How to Switch Static IP Proxy Providers Safely

Move your outbound traffic with a practical plan for allowlists, application testing, connection draining and rollback.

Changing proxy providers is easier when you separate the migration into decisions you can verify: which workloads need the new route, which destinations must accept it, how you will test it, and when the old route can be retired.

The aim is to keep your application working while you change how it connects. Prepare the new path alongside the old one, prove it with representative operations, and move traffic in controlled stages instead of treating a new proxy URL as a finished migration.

This guide covers switching outbound proxy providers, including a move to QuotaGuard. Inbound reverse-proxy migration requires a separate plan. It does not assume that another provider's IP addresses transfer to your new subscription. Plan for a new egress identity unless the providers explicitly confirm otherwise.

The migration at a glance

Inventory → Prepare → Allowlist → Test → Cut over → Observe → Retire

Keep the old route available for the agreed overlap period. Add the new addresses before moving traffic. Decide what success and rollback look like before changing production.

Choose the right migration path

  • Changing outbound proxy providers: use the sequence below.
  • Moving from shared to dedicated QuotaGuard service: follow the shared-to-dedicated guide. Support provisions a new IP pair and connection URL, and the old access is available for an agreed transition period. Confirm its end date and request any extension before it expires.
  • Moving an existing QuotaGuard subscription from Heroku to direct billing: use the Heroku migration guide. That supported subscription transfer can preserve your existing connection identity. It is not the same operation as replacing one proxy provider with another.
  • Switching between QuotaGuard Static and Shield: the products use different address pools, so your IPs change. Arrange the transition, add both new IPs at every affected destination, and keep the old service available until the new route is verified.

1. Inventory the connections, not just the application

Start with a small migration worksheet. For each workload, record:

  • Runtime and owner: web service, worker, scheduled job, serverless function or administrative tool.
  • Destination hostname, port and protocol.
  • Current proxy integration and where its configuration is stored, including explicit client settings and bypass rules such as NO_PROXY.
  • Who controls the destination's allowlist and how long changes take.
  • Whether the destination permits shared source IPs or requires exclusive or separately registered addresses.
  • Any allowlist capacity, source-IP rate limits or provider-range restrictions that could affect the new route.
  • A representative operation that will demonstrate success.
  • Whether the workload holds connections open or creates them for each operation.
  • The previous configuration and the steps needed to restore it.

A web application and its overnight export may share a repository without sharing proxy settings. A successful browser-facing request does not prove that the export worker has migrated.

Record a baseline before changing anything: normal completion rates, connection errors, latency, queue depth and job duration. Choose measurements that reveal a problem in your application, rather than collecting every available metric.

2. Match the integration to the traffic

A replacement proxy must work with your actual client and protocol. Do not assume that replacing the hostname in an old URL preserves the scheme, port, authentication method or TLS behavior.

Keep the change limited to the traffic that needs the new identity, and preserve the required bypasses for local and private-network destinations.

For HTTP and HTTPS APIs, an HTTP proxy is often the straightforward choice. Database connections, SFTP and other non-HTTP TCP traffic need an appropriate SOCKS or tunnel integration. QuotaGuard's HTTP versus SOCKS guide explains that distinction.

QuotaGuard Static offers HTTP and SOCKS5; QuotaGuard Shield adds TLS to the proxy through HTTPS proxying and SOCKS over TLS. A client that supports HTTPS destinations does not necessarily support an HTTPS proxy, and ordinary SOCKS5 support does not establish SOCKS-over-TLS compatibility. Verify your actual library and version, destination ports, capacity and timeout requirements. Do not infer UDP or QUIC support from documented TCP use cases.

If you use QGTunnel, review the SOCKS quick start against your deployment. Check the startup command, destination mappings and any local configuration file, not only an environment variable. A configuration change is useful only if the process handling the traffic actually loads it.

Also check:

  • Network reachability: the destination must be reachable from the new proxy path. A static public egress address does not itself provide access to a private network or private DNS zone.
  • Application support: confirm that your client uses the selected proxy configuration. A proxy variable is not a universal instruction understood by every library.
  • TLS: retain certificate and hostname verification. An HTTPS destination accessed through an HTTP CONNECT proxy can still have end-to-end application TLS; that is separate from encryption of the client-to-proxy connection and its authentication exchange. See the Static setup guide's HTTPS explanation.
  • Security boundaries: Static HTTP CONNECT does not encrypt the initial proxy authentication exchange or CONNECT target. Shield protects the client-to-proxy hop, but a plaintext destination protocol remains plaintext beyond the proxy unless you separately configure end-to-end encryption. Keep destination authentication and TLS verification intact, and preserve SSH host-key verification for SFTP.
  • Secrets: keep proxy credentials in your approved secret store. Redact full connection URLs from screenshots, tickets and logs.

Use the complete connection details supplied for the new service. Avoid guessing credentials or building a new URL by copying pieces from the old provider. For QuotaGuard, obtain the protocol-specific URLs and assigned IPs from Setup → Outbound. The proxy connection host is not interchangeable with the source IPs a destination sees.

3. Prepare an overlap period and update allowlists

Agree how long both services will remain usable, who pays for the overlap, and the deadline for cancellation or retirement. A new-provider trial is not a promise that the previous provider will keep your old route active. Cross-provider overlap depends on both services and your subscriptions.

Ask destination owners to add the new addresses while retaining the old entries for the migration. Confirm that the rules are active, not merely that someone submitted a change request.

These are the destination's inbound source-IP rules. Separately check that your application's own network permits outbound connections to the new proxy host and port.

Your application reaches the destination allowlist over the old route and the new route during the overlap period. The allowlist keeps the old entries and adds the new addresses.Yourapplicationold routenew routeDestinationallowlistold entriesnew addressesoverlap periodYour application reaches the destination allowlist over the old route and the new route during the overlap period. The allowlist keeps the old entries and adds the new addresses.Your applicationold routenew routeDestination allowlistold entriesnew addressesoverlap period

QuotaGuard provides two assigned static IP addresses for its normal high-availability setup. Allowlist both: traffic can use either address. Seeing one address in a brief check does not establish that all subsequent connections will use that address. See why QuotaGuard provides two IPs.

If a partner cannot permit old and new addresses together, agree a coordinated change window and rollback procedure before proceeding. Do not describe that situation as a routine zero-downtime cutover.

4. Prove the new route before moving production

Test from the runtime that will use the proxy, with the same client, configuration and network restrictions wherever practical. A successful laptop check is useful, but it is not proof that a deployed worker uses the same route.

Use two layers of evidence:

Route evidence. For an HTTP-capable integration, make a proxied request to an IP-check endpoint and compare the result with the assigned pair. For a QuotaGuard check, request https://ip.quotaguard.com through the selected proxy, parse the JSON response’s ip field, and confirm that it matches one of your two dashboard addresses. Where available, confirm the source address in the destination's logs. This checks the path used by that request, not every connection the application may create.

Application evidence. Exercise the destination operation you actually depend on: an authenticated API read, a new database connection and harmless query, or an approved file-transfer test. Start with non-destructive operations. If a write is necessary, use a designated test resource and a cleanup plan.

Include redirects, database-discovered hosts and hostname resolution where your application uses them. For file transfers, verify the resulting file's integrity. Check who resolves destination names in your chosen proxy integration rather than assuming every client resolves them in the same place.

Record what ran, which environment ran it, the expected result and the observed result. A proposed test is not a completed test. An HTTP success is not a database or SFTP validation, and observing one egress address is not a controlled failover test.

Use destination records to confirm both IPs are authorized and, where traffic evidence is available, which paths were exercised. For controlled per-IP testing, use the documented node-specific hostname derived from your subscription’s proxy hostname: PROXY-a.quotaguard.com or PROXY-b.quotaguard.com. These connections bypass load balancing and failover. Keep this override confined to the test, then return to the normal proxy hostname. Do not substitute a bare outbound IP for the proxy hostname. Repeated requests alone do not guarantee both paths were tested.

Include failures in the plan: an invalid credential, a rejected connection or an unavailable destination should be visible to the operator. Do not create those failures in production unless you have explicitly arranged a safe test.

5. Cut over new connections and handle existing ones deliberately

Changing proxy settings does not move an established connection onto a different network path. A connection already using the old proxy may continue there until it closes, or it may be interrupted by a restart, deployment, timeout or service retirement.

An established connection stays on the old proxy until it closes. New connections use the new route. The established connection does not move between routes.Yourapplicationestablished connectionnew connectionsold proxynew routeuntil it closesAn established connection stays on the old proxy until it closes. New connections use the new route. The established connection does not move between routes.Your applicationestablishedconnectionnewconnectionsold proxynew routeuntil it closes

Connection pools can keep reusing old connections after a configuration change. Long-running transfers and persistent sessions need their own handling. Check how your runtime reloads configuration and how your client creates or drains its pool.

The QuotaGuard Static technical FAQ documents a fixed 10-minute HTTPS CONNECT lifetime that is not configurable per account, including when traffic is flowing. It recommends setting a pool's maximum connection lifetime below that limit. Check how your client enforces that setting; it is not a promise that a long in-flight operation will finish before the tunnel closes. For genuinely long-lived connections, evaluate a compatible SOCKS5 or QGTunnel integration, which the FAQ distinguishes from this fixed cap.

Where your deployment supports it, start with a small group of workers or instances. Confirm their new connections and representative operations before expanding the rollout. If you cannot isolate a subset, use an explicitly planned change window instead of claiming a staged rollout you cannot perform.

Before the switch:

  • Stop assigning new work to instances you intend to drain, where supported.
  • Give in-flight operations a defined completion window.
  • Restart or recycle clients and processes in the way your platform requires.
  • Confirm that replacement connections use the new configuration.
  • Stagger client reconnections where possible so the cutover does not create a sudden connection surge.

Do not blindly retry every interrupted operation. A failed response does not always mean a remote write failed. Use the destination's idempotency or reconciliation mechanism when available, especially for payments, record creation and other state-changing requests. QuotaGuard does not retry requests on your behalf; retry handling belongs to your client. Changes to connection reuse can also change counted request usage.

High availability at the proxy layer is not a guarantee that an individual TCP session survives a failure or a provider change. Build the cutover around your application's reconnection behavior.

6. Watch outcomes and keep rollback usable

Compare the migrated workload with its baseline. Watch authentication failures, destination allowlist rejections, timeouts, TLS errors, latency and job completion. Check background work as well as interactive traffic.

Choose an observation period that includes your important workload cycles. A few successful requests during business hours do not cover a nightly synchronization or weekly export. If those cycles extend beyond the overlap deadline, arrange more time or explicitly test the affected job before retiring the old service.

Write down rollback triggers in advance, such as sustained connection failures or a critical job failing its verification. Name the person who can pause or reverse the rollout. Set thresholds from your application's requirements rather than copying arbitrary percentages.

A usable rollback requires the old account, credentials, destination permissions and configuration to remain available. Restore the previous configuration using the same controlled rollout discipline, reconnect affected clients as needed, and verify actual operations again. Rolling back routing does not undo writes already completed at a destination.

7. Retire the old route only after every workload is accounted for

Use your inventory to confirm that no application, worker, scheduled task or administrative process still depends on the previous service. Review available old-provider usage alongside application evidence; neither should be treated as complete proof on its own.

Confirm that old pools and persistent sessions have drained or closed, and that the agreed observation and rollback windows have elapsed. Investigate unexplained traffic on the old route before withdrawing access.

Then coordinate removal of obsolete allowlist entries, revoke unused credentials and cancel the old service under its terms. Update the runbook and ownership records so a future deployment does not restore an obsolete URL.

Keep the new subscription active while your integration depends on its addresses. If you cancel QuotaGuard and later subscribe again, do not expect the same IPs; plan any later retirement with the downstream allowlist owners.

Final cutover checklist

  • Every affected workload and destination has an owner.
  • The new client/protocol combination has been checked.
  • Both new QuotaGuard IPs are allowed at every relevant destination.
  • Representative operations have succeeded from the deployment environment.
  • Existing connections, pools and in-flight work have an explicit plan.
  • Rollback configuration is saved securely and the old route still works.
  • The overlap deadline and observation period are compatible.
  • Background jobs have been verified before cleanup.

The safest migration is not the one with the fewest steps. It is the one where every step answers a specific question before you remove your fallback.

If you are moving to QuotaGuard, contact support@quotaguard.com with your protocols, deployment platform, destination requirements and intended cutover date. Share redacted configuration when needed, not passwords or full credential-bearing URLs.

Reliability Engineered for the Modern Cloud

For over a decade, QuotaGuard has provided reliable, high-performance static IP and proxy solutions for cloud environments like Heroku, Kubernetes, and AWS.

Get the fixed identity and security your application needs today.