[netomi]

What

MigrationItemJobRequestHandler used to pull up to 25000 not-yet-migrated migration_item rows every 15 minutes and enqueue() all of them immediately (due "now"). I confirmed by inspecting JobRunr's own polling code (Paging.AmountBasedList.ascOnUpdatedAt, used by OnboardNewWorkTask) that every ENQUEUED/due job across the whole server sits in one strict FIFO queue ordered by due time — there's no per-job-type priority in open-source JobRunr. So a burst of 25000 jobs all due "now" meant any real, user-triggered job (extension published, signature generated, etc.) enqueued around the same time would sort in behind however many of those 25000 came first, delaying it until the burst drained.

How

• MigrationItemJobRequestHandler.batchSize: configurable via ovsx.migrations.batch-size (default 200, down from the hardcoded 25000) — directly bounds how many migration jobs can ever be due ahead of a real job at once.
• Each item in a batch is now scheduler.schedule()d at a staggered offset (200ms apart) instead of enqueue()d all for "now" — a full batch at the default size spreads over ~40s, comfortably inside the default 15-minute recurring interval.
• MigrationScheduler's recurring cron: configurable via ovsx.migrations.cron (default */15 * * * *, identical to the previous hardcoded Cron.every15minutes()).
• MigrationService.enqueueMigration now takes the target Instant and calls scheduler.schedule(...) instead of scheduler.enqueue(...) so the staggering above actually takes effect.
• Logging: now reports how many items remain after each batch (new RepositoryService.countNotMigratedItems()), since draining a large backlog can now take many recurring runs instead of one, and the old per-batch count alone gave no sense of overall progress. Skipped entirely when a batch finds nothing to do, so it doesn't log a pointless "scheduled 0 items" line (and doesn't run the count query behind it) on ticks where there's nothing left — an empty slice already means hasNext() is false too, so the "migration completed" branch still fires correctly on its own.

Testing

• MigrationItemJobRequestHandlerTest (new): read-only short-circuit, configured batch size is honored, items get strictly-increasing evenly-spaced schedule times, remaining-count query runs only when items were found, recurring-job deletion logic in both directions.
• MigrationSchedulerTest (new): configured cron is passed through to scheduleRecurrently.
• Full unitTests suite passes (957 tests).

Note

I also looked into detecting "the current batch has fully drained" and immediately scheduling the next one, instead of relying on the cron interval at all. That's possible but needs a schema change (a way to distinguish "still in flight" from "permanently failed and parked for investigation", which currently look identical in migration_item) — happy to pursue as a follow-up if the cron-based approach here turns out not to be responsive enough in practice.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR reduces the impact of migration backlogs on user-triggered JobRunr jobs by throttling how many migration items are scheduled per run and by staggering their due times, while also making migration scheduling configurable and improving progress logging/visibility.

Changes:

• Introduces configurable migration batch sizing and staggered scheduling so migration items don’t all become due “now”.
• Makes the recurring migration scheduling cron configurable.
• Adds backlog-counting and new unit tests covering the scheduling/throttling behavior.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
server/src/main/java/org/eclipse/openvsx/migration/MigrationItemJobRequestHandler.java	Limits batch size via config and staggers scheduled instants; adds backlog logging.
server/src/main/java/org/eclipse/openvsx/migration/MigrationScheduler.java	Makes recurring cron configurable and passes it to JobRunr.
server/src/main/java/org/eclipse/openvsx/migration/MigrationService.java	Changes migration job submission from immediate enqueue to scheduled-at-instant.
server/src/main/java/org/eclipse/openvsx/repositories/RepositoryService.java	Adds a service method to count not-yet-migrated items for logging/progress.
server/src/main/java/org/eclipse/openvsx/repositories/MigrationItemRepository.java	Adds a Spring Data count query for not-yet-migrated items.
server/src/test/java/org/eclipse/openvsx/migration/MigrationItemJobRequestHandlerTest.java	New tests for batch sizing, staggering, backlog-count query behavior, and recurring-job deletion behavior.
server/src/test/java/org/eclipse/openvsx/migration/MigrationSchedulerTest.java	New test verifying the configured cron is passed through to scheduleRecurrently.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.