Moving from an on-premises PBX to a hosted phone system removes patching, SIP trunk firewall rules and hardware failures from your list, but a rushed switch loses calls. To migrate PBX to cloud safely, treat it like any server migration: inventory everything, test in parallel, cut over at a planned time and keep a rollback path until numbers have ported. The plan below works whether you are leaving 3CX, FreePBX or plain Asterisk.
This page contains partner links. See our affiliate disclosure.
Short answer: Export the extension list, ring groups, queues, IVR menus, business hours and DIDs from the old PBX; check upload bandwidth, QoS and outbound firewall rules at every site; build the same call flows in the new system and pilot with one team using temporary numbers; then submit number porting, forward calls during the port window, and keep the old PBX and its call records until every number has moved and been tested. (Reviewed November 2026.)
Table of Contents
1. Inventory the old system
List extensions and users, ring groups, queues, IVR menus with their prompts, time conditions, voicemail boxes, call recording rules, every DID and which route it hits, and integrations such as CRM click-to-call. Do not forget analogue lines on ATAs: fax machines, door entry, lift phones and alarm diallers often need special handling or a dedicated line.
# FreePBX / Asterisk: quick exports
# extensions and DIDs: Admin » Bulk Handler » Export in the GUI
# (fwconsole bi / bulkimport only imports CSV files)
asterisk -rx "pjsip show endpoints" > endpoints.txt
mysqldump asteriskcdrdb > cdr-backup.sql
tar czf recordings.tgz /var/spool/asterisk/monitor
On 3CX take a full backup from the admin console and export the user list and call reports before you begin.
Before rebuilding anything, use the call records to find numbers nobody calls, so you do not port or pay for them. FreePBX stores the DID each inbound call arrived on in the CDR table:
# inbound calls per DID over the last 90 days
mysql asteriskcdrdb -e "SELECT did, COUNT(*) AS calls, MAX(calldate) AS last_call
FROM cdr WHERE did != '' AND calldate > NOW() - INTERVAL 90 DAY
GROUP BY did ORDER BY calls;"
A DID that is in the Bulk Handler export but missing from this output had no inbound calls in that period. Check with the business before dropping it: some numbers are kept for printed material, seasonal campaigns or outbound caller ID only.
2. Choose the platform
Compare how each provider handles the features your inventory shows you rely on: call queues, IVR, CRM integration, call recording retention, local numbers in your countries and per-user pricing. Our reviews of JustCall and CloudTalk cover plans, prices and limits, and the JustCall vs CloudTalk comparison puts them side by side.
3. Check the network
Size upload bandwidth for concurrent calls (the VoIP codec bandwidth guide has the figures), enable QoS on the edge router, disable SIP ALG, and allow the provider’s outbound HTTPS and UDP media ranges if your firewall filters outbound traffic. Measure jitter and loss from each site with the checks in the VoIP call quality guide.
4. Rebuild call flows
Recreate ring groups, queues, IVR and business hours in the new system from your inventory. Record fresh prompts rather than reusing old low-bit-rate files, and write down every routing decision so it can be tested line by line.
5. Pilot with one team
Give one team the new apps or desk phones with temporary numbers for a week. Test inbound, outbound, transfers, voicemail to email, recording, and mobile apps on 4G. Fix issues before anyone else moves.
6. Port the numbers
Submit porting requests with the latest bill from the current carrier; names, addresses and account numbers must match exactly. Porting takes days to weeks depending on country and carrier, so plan the cutover date around the confirmed port date. The phone number porting guide lists the documents and common rejections.
7. Cut over
On port day, change the old PBX’s inbound routes to forward to the new temporary numbers as a safety net, then test every ported DID as it goes live. Keep the old system running and registered until the last number is confirmed.
8. Decommission carefully
Export call detail records and recordings you must retain for compliance, cancel old SIP trunks only after porting completes (cancelling first can release the numbers), and remove the SIP firewall rules. Before you wipe the old server’s disks, prove the retained copies can be read: restore cdr-backup.sql into a test database on another machine and compare its row count and date range with the old server, play a sample of recordings from the archive, including the oldest you must keep, and check that whoever handles retention requests can find them. Keep at least two copies, one off-site, for the full retention period. Only then wipe the disks.
Emergency calling checks
A cloud phone system is a new multi-line telephone system (MLTS), so check emergency calling before the first user moves, not after. In the US, Kari’s Law requires an MLTS installed after 16 February 2020 to let users dial 911 directly, with no prefix such as 9, and to notify a central point, such as reception or security, when someone calls 911. RAY BAUM’S Act Section 506 adds a dispatchable location with each 911 call: the validated street address plus floor or room where that is needed to find the caller.
- Register an emergency address with the new provider for every site, with floor or room detail for large buildings.
- Decide how locations for home and roaming users are kept current. Where automatic location is not technically feasible, the FCC rules let providers of non-fixed VoIP use a registered location that the user must be able to update.
- Set the 911 notification target and confirm someone actually sees it.
- Ask the provider how to test emergency calling. Do not place unannounced test calls to the emergency number.
- Outside the US, check the national regulator’s rules for emergency calls and caller location from VoIP.
Cutover test plan
Run the same list during the pilot and again on port day, and tick off each line against your inventory:
| Test | How | Pass when |
|---|---|---|
| Every DID | Call each number from a mobile | It reaches the destination listed in the inventory |
| Out-of-hours routing | Force the time condition to closed, then call | Caller hears the closed message or reaches voicemail |
| IVR menus | Press every option, an invalid key and no key | Each option reaches the right target; invalid input repeats as designed |
| Outbound caller ID | Call a mobile from several extensions | The number shown is the one users should present |
| Transfers and voicemail | Blind and attended transfer to an internal and an external number; leave a voicemail | Transfers complete; the voicemail email arrives with audio |
| Emergency calling | Follow the provider’s test procedure | Correct address on file; the notification reaches the central point |
| Fax, alarm and lift lines | Send a multi-page test fax; ask the alarm company for a test signal | Fax completes; the monitoring centre confirms the signal |
| Mobile app | Inbound and outbound calls on 4G with Wi-Fi off | Two-way audio for several minutes without drops |
Rollback plan
Agree before port day what triggers a rollback and who makes the call, for example the main number failing to ring for longer than an agreed time, or a queue that cannot be answered. Write the trigger and the name down.
What rollback means depends on whether the number has moved. Before the port completes, it means removing the forwarding on the old PBX. After the port, the number belongs to the new provider and porting it back is a new port request with the same paperwork and lead time. The fast fallback is to forward the ported number from the new system to a number that still reaches the old PBX, such as a temporary DID on the old trunk. Point that fallback DID straight at the old queues or ring groups, not at an inbound route that still carries the cutover forwarding, or a call can bounce between the two systems in a loop. That only works while the old trunk is still in service, which is one more reason to cancel it last.
Migrate PBX to cloud at a glance


Official documentation: 3CX documentation, Asterisk documentation, FreePBX documentation, National 911 Program: Kari’s Law and RAY BAUM’S Act, FCC rules for MLTS 911 and dispatchable location (Federal Register).
Related guides: Port a phone number to VoIP · VoIP codec bandwidth · VoIP call quality.
Frequently asked questions
How long does a PBX to cloud migration take?
For a single office with up to about 50 users, plan two to six weeks. Most of that time is number porting; the technical build and pilot usually take a few days.
Can I keep my existing desk phones?
Often yes, if the new provider supports your phone model for provisioning. Otherwise switch users to the desktop and mobile apps, which avoids provisioning work entirely.
What happens to calls during number porting?
The number stays with the old carrier until the port completes, then switches within minutes to hours. Keep the old PBX active and forward to temporary numbers so no calls are lost if the switch is delayed.
Can I roll back after a number has ported?
Not quickly. Porting back is a new port request with the same paperwork and lead time. Until testing is complete, keep the old PBX and a temporary number on its trunk, and forward the ported number to it from the new system if something breaks.
Will fax machines and alarm panels work on a cloud PBX?
Not always. Ask the provider whether it supports T.38, the ITU-T protocol for fax over IP, and test a multi-page fax during the pilot. For alarm panels and lift phones, confirm with the alarm or lift company before moving the line; keeping a dedicated line for them is often simpler.
Maintenance record
This guide changes servers, data or security settings, so we re-check it against current versions on a fixed schedule. Take a backup or snapshot before you start.
- Maintained by
- srvScripts editorial team
- Last full review
- Next review
- Sources
- docs.asterisk.org
