CertSafari
    Snowflake SnowPro Advanced: Security Engineer (SEA-C01)· Lessons

    Domain 2 · Lesson 11/21

    Failover execution: readiness audit, promotion and validation

    Manage secure replication and failover operations.

    14 min read
    4.29% of exam
    5 sources
    Published 20 Oct 2026
    Docs as of 18 Oct 2026

    What you will be able to do

    • Audit a failover group's edition, object set, privileges and refresh currency before a failover
    • Rehearse promotion safely with a dedicated non-production failover group
    • Promote secondary failover groups, handling refreshes that are still in progress
    • Validate the promoted account: resumed refreshes, re-trusted storage integrations and working SSO, OAuth and SCIM

    1.Auditing readiness before a failover

    A failover only recovers what the failover group replicates, into accounts that can actually receive it. The audit therefore starts with edition. Database and share replication work on every edition. Replication of every other object (users, roles, network policies, security and other integrations, account parameters) is Business Critical Edition or higher, and so are failover groups themselves. Plan every account in a failover group at Business Critical or above, the source and every target. Enabling ENABLE_ACCOUNT_DATABASE_REPLICATION for the organization is a prerequisite, and it does not lift the edition requirement.

    Next, prove what is in the group. Run SHOW FAILOVER GROUPS on the primary account and read the definition itself. Login history shows who signed in, and query history shows statements someone ran. Neither one tells you what the group replicates. A REPLICATE grant only shows who may refresh the group.

    Group definition columns an auditor reads, and what each one proves
    ColumnWhat it showsReadiness question it answers
    object_typesObject types enabled for replication and failoverAre users, roles, network policies and integrations in the set, or only databases?
    allowed_integration_typesIntegration types enabled for replication (always present, even when none were specified)Are security integrations actually replicated?
    allowed_accountsAccounts enabled for replication and failoverIs the intended DR account a target?
    replication_scheduleRefresh interval, NULL if none is setDoes the refresh cadence meet the recovery point objective?
    secondary_statestarted or suspendedAre scheduled refreshes actually running?

    Then check who can pull the trigger. Replication and failover group privileges and ownership covers the grants. For readiness, confirm that the on-call role holds FAILOVER on each secondary failover group in the target account, and FAILOVER on the connection too if clients are redirected. Finally, check currency in the target account. The query below should return no rows. A row means a refresh is still running, and that blocks promotion. Also confirm that private endpoints and Tri-Secret Secure exist in the target, because neither is enabled there by default.

    Target account: confirm no refresh is in progress for the groupsql
    SELECT phase_name, start_time, job_uuid FROM TABLE(INFORMATION_SCHEMA.REPLICATION_GROUP_REFRESH_HISTORY('myfg')) WHERE phase_name <> 'COMPLETED' and phase_name <> 'CANCELED';

    Checkpoint 1 of 9· Fill the gap

    Before failover, you run this query in the target account. When no refresh is blocking promotion, it should return no rows. Which phase name completes the filter?

    SELECT phase_name, start_time, job_uuid FROM TABLE(INFORMATION_SCHEMA.REPLICATION_GROUP_REFRESH_HISTORY('myfg')) WHERE phase_name <> ' ? ' and phase_name <> 'CANCELED';

    Checkpoint 2 of 9· Exam question

    A security team audits its failover group each quarter. They must prove that network policies, security integrations, users, and roles are all in the replicated object set, not only databases. Which audit step answers this directly?

    Checkpoint 3 of 9· Exam question

    A company wants to add network policies, users, roles, and security integrations to an existing failover group. What account requirement applies?

    Sources123

    2.Rehearsing failover without touching production

    A refreshed secondary is not a test of failover. Its objects are read-only until the group is promoted, so inspecting them says nothing about how users, roles and network policies behave once the account becomes writable and starts serving logins. Promoting the production secondary as a test has the opposite problem. The old primary's objects go read-only, and scheduled refreshes on every secondary are suspended, so the test is itself a production failover.

    Promotion is per group, and replicated objects can be split across several failover groups. That allows a safe rehearsal. Create a dedicated failover group with the same security object types between two non-production accounts. Refresh it, promote the secondary, and run functional checks against the promoted account. For the security objects, follow the pattern used when Replicating security integrations (SAML2, OIDC, OAuth, SCIM): identify source, target and connection URL, prepare each side, then test failover and failback. The pass criteria are working SAML SSO through the new primary connection, OIDC SSO for custom providers once the target's redirect URI is registered, and SCIM provisioning landing on the promoted account.

    Checkpoint 4 of 9· Check yourself

    Your security team wants proof that replicated users, roles, network policies and SSO integrations actually work after promotion. Production must not be disrupted. Which approach gives that proof?

    Checkpoint 5 of 9· Exam question

    Before an annual DR exercise, an engineer wants to confirm that replicated users, roles, and network policies are fully functional after promotion, without touching the production failover group. What is the most reliable controlled test?

    Sources34

    3.Promoting the secondary failover group

    Failover is always run from the target account. In Snowsight, choose Admin » Accounts » Replication » Initiate failover, select the failover groups, then select the connections. Promoting both together, called a bulk failover, is the most consistent option. Groups and connections are separate objects. Promoting a group makes its objects writable but redirects no clients, and promoting a connection moves clients without making any data writable. In SQL, a role with FAILOVER runs ALTER FAILOVER GROUP myfg PRIMARY for each group, followed by ALTER CONNECTION … PRIMARY.

    This happens because the replication service can stay up during a partial regional outage and keep refreshing. In SQL there are two ways to clear the block. SUSPEND stops future refreshes but makes you wait for the one in flight. SUSPEND IMMEDIATE also cancels a scheduled refresh that is running. Cancelling has a cost. A refresh stopped in SECONDARY_DOWNLOADING_METADATA or SECONDARY_DOWNLOADING_DATA can leave the target inconsistent. Once a refresh enters those phases it runs to completion even without the source, so letting it finish is the safer choice. Watch the phase before you decide.

    Check the current phase of an in-progress refreshsql
    SELECT phase_name, start_time, end_time
      FROM TABLE(
        INFORMATION_SCHEMA.REPLICATION_GROUP_REFRESH_PROGRESS('myfg')
      );

    Snowsight handles the same conflict differently. By default it waits for running refreshes, polling with your selected warehouse. Under advanced options, Exit with current progress promotes only the groups that are not refreshing and leaves the rest for later. Cancel refreshes and force failover promotes everything, but you may have to clean up inconsistencies from the interrupted refreshes afterwards.

    Checkpoint 6 of 9· Check yourself

    An on-call operator signs in to the Region B account and runs ALTER FAILOVER GROUP myfg PRIMARY. The command succeeds, but BI tools using the connection URL still fail against Region A. What went wrong?

    Checkpoint 7 of 9· Check yourself

    In a partial regional outage, REPLICATION_GROUP_REFRESH_PROGRESS shows that the scheduled refresh of myfg is in the SECONDARY_DOWNLOADING_DATA phase. You want the promoted account to be consistent. What should you do?

    Sources3

    4.Validating the promoted account

    Start by checking that the failover completed. Every intended group should now be primary in the promoted account, and SHOW CONNECTIONS should report is_primary as true there. Groups that were skipped can be promoted in another bulk failover or one at a time. Next, restore replication. Failover suspends scheduled refreshes on all secondary failover groups, and nothing resumes them automatically. Run the statement below in every target account that holds a secondary, then confirm that secondary_state reads started.

    Run in each target account holding a secondary groupsql
    ALTER FAILOVER GROUP myfg RESUME;

    Then re-establish trust with external resources. A replicated storage integration has its own cloud IAM entity, separate from the primary's. Until you grant that entity access in your cloud provider, stages in the promoted account cannot read or write storage. The grant is needed only once per target account, so do it before a real outage, not during one. The integration itself is configured as covered under Storage integrations and allowed storage locations.

    Identity checks come next, one per integration type:

    • SAML SSO should work through the new primary connection.
    • Custom-provider OIDC integrations are not Client Redirect-aware, so verify OIDC SSO again in the promoted account.
    • SCIM provisioning pointed at the connection URL follows the promotion without any IdP change, and existing SCIM tokens and PATs keep working.

    Finish with the object-level audit from Validating replicated users, roles and grants, and refresh status, run against the now-writable account. SHOW USERS, SHOW INTEGRATIONS and DESCRIBE INTEGRATION should match the source. Then resume or replay ingest and transformation pipelines to bring the data current.

    Checkpoint 8 of 9· Fill the gap

    A failover has finished. The old primary is now a secondary, and its secondary_state reads suspended. Which keyword, run in each target account that holds a secondary group, starts automatic refreshes again?

    ALTER FAILOVER GROUP myfg  ? ;

    Checkpoint 9 of 9· Match them up

    Your post-failover audit covers three replicated identity integrations. Match each one to what you should expect or verify on the promoted account.

    Tap a term, then the definition that fits it.

    Sources3254

    Practise it for real

    Rehearse a full failover of a dedicated non-production failover group named myfg and its connection, then validate the promoted account.

    1. 1.In the target account, run SHOW FAILOVER GROUPS and read object_types, allowed_integration_types and secondary_state for myfg.

      Why: It confirms that the group carries the security object types you mean to test, and that refreshes are running.

      You should see: object_types lists the account objects you configured, and secondary_state reads started.

    2. 2.Run the REPLICATION_GROUP_REFRESH_HISTORY query that filters out COMPLETED and CANCELED phases.

      Why: Promotion fails while a refresh is in progress.

      You should see: No rows. If a row appears, check REPLICATION_GROUP_REFRESH_PROGRESS and wait out any SECONDARY_DOWNLOADING phase.

    3. 3.Run ALTER FAILOVER GROUP myfg PRIMARY, then ALTER CONNECTION myconnection PRIMARY, with a role that holds FAILOVER on both.

      Why: Data becomes writable and clients are redirected as two separate promotions.

      You should see: Both statements succeed, and SHOW CONNECTIONS shows is_primary true in this account.

    4. 4.Sign in through the connection URL with SSO, and run SHOW USERS and SHOW INTEGRATIONS.

      Why: It proves identity and replicated objects work on the promoted account.

      You should see: SSO succeeds, and the user and integration counts match the source.

    5. 5.In each account that now holds a secondary myfg, run ALTER FAILOVER GROUP myfg RESUME.

      Why: Failover suspended scheduled refreshes.

      You should see: SHOW FAILOVER GROUPS reports secondary_state as started.

    Stuck? Get a nudge

    If the PRIMARY statement errors with 'is being refreshed', don't reach for SUSPEND IMMEDIATE first. Check the refresh phase.

    Sources

    (5)

    Every claim above is drawn from one of these pages, quoted as it was written on the date shown.

    1. 1.
      “Database replication and share replication are available on all editions.”
      ↩︎ Auditing readiness before a failover
      “Replication of all other objects is only available for Business Critical Edition (or higher).”
      ↩︎ Exam trap 1
    2. 2.
      “List of specified object types enabled for replication (and failover in the case of a FAILOVER group).”
      ↩︎ Auditing readiness before a failover
      “Snowflake always includes this column in the output even if integrations were not specified”
      ↩︎ Auditing readiness before a failover
      “Current state of scheduled refresh. Valid values are started or suspended.”
      ↩︎ Validating the promoted account
    3. 3.
      “The example in this section must be executed by a role with the FAILOVER privilege.”
      ↩︎ Auditing readiness before a failover
      “Those objects become read-only on the account that formerly was the primary and is now a secondary account.”
      ↩︎ Rehearsing failover without touching production
      “You can optionally divide the replicated objects across multiple failover groups”
      ↩︎ Rehearsing failover without touching production
      “the objects specified in those failover groups become writable on the newly promoted primary account”
      ↩︎ Promoting the secondary failover group
      “During a partial outage in your source region, the replication service might continue to be available”
      ↩︎ Promoting the secondary failover group
      “canceling a refresh operation in the SECONDARY_DOWNLOADING_METADATA or SECONDARY_DOWNLOADING_DATA phase might result in an inconsistent state on the target account.”
      ↩︎ Promoting the secondary failover group
      “Snowflake uses your currently selected warehouse to poll the status of the ongoing refreshes.”
      ↩︎ Promoting the secondary failover group
      “you might need to clean up any inconsistencies on the secondary system from the interrupted refreshes”
      ↩︎ Promoting the secondary failover group
      “must be executed in each target account with a secondary failover group to resume automatic refreshes”
      ↩︎ Validating the promoted account
      “Or you can fail over the remaining failover groups one at a time”
      ↩︎ Validating the promoted account
      “you can resume or replay your ingest and transformation pipelines to update the replicas to the current state”
      ↩︎ Validating the promoted account
      “For the most consistent and reliable failover experience, select all the applicable failover groups and connections and promote them all at the same time.”
      ↩︎ Exam trap 2
      “On failover, scheduled refreshes on all secondary failover groups are suspended.”
      ↩︎ Exam trap 3
      “Snowflake prevents failover if a refresh operation is in progress.”
      ↩︎ Prediction
      “Select any connections to promote. After the failover, those connections connect to the account that you’re promoting to be the new primary account.”
      ↩︎ Checkpoint
      “Allowing the phase to complete before you fail over ensures replicas are in a consistent state.”
      ↩︎ Checkpoint
    4. 4.
      “The general approach to test replication and failover/failback with each network policy and security integration is as follows:”
      ↩︎ Rehearsing failover without touching production
      “After failover, SAML SSO works on the new primary connection.”
      ↩︎ Validating the promoted account
      “If you completed the previous step and use a custom provider, verify OIDC SSO again in the target account after failover.”
      ↩︎ Validating the promoted account
      “promoting the connection to the new primary account reroutes provisioning requests without requiring any reconfiguration of the identity provider”
      ↩︎ Validating the promoted account
      “continue to work on the new primary account after failover”
      ↩︎ Validating the promoted account
      “OIDC integrations using OIDC_PROVIDER='CUSTOM' are not Client Redirect-aware.”
      ↩︎ Checkpoint
    5. 5.
      “The replicated integration has its own identity and access management (IAM) entity”
      ↩︎ Validating the promoted account
      “you must update your cloud provider permissions to grant the replicated integration access to your cloud storage”
      ↩︎ Validating the promoted account
      “You only need to configure this trust relationship on target accounts one time.”
      ↩︎ Validating the promoted account

    Ready to test yourself?

    Practise the 15 questions on this subdomain.

    Spotted a mistake, or was something unclear? Tell us.