[Secrets] Add jitter to SqlServerSecretsStoreMigrator retry backoff + 8-way concurrent-migrate test #18

Open
opened 2026-11-06 15:25:44 -05:00 by dohertj2 · 0 comments
Owner

SqlServerSecretsStoreMigrator retries its DDL with a flat 100*attempt backoff and no jitter (ZB.MOM.WW.Secrets/src/ZB.MOM.WW.Secrets.Replicator.SqlServer/SqlServerSecretsStoreMigrator.cs:67, verified 2026-11-06). The identical loop ported into Auth 0.3.0's SqlServerApiKeyStoreMigrator exhausted its retries ~1 in 60 eight-way concurrent boots; adding jitter took it to 0/180. Secrets has no concurrent-migrate test, so the defect is presumed latent here (a 2-node central pair is only 2-way, so low practical risk).

Fix: mirror the Auth fix (jittered backoff) and add a Migrate_Concurrently_FromEightTasks_AllSucceed-shaped live test.

Source: ScadaBridge docs/plans/2026-07-08-deferred-work-register.md, register row 109 (belongs to scadaproj).

SqlServerSecretsStoreMigrator retries its DDL with a flat 100*attempt backoff and no jitter (ZB.MOM.WW.Secrets/src/ZB.MOM.WW.Secrets.Replicator.SqlServer/SqlServerSecretsStoreMigrator.cs:67, verified 2026-11-06). The identical loop ported into Auth 0.3.0's SqlServerApiKeyStoreMigrator exhausted its retries ~1 in 60 eight-way concurrent boots; adding jitter took it to 0/180. Secrets has no concurrent-migrate test, so the defect is presumed latent here (a 2-node central pair is only 2-way, so low practical risk). Fix: mirror the Auth fix (jittered backoff) and add a Migrate_Concurrently_FromEightTasks_AllSucceed-shaped live test. Source: ScadaBridge docs/plans/2026-07-08-deferred-work-register.md, register row 109 (belongs to scadaproj).
dohertj2 added the type:bugpriority:p3migrated labels 2026-11-06 15:25:44 -05:00
Sign in to join this conversation.