apollo_storage: always spawn storage reader server, return 503 when disabled

[nadin-Starkware]

Note

Medium Risk
Touches shared storage/server wiring used across multiple components and changes disabled-mode behavior from “no listener” to “listener that 503s”, which may affect operational expectations and port allocation.

Overview
Storage now always creates and spawns the storage reader server (no more Option), and a disabled server returns 503 Service Unavailable to all requests instead of not starting.

This updates batcher, state-sync, and class-manager to hold a mandatory keep-alive AbortHandle, switches call sites to storage_reader_server.spawn(), and adjusts tests/integration configs to allocate distinct storage-reader ports (including a new class-manager test identifier and updated hybrid overlay port).

^{Written by Cursor Bugbot for commit ecdc945. This will update automatically on new commits. Configure here.}

[reviewable-StarkWare]

This change is 

[nadin-Starkware]

• apollo_storage: always spawn storage reader server, return 503 when disabled #12836 : 3 dependent PRs (#12876 , #12940 , #13117 ) 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[github-actions]

Artifacts upload workflows:

• Started at 2026-02-24T13:17:27Z
• Started at 2026-02-24T14:36:21Z
• Started at 2026-02-24T16:45:26Z
• Started at 2026-02-24T16:59:18Z
• Started at 2026-02-25T06:38:31Z
• Started at 2026-02-25T06:39:51Z
• Started at 2026-02-25T06:47:05Z
• Started at 2026-02-25T07:08:52Z
• Started at 2026-02-25T07:26:46Z
• Started at 2026-02-25T07:47:50Z
• Started at 2026-02-25T08:51:41Z
• Started at 2026-02-25T09:08:18Z
• Started at 2026-02-25T09:31:30Z
• Started at 2026-02-25T10:17:40Z
• Started at 2026-02-25T10:51:20Z
• Started at 2026-02-25T10:56:34Z
• Started at 2026-02-25T10:56:34Z
• Started at 2026-02-25T11:18:24Z
• Started at 2026-02-25T11:19:43Z
• Started at 2026-02-25T11:19:42Z
• Started at 2026-02-25T12:15:00Z
• Started at 2026-02-25T12:26:48Z
• Started at 2026-02-26T05:41:08Z
• Started at 2026-03-04T09:23:47Z
• Started at 2026-03-04T09:33:24Z
• Started at 2026-03-04T10:18:57Z
• Started at 2026-03-04T10:40:22Z
• Started at 2026-03-04T13:16:16Z
• Started at 2026-03-04T13:46:07Z
• Started at 2026-03-05T08:05:05Z
• Started at 2026-03-05T12:39:27Z
• Started at 2026-03-09T13:25:17Z
• Started at 2026-03-09T14:22:22Z
• Started at 2026-03-09T14:29:57Z
• Started at 2026-03-10T08:35:49Z
• Started at 2026-03-10T11:33:10Z
• Started at 2026-03-10T13:09:47Z

[graphite-app]

Merge activity

• Feb 25, 6:40 AM UTC: Graphite rebased this pull request, because this pull request is set to merge when ready.
• Mar 4, 9:24 AM UTC: This pull request can not be added to the Graphite merge queue. Please try rebasing and resubmitting to merge when ready.
• Mar 4, 9:24 AM UTC: Graphite disabled "merge when ready" on this PR due to: a merge conflict with the target branch; resolve the conflict and try again..

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

[Itay-Tsabary-Starkware]

@Itay-Tsabary-Starkware made 2 comments.
Reviewable status: 0 of 20 files reviewed, 1 unresolved discussion (waiting on nadin-Starkware).

---

crates/apollo_storage/src/storage_reader_server.rs line 161 at r12 (raw file):

            storage_reader: self.storage_reader.clone(),
            enabled: self.enabled,
            _phantom: PhantomData,

Is this still required? Even if not, can be in a different pr

Code quote:

_phantom: PhantomData,

---

crates/apollo_storage/src/storage_reader_server.rs line 202 at r12 (raw file):

    {
        let socket = SocketAddr::from((self.config.ip(), self.config.port()));
        let enabled_status = if self.config.is_enabled() { "enabled" } else { "disabled" };

IMO:

1. either define a proper enum instead of using a boolean value
2. print "is_enabled: {bool value}
   2 is simpler

[Itay-Tsabary-Starkware]

@Itay-Tsabary-Starkware made 2 comments.
Reviewable status: 0 of 20 files reviewed, 3 unresolved discussions (waiting on nadin-Starkware).

---

crates/apollo_storage/src/storage_reader_server_test.rs line 114 at r12 (raw file):

    let mut available_ports =
        AvailablePorts::new(TestIdentifier::StorageReaderServerUnitTests.into(), 2);

Should be using the unique_id! macro

Code quote:

2

---

crates/apollo_storage/src/storage_reader_server_test.rs line 169 at r12 (raw file):

    let mut available_ports =
        AvailablePorts::new(TestIdentifier::StorageReaderServerUnitTests.into(), 4);

Should be using the unique_id! macro
(throughout)

Code quote:

4

[Itay-Tsabary-Starkware]

@Itay-Tsabary-Starkware made 1 comment.
Reviewable status: 0 of 20 files reviewed, 4 unresolved discussions (waiting on nadin-Starkware).

---

deployments/sequencer/configs/overlays/hybrid/testing/node-0/services/core.yaml line 104 at r12 (raw file):

  - name: state-sync-storage-reader
    port: 55014
    targetPort: 55014

why?

Code quote:

    port: 55014
    targetPort: 55014

[Itay-Tsabary-Starkware]

@Itay-Tsabary-Starkware made 1 comment.
Reviewable status: 0 of 20 files reviewed, 4 unresolved discussions (waiting on nadin-Starkware).

---

crates/apollo_class_manager/src/class_storage.rs line 289 at r12 (raw file):

        let storage_reader_server_handle = Arc::new(tokio::spawn(async {}).abort_handle());
        Ok(Self { reader, writer: Arc::new(Mutex::new(writer)), storage_reader_server_handle })
    }

Let's keep as is for this pr, but we should change this, this essentially having another config value (run the server or the mock process) manifested as a different function.
If no other alternatives I'd even prefer we actually have this as a config value.
Could you come up with some other way? A dependency injection of some sort?

Code quote:

    pub fn new(
        storage_config: StorageConfig,
        storage_reader_server_config: ServerConfig,
    ) -> ClassHashStorageResult<Self> {
        let (reader, writer, storage_reader_server) =
            apollo_storage::open_storage_with_metric_and_server::<
                GenericStorageReaderServerHandler,
                StorageReaderRequest,
                StorageReaderResponse,
            >(
                storage_config,
                &CLASS_MANAGER_STORAGE_OPEN_READ_TRANSACTIONS,
                storage_reader_server_config,
            )?;

        let storage_reader_server_handle = Arc::new(storage_reader_server.spawn());

        Ok(Self { reader, writer: Arc::new(Mutex::new(writer)), storage_reader_server_handle })
    }

    /// Opens the storage without a storage reader server or metrics.
    #[cfg(any(feature = "testing", test))]
    pub(crate) fn new_plain(storage_config: StorageConfig) -> ClassHashStorageResult<Self> {
        let (reader, writer) = open_storage(storage_config)?;
        let storage_reader_server_handle = Arc::new(tokio::spawn(async {}).abort_handle());
        Ok(Self { reader, writer: Arc::new(Mutex::new(writer)), storage_reader_server_handle })
    }

[nadin-Starkware]

@nadin-Starkware made 4 comments.
Reviewable status: 0 of 20 files reviewed, 4 unresolved discussions (waiting on Itay-Tsabary-Starkware).

---

crates/apollo_storage/src/storage_reader_server.rs line 202 at r12 (raw file):

Previously, Itay-Tsabary-Starkware wrote…

IMO:

1. either define a proper enum instead of using a boolean value
2. print "is_enabled: {bool value}
   2 is simpler

Done.

---

crates/apollo_storage/src/storage_reader_server_test.rs line 114 at r12 (raw file):

Previously, Itay-Tsabary-Starkware wrote…

Should be using the unique_id! macro

Done in separate PR

---

crates/apollo_storage/src/storage_reader_server_test.rs line 169 at r12 (raw file):

Previously, Itay-Tsabary-Starkware wrote…

Should be using the unique_id! macro
(throughout)

Done in separate PR

---

deployments/sequencer/configs/overlays/hybrid/testing/node-0/services/core.yaml line 104 at r12 (raw file):

Previously, Itay-Tsabary-Starkware wrote…

why?

55012 is also used by the proof_manager component. There were no conflicts before because the server was disabled, but once we changed it to always spawn, a conflict occurred. So I changed the port to make it unique

[Itay-Tsabary-Starkware]

@Itay-Tsabary-Starkware reviewed 13 files and all commit messages, made 1 comment, and resolved 4 discussions.
Reviewable status: 13 of 20 files reviewed, all discussions resolved.

[nadin-Starkware]

@nadin-Starkware reviewed 7 files.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on nadin-Starkware).