# Upgrading to v0.7

This release introduces a rules-based Pro Pruner, chunk workers, and an Erlang notifier. It
requires Oban v0.7 and adds new tables and an index to the schema.

Reinstalling the schema is required, and the pruner's `interval` option was removed. Clusters
shared with Elixir nodes should also switch the pruner to manual sync.

## Bump Your Deps

Update Oban and Pro to the latest versions:

```toml
dependencies = [
  "oban>=0.7.0",
  "oban-pro>=0.7.0",
]
```

Then upgrade both packages, reinstalling Pro to renew it:

```bash
uv sync --upgrade-package oban --upgrade-package oban-pro --reinstall-package oban-pro
```

Oban v0.7 adds job timeouts, a shutdown grace period, and the extension points the Pro Pruner is
built on.

## Reinstall the Schema (Required)

Oban Pro v0.7 adds the `oban_pruners` and `oban_jobs_archive` tables, an index for chunk lookups,
and an updated `oban_state_to_bit` function. Run the install command again:

```bash
obanpro install
```

Or apply `install_sql()` through a new migration in your migration framework. Installation is
idempotent, and only the missing tables, indexes, and functions are created.

On large `oban_jobs` tables, creating `oban_jobs_chunk_index` blocks writes while it builds.
Create the index concurrently beforehand, and the install will skip it:

```sql
CREATE INDEX CONCURRENTLY IF NOT EXISTS oban_jobs_chunk_index
ON oban_jobs ((meta->>'chunk_id'), state, priority, scheduled_at, id)
WHERE meta ? 'chunk_id';
```

## Replace the Pruner Interval

The Pro Pruner runs on a cron schedule rather than an interval, and the `interval` option now
raises an error. Replace it with an equivalent `schedule`:

```diff
 [pruner]
-interval = 60
+schedule = "* * * * *"
```

The top-level `max_age` still applies to the default rule. Pruning now skips jobs that belong to
an active workflow, so jobs from long-running workflows are kept longer than before.

## Use Manual Pruner Sync in Mixed Clusters

Configured pruner rules are the source of truth by default, and persisted rules missing from the
configuration are deleted on start. In a cluster shared with Elixir nodes, or where rules are
managed from Oban Web, switch to manual sync before upgrading so Python nodes don't delete rules
they didn't configure:

```toml
[pruner]
sync_mode = "manual"
```