5900-1 proves itself, then you hand it back |
Final | r5900-1 | r5900-2 | r5900-1 |
THE GOLDEN RULE: Rule of thumb for a first-timer: NEVER remove the ACTIVE box. Every phase only ever touches the STANDBY one. Active boxes only change hands during a deliberate, planned failover you trigger yourself. |
1.1 FAQ a first-timer usually asks
Question | Straight answer |
Will the website go down? | Not if HA is healthy. Traffic moves between members via a heartbeat you test before trusting it. |
What if the new box is wrong/misconfigured? | It stays standby, users unaffected. You fix it with zero urgency. |
What is the worst that can happen? | A failed proof failover — you simply fail traffic back to the old i7800, still racked and staged. That's the whole design. |
How long does it take? | Phase 1: one evening change window. Phase 2: another window, usually days later. |
2. What You Need Before Touching Anything
2.1 On your laptop
A browser (Chrome/Edge) — all F5 management is done over HTTPS web pages.
An SSH client — Windows: use PuTTY (Host Name: the box's mgmt IP, Port 22, Connection type SSH) or PowerShell's built-in ssh. macOS/Linux: ssh admin@<ip>.
A screenshot folder and a plain-text scratch file — every step you'll capture before/after evidence; rollback instructions will paste them back later.
A printed/written rollback trigger list agreed with your lead (examples: sync stuck more than 15 min → stop; test VIP not answering in 2 min → roll back).
2.2 Accounts, addresses and names to pre-write here
Item | Value (fill in BEFORE the night) |
i7800-A mgmt IP / login | ______________________ |
i7800-B mgmt IP / login | ______________________ |
F5OS mgmt IP of r5900-1 (new!) | ______________________ |
Tenant (BIG-IP) mgmt IP on r5900-1 | ______________________ |
3 test applications to watch | ______________________ |
Your core switch / distribution IPs for ping tests | ______________________ |
3. The Jargon You'll Meet — In Plain Words
Term you'll see | What it really is |
UCS file | A full backup archive of the BIG-IP config (kind of like a zip of your whole setup) — the restore file for the new box. |
platform-migrate | A UCS-restore option meaning: 'this came from different hardware' — adapts interface naming etc. |
Device trust | The SSL-certificate handshake that lets two boxes sync safely; you click to accept each other, like friending on a network. |
Device group + config-sync | The set of boxes sharing config automatically. Whenever A changes something, B is copied within seconds. |
Traffic group | The bundle of IPs/services that fails over together ('traffic-group-1' by default). |
Failover (forced) | You deliberately make the active box stand down so the standby takes over — the main tool of this migration. |
Self-IP | An IP the balancer itself owns on a VLAN — this is what neighbor devices talk to. |
Virtual Server (VIP) | The public-facing IP+port combo your apps sit behind. |
Tenant (rSeries only) | The virtual BIG-IP instance running inside the r5900's operating system (F5OS). You build the tenant first, then log into it like a normal BIG-IP. |
4. The One Big Thing to Understand About the r5900
The old i7800 boots straight into BIG-IP — one screen does everything. But rSeries boxes run two layers: F5OS, the appliance's own operating system (where the power, fans, software images and tenants live), and inside it a BIG-IP tenant — which is where your real work happens.
So there are two admin pages (two different IPs on the same box):
Layer | What it looks like | What you do there |
F5OS (outer shell) | Blue-toned platform UI at https://<r5900 mgmt IP> | Install the tenant image, deploy the tenant (click 'Deploy'), manage power/diag |
BIG-IP tenant (inner) | Classic F5 TMUI you already know from the i7800s | Everything else: config, VIPs, HA, sync |
FIRST-LOGIN: All F5OS default-account changes: First login to F5OS forces a password change for the admin account. The tenant's BIG-IP mgmt creds are set when you fill the tenant form. |
5. Phase 0 — Pre-work Weekend (nothing is unplugged)
5.1 Back up the configuration on both i7800s
You make UCS archives (full backups) BEFORE anything changes. Do this on BOTH boxes, one after the other.
Step 1 — Open the BIG-IP web page of i7800-A
In your browser: https://<i7800-A mgmt IP>. Log in with admin.
Step 2 — Find the archive manager
CLICK PATH: System > Archives > UCS List (then click the "Create" button)
Step 3 — Name it clearly, create it
Example file name: pre-migration-i7800A-2026-10-27.ucs — then click OK. BIG-IP builds the archive in about a minute.
Step 4 — Download the archive to your laptop
CLICK PATH: System > Archives > UCS List > click the file name > "Download"
Step 5 — Repeat on i7800-B
Same four steps. You should end with two files on your laptop, dated today. Call them out in your change ticket.
GOOD RESULT: You now have one .ucs file per old box. If everything else goes wrong tonight — these files are the safety net. |
5.2 Stand up the r5900 (in the lab first if you can)
Step 1 — Log into F5OS on r5900-1
Point your browser at https://<r5900-1 F5OS mgmt IP> (you / your DC team set this IP when racking). Log in; first login makes you set a new admin password.
Step 2 — Upload the BIG-IP tenant image
CLICK PATH: Software Management > Tenant Images > Import > choose the BIG-IP ISO matching your i7800's TMOS version (or newer, F5-supported) > Upload
NOTE: Match or exceed the i7800's TMOS version. Example: pair runs 17.1.x → pick 17.1.x or a maintenance release of the same train. Versions live at System > About on the old UI if you need to check. |
Step 3 — Deploy the tenant
CLICK PATH: Tenant Management > Tenant Deployments > "Deploy" button
In the form fill in:
Name: bigip1 (whatever you like — the tenant name will show up on the platform UI)
Type: BIG-IP · Image: pick the ISO you just imported
Management IP / netmask / gateway: the tenant mgmt details you planned in Section 2.2
VLANs to trust + CPU/RAM: leave at the recommended sizing for this platform
Click Deploy, then wait — starting takes a few minutes. Watch the Status column.
GOOD RESULT: Tenant Status flips to Running. Bonus: hit the tenant mgmt IP in a fresh tab — you should see the classic BIG-IP login page (same look as the i7800s!). |
Step 4 — Restore the i7800 config onto the tenant (the clever bit)
SSH to the TENANT (not F5OS this time — the tenant's mgmt IP):
ssh admin@<tenant-mgmt-ip>
tmsh load sys ucs /var/local/ucs/<your-backup>.ucs platform-migrate
NOTE: You'll be asked to copy the UCS file up first: from your laptop, scp the archive to admin@<tenant-mgmt-ip>:/var/local/ucs/ — use your file's real name. |
When it finishes, log back into the tenant's web UI and edit interface references: port numbers/names on the r5900 are different from the i7800 — open:
CLICK PATH: Network > Interfaces > Interface List — check every entry shows a real physical port of the r5900
CLICK PATH: Network > VLANs > Self IPs — make sure each self-IP's VLAN binds to the ports you-cabled-for-it
Do NOT enable config-sync to the old pair yet — that belongs to Night 1.
6. Phase 1 — Night 1: Swap the Standby (i7800-B out, r5900-1 in)
Remember the Golden Rule: tonight you only ever touch the STANDBY box. i7800-A keeps serving every user all night.
6.1 Confirm who's who before unplugging anything
Step 1 — Check status via web
On i7800-A's web UI, top panel shows: BIG-IP - Active. Also:
CLICK PATH: Device Management > Devices — you'll see both boxes listed; the column 'Sync Status' should say In Sync
GOOD RESULT: If everything says Active on A and In Sync — you're good to proceed. |
Step 2 — Same check via SSH (this is the test you'll use all night)
ssh admin@<i7800-A-mgmt-ip>
tmsh show cm failover-status
Look at the Failover traffic-group-1 line — it should say 'active device i7800-A' and ConfigSync 'In Sync'. Text output like this means YES:
Failover: active | ConfigSync: In Sync
6.2 Park i7800-B safely
Step 1 — Take i7800-B out of the data path
In the middle of the UI click: Device Management > Devices > i7800-B > open the device page > click the Force Offline button. Go to its interfaces (Network > Interfaces) and disable any VLAN-facing ports. Leave the box powered on with management live for rollback reasons.
Step 2 — Remove it from the device group
CLICK PATH: Device Management > Device Groups > click your sync group > remove i7800-B from Members list > Save
Step 3 — Break the trust from i7800-A's side
CLICK PATH: Device Management > Device Trust > Peer List > tick i7800-B > remove/delete
GOOD RESULT: Device Management > Devices on i7800-A now shows ONLY i7800-A, In Sync, Active. Users are stale to this — nothing in production changed. |
6.3 Introduce r5900-1 into the team
Step 1 — Cable it per the interface plan
Physical move per Section 5.2's port table. On the r5900 tenant: Network > Interfaces lights up the new links.
Step 2 — Let i7800-A trust r5900-1 (the friend request)
On i7800-A's web UI:
CLICK PATH: Device Management > Device Trust > Peer List > "Add ..." > type r5900-1's tenant mgmt IP and admin creds > Retrieve Device Information > Device Certificate matches? Yes > Add Device
Step 3 — Add r5900-1 to the sync group
CLICK PATH: Device Management > Device Groups > your group > move r5900-1 into the Members list > Save
Step 4 — Fire the first full sync (copies EVERYTHING from A to r5900-1)
CLICK PATH: Device Management > Overview > click r5900-1 > "Sync Device to Group" (or: it may already be buttoned as 'Sync' on the overview row)
Prefer the CLI? Same thing in one line (run this on i7800-A):
tmsh run cm config-sync to-group <your-device-group-name> full-sync
Step 5 — Wait — then check sync on BOTH sides
Give it a minute or two, then on i7800-A and r5900-1 run:
tmsh show cm failover-status # 'ConfigSync: In Sync' is what you want
tmsh show cm device | grep -i sync # quick view both members show Enabled sync
GOOD RESULT: Both screens: 'In Sync'. Devices page shows 2 boxes: A Active, r5900-1 Standby. |
6.4 THE proof: force traffic onto r5900-1 — deliberately
This is the moment of truth: you're going to ask A to stand down for a few minutes, so users actually run on the NEW hardware — and you'll watch them do it safely. Do it inside the low-traffic change window you booked.
Step 1 — On i7800-A, force yourself to standby
ssh admin@<i7800-A-mgmt-ip>
tmsh run sys failover standby
The pair flips: r5900-1 becomes ACTIVE. i7800-A shows Standby. Gratuitous ARPs announce: traffic starts hitting r5900-1 automatically.
Step 2 — Watch users flow — the five-minute checklist
Run these from your SSH to r5900-1 AND from inside your network:
Test | How | Looks good when |
VIP answers | curl -k http://<one-VIP-IP>/ from your laptop | App homepage HTML/headers return |
Pools healthy | tmsh show ltm pool (on r5900-1) | Availability for each pool = green / available |
Live traffic | tmsh show sys connection | head | Real session rows flying past |
Peer pinging | ping -I <r5900-1 self-IP> <core gateway IP> | 0% loss |
GUI view | UI on r5900-1: Local Traffic > Virtual Servers list | All circles/icons green (available) |
Step 3 — Hand traffic back
ssh admin@<i7800-A-mgmt-ip> then: tmsh run sys failover online
Wait 30 seconds, re-run tmsh show cm failover-status on both, confirm A is active again, r5900-1 standby, In Sync. Screenshot both.
GOOD RESULT: Milestone! r5900-1 officially carried 100% of real production and nobody noticed. This proof is what makes Phase 2 a formality, not a gamble. |
7. Phase 2 — Night 2: Remove i7800-A, Add r5900-2
Everything is the same pattern, roles reversed: r5900-1 has proven itself so it's allowed to become the active member, and this time the STANDBY being replaced is i7800-A.
7.1 Hand control to r5900-1 (permanently this time)
ssh admin@<i7800-A-mgmt-ip> then: tmsh run sys failover standby
Step 1 — Confirm r5900-1 active & healthy: same commands as 6.1/6.4
Step 2 — Park i7800-A like you parked B: Force Offline, remove from group, remove from trust
Click paths are identical to 6.2 — just pointing at i7800-A this time.
Step 3 — Leave i7800-A powered-on in the rack for now (don't pull it yet — see Rollback)
7.2 Stand up r5900-2 as the new standby
Clone of Section 5.2 and 6.3 applied to r5900-2:
Rack r5900-2; deploy tenant per your file; platform-migrate the UCS; fix interface mapping; license the tenant.
Cable its ports in the same pattern; confirm link on Network > Interfaces.
From r5900-1: Add peer (Device Trust > Peer List > Add) → move r5900-2 into the group → run full-sync.
Verify: Devices page shows r5900-1 = Active, r5900-2 = Standby, In Sync on both; run the five-minute checklist of Section 6.4 one more time — now aimed at the FINAL pair.
GOOD RESULT: Final state: two r5900s in a healthy pair. Both i7800s retired but still racked, staged for rollback. |
8. Your Test Toolbox (keep this open on night shift)
8.1 SSH commands — copy/paste ready
What you want to know | Command | Good looks like ~ |
Which box is active / is sync healthy? | tmsh show cm failover-status | 'active device' = expected unit · 'ConfigSync: In Sync' |
Detailed per-member status | tmsh show cm device | Each device: Enabled · In Sync |
Are my apps (VIPs) up on this unit? | tmsh show ltm virtual | grep -E "Virtual Server|Availability" | Availability: available for each VIP |
Are pool members alive? | tmsh show ltm pool | grep -E "Ltm::Pool::\|Availability" | each pool/member Available / green |
Is traffic really flowing? | tmsh show sys connection | head | fresh rows (not empty) |
ARP answered for our VIPs? | tmsh show net arp | grep <VIP-IP> | one row with a MAC address per VIP |
Can the box ping upstream? | ping -I <self-IP> <gateway-IP> | 0% packet loss |
The web hit users would do | curl -k http://<VIP-IP>/ (laptop) | your app's HTML/loading redirect comes back |
8.2 GUI equivalents (if you prefer clicking)
Question | Where to click | Green means |
Is this box Active? | top banner of its web UI | Says 'Active' opposite the box name |
Sync state | Device Management > Devices table, Sync Status column | In Sync |
VIP health | Local Traffic > Virtual Servers > Virtual Server List, Availability icon | Green circle = available |
Pool/member health | Local Traffic > Pools > Pool List > select > Members | Green icons on every member |
9. The Panic Button (Rollback)
Both phases share one core recovery move — the magic of keeping the old box racked:
9.1 During/after Phase 1 (mixed pair misbehaving)
Step 1 — Fail traffic back to the old active
ssh admin@<r5900-1-tenant-ip> then: tmsh run sys failover standby
(or on A: tmsh run sys failover online)
Step 2 — Pop r5900-1 out of the group/trust (clicks of 6.2, reversed)
Step 3 — Re-cable i7800-B's interfaces, bring Back Online, re-add to trust + group, full-sync
Exact same buttons you already used in 6.3 — this time adding the OLD box.
9.2 During/after Phase 2
Same move: i7800-A is still racked next door. Bring it Back Online, trust it, add it to the group, sync. Traffic never left its side of the room to begin with.
WHEN TO PULL THE TRIGGER: Decision gate: any test in Section 8 failing for more than 5 minutes inside the window = execute rollback, no debate. Documenting and re-planning later beats hero debugging in front of users. |
10. Wrap-Up, the 'Being a Pro' Checklist
Two weeks later: Schedule the i7800s' quarantine (rack removal, licenses, warranty/RMA letters) only AFTER the final pair has been stable for days.
Update device lists in monitoring/SNMP/syslog, TAC cases, and license-portal notes; both i7800s decommissioned everywhere.
Refresh your backups: Export new UCS archives from both r5900s and file them next to the pre-migration ones.
Send the before/after facts (sized VLAN ports per the plan, TMOS versions, sync status screenshots) to your change ticket as evidence of a clean cutover.
You did a phased hardware migration and no user opened a ticket — that's the goal.
www.thenetworkdna.com · F5 BIG-IP Migration i7800 → r5900
