Skip to content
KiweelyKiweelyKiweely

Disaster Recovery and Failover for a Company-Wide Workday Connector

Bridge, coming next, does not keep a copy of Workday to fail over to. Kiweely Cloud uses stateless workers. A dedicated static IP has to stay allowlisted in Workday. Self-hosted failover belongs to the customer’s own platform. No recovery-time objective is on the product pages.

  • By Kiweely
  • 4 min read
On this page
Disaster Recovery and Failover for a Company-Wide Workday Connector

The failure you can plan from the published pages is a connector that stores nothing, in front of Workday, which stores everything. There is no second HR system inside the connector to promote when a region blips.

Kiweely Bridge is coming next. The security and hosting copy is the whole of the recovery story Kiweely has published. It does not include a recovery-time objective, a recovery-point objective, or a status-page promise. Do not invent those numbers in a design review.

What there is to recover

Data Where it lives if Bridge is down What Bridge holds
DataWorkers, pay, elections, business processes Where it lives if Bridge is downWorkday, the system of record What Bridge holdsNothing on disk. Responses pass through memory and are not written to a database or to logs
DataThe employee’s Workday grant Where it lives if Bridge is downSealed with AES-256-GCM inside the client session, decrypted only in memory during a call What Bridge holdsNo table of employee credentials. An idle worker holds zero of them
DataAdmin policy Where it lives if Bridge is downTenant URL, enabled tool rules, and payload-free call metadata, as the trust table states What Bridge holdsNot the tool arguments or the answer
DataThe chat Where it lives if Bridge is downThe assistant the employee was using What Bridge holdsNot a Bridge transcript

A Bridge outage does not roll Workday back, and a Workday outage is not repaired by restarting Bridge. You fail the assistant back to the Workday UI, which the employee can already open. That is an unsatisfying sentence and it is the accurate one.

Three hosting modes, three different failovers

The hosting section on the Bridge pages:

Kiweely Cloud. Stateless multi-region workers on a dedicated tenant route. No infrastructure for your IT team to run, and no stored payloads to replicate. What is not published: which regions, how traffic moves, or how long a regional loss lasts. Call it multi-region and stateless. Do not call it an active-active design with a measured RTO.

Dedicated single-tenant. An isolated worker pool and a dedicated static outbound IP, so you can enter that address in Workday’s Maintain IP Ranges and enforce it in Manage Authentication Policies. Workday then rejects token use from anywhere else. Failover that issues a new outbound address is incomplete until that Workday allowlist matches the new address. The page does not say the IP is anycast or that a second address is pre-registered. Treat an IP change as a Workday change window.

Self-hosted container. The signed, stateless container runs in your AWS, Google Cloud, Azure, or Kubernetes cluster. Workday traffic can stay in your perimeter. Restart, capacity, and zone failure are your orchestrator’s job. Kiweely’s published claim is the same admin setup workflow, not a clustered control plane you do not operate. This mode is an enterprise quote, as is the dedicated IP.

What to rehearse

  1. Assistant vendor down, Bridge up, Workday up. Employees use Workday. Bridge does not embed Claude, ChatGPT, or Gemini; those products have their own availability, which these pages do not SLA.
  2. Bridge workers down, Workday up. Same fallback. In-flight reads disappear with the connection, because they were not stored. An in-flight write that has not been allowed on the browser card expires with the short-lived action hash. It should not land in Workday. Confirm that on a non-production tenant when the product exists. It does not exist yet.
  3. Workday down. Nothing in the connector answers from cache of HR rows, because that cache is not part of the design. Stale answers would be worse than an error.
  4. Dedicated IP rotated. Update Maintain IP Ranges before you expect calls to succeed.
  5. Sandbox and production are different tenants. A disaster test in sandbox does not move production data, and a sandbox copy of production is still real worker data. Do not “fail over” a demo onto production to see if it works.

Admin policy is the piece you would need to re-enter if the configuration store were lost. The pages say Bridge stores tenant URL, enabled tool rules, and payload-free metadata. They do not publish a backup schedule for that configuration. Ask for it when you register interest. Until then, keep your own record of which tools and fields you intended to enable.

Frequently asked questions

Can we point Bridge at a Workday disaster-recovery tenant?

Only if that environment is a real Workday tenant with its own URL, its own API client, and workers who can sign in there. Bridge’s plan includes a company’s non-production tenants. It does not document Workday’s disaster-recovery tenant product. Follow Workday’s tenant model, and do not assume the production client works on another host.

Does zero storage make recovery faster?

It removes HR data from the connector’s recovery. It does not make Workday or the assistant vendor faster. Kiweely Bridge is coming next, and it does not publish a recovery-time number.

Your Workday work, done by AI.

Kiweely Vine is the AI agent that builds, deploys and tests Extend apps as real people, makes changes in any Workday task screen, answers from your real tenant and builds reports.

Get the next one in your inbox.

One email a month for Workday teams: what’s changing in Workday and what to retest, one tip tried in a real tenant, and what’s new at Kiweely.

We’ll email you once to confirm. Unsubscribe in one click from any issue.