# Embedded verifier and post-migration data verification cost > Source-anchored research for Embedded verifier and post-migration data verification cost. Run status: COMPLETED. Parent: [Migration and Cutover Cost (Live Migration, mongosync, Dual-Running, Egress)](https://llms-explorer.com/tree/migration-and-cutover-cost-live-migration-mongosync-dual-running-egress/) · 5 facets · 18 facts · page: https://llms-explorer.com/tree/embedded-verifier-and-post-migration-data-verification-cost/ ## Definitions - Embedded verification state is not resumable: restarting verification repeats its work and can leave it behind migration. The original design deliberately chose in-memory state over persistence. (medium confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/mongosync/current/reference/verification/embedded.md) ## How it works - The standalone MongoDB Labs verifier warns that change-stream lag can extend cutover or exceed the oplog window and force verification to restart. Growing recheck generations can also lengthen cutover even when the change stream keeps up. (low confidence) (verified as of 2026-10-27) — [source](https://github.com/mongodb-labs/migration-verifier) - MongoDB recommends verifying each sync before moving application traffic; both its documentation and migration-verifier repository prefer the embedded verifier when its requirements are met. (medium confidence) — [source](https://www.mongodb.com/docs/mongosync/current/reference/verification.md) ## Problems, failure modes and limitations - In its October 13, 2026 migration account, RightTouch reports TTL-related verification failures and an autoscaling rolling restart that interrupted verification during rehearsal. It used document-count checks and fixed a larger destination tier for production; these are case-specific operational costs, not universal failure rates. (low confidence) (verified as of 2026-10-13) — [source](https://tech.righttouch.co.jp/entry/2026/10/13/123032) - Matching document counts cannot establish correct propagation of updates. MongoDB limits count-based verification to insert-only workloads, and RightTouch explicitly accepted this coverage gap when substituting count checks for embedded verification. (medium confidence) — [source](https://www.mongodb.com/docs/mongosync/current/reference/verification.md) - MongoDB documents that post-sync dbHash comparison requires COMMITTED state, blocks writes while hashing, and cannot verify sharded clusters. It therefore carries an availability constraint that a count comparison does not address. (low confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/mongosync/current/reference/verification.md) - SERVER-134745 records a sharded MongoDB 9.0.2 database-drop/change-stream reproduction that was fatal to the embedded verifier in CI. This is historical bug evidence, not evidence that supported production migrations generally fail. (low confidence) — [source](https://jira.mongodb.org/browse/SERVER-134745) ## Comparisons and alternatives - The documented verifier memory requirement is about 10 GB plus 500 MB per million documents. At 100 million documents this implies about 60 GB on the mongosync host, an arithmetic sizing estimate rather than a dollar-cost benchmark. (medium confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/mongosync/current/reference/verification/embedded.md) ## Facts and statements - Mongosync 1.9.0 introduced the embedded verifier on November 13, 2024, enabled by default for replica-set migrations; live upgrades from earlier releases disable embedded verification. (medium confidence) — [source](https://www.mongodb.com/docs/mongosync/current/release-notes/1.9.md) - The implementation lead describes a full scan and change-stream hashing on each cluster, with per-document IDs and hashes retained in memory. The fast final checksum comparison therefore does not eliminate the preceding scan and memory work. (low confidence) — [source](https://www.mcclimon.org/portfolio/embedded-verifier/) - The embedded verifier excludes capped collections, TTL-indexed collections, and collections with nondefault collation. These need additional checks; enabled verification alone does not establish complete coverage. (low confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/mongosync/current/reference/verification/embedded.md) - The general verification overview directs users to separate index and metadata comparisons, while the embedded-verifier page says version 1.15 added metadata, index, and view examination. Verification planning must resolve this documentation mismatch against the deployed version. (medium confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/mongosync/current/reference/verification.md) - The standalone verifier offers hash-and-length comparison to reduce network traffic, but documents reduced sensitivity to some numeric type changes and less mismatch detail. Faster verification can change what discrepancies it detects. (low confidence) (verified as of 2026-10-27) — [source](https://github.com/mongodb-labs/migration-verifier) - AWS DMS documents validation queries consuming extra source, target and network resources, with additional runtime dependent on migrated data volume. This supports budgeting verification separately, without quantifying MongoDB overhead. (low confidence) (verified as of 2026-10-30) — [source](https://docs.aws.amazon.com/dms/latest/userguide/CHAP_Validating.html) - From mongosync 1.15, embedded verification examines collection metadata, indexes and views; from 1.22 it starts at Change Event Application by default. (low confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/mongosync/current/reference/verification/embedded.md) - The standalone migration-verifier stores metadata on the destination by default, adding concurrent verification workload. Its README suggests a separate metadata cluster through --metaURI if that workload overwhelms the destination. (low confidence) (verified as of 2026-10-27) — [source](https://github.com/mongodb-labs/migration-verifier) - AWS DMS data-validation endpoint lists omit MongoDB and DocumentDB. Its validation functionality therefore must not be assumed to validate MongoDB-target migrations. (low confidence) (verified as of 2026-10-30) — [source](https://docs.aws.amazon.com/dms/latest/userguide/CHAP_Validating.html) - Planning inference: a verification budget should include verifier compute, database workload and operator effort for exceptions or restarts; the sources do not justify a fixed percentage or dollar surcharge. (medium confidence) — [source](https://www.mongodb.com/docs/mongosync/current/reference/verification/embedded.md)