Alloc/Free Job Resource done by Prolog/Epilog phase of Plugin (#28)

Author: ohtanim

commands/qrun/daapi/README.md

@@ -87,6 +87,8 @@ Once you specify all required configuration parameters, you can create a Client
 let client = ClientBuilder::new("http://localhost:8290")
     .with_auth(AuthMethod::IbmCloudIam {
         apikey: "your_iam_apikey".to_string(),
+        service_crn: "your_service_crn".to_string(),
+        iam_endpoint_url: "your_iam_endpoint_url".to_string(),
     })
     .with_timeout(Duration::from_secs(60))
     .with_retry_policy(retry_policy)

commands/qrun/daapi_c/build.rs

@@ -1,3 +1,14 @@
+//
+// (C) Copyright IBM 2025
+//
+// This code is licensed under the Apache License, Version 2.0. You may
+// obtain a copy of this license in the LICENSE.txt file in the root directory
+// of this source tree or at http://www.apache.org/licenses/LICENSE-2.0.
+//
+// Any modifications or derivative works of this code must retain this
+// copyright notice, and modified files need to carry a notice indicating
+// that they have been altered from the originals.
+
 fn main() {
     for (key, value) in std::env::vars() {
         eprintln!("{key}: {value}");

commands/qrun/qrun/Cargo.toml

@@ -6,6 +6,7 @@ edition = "2021"
 [features]
 default = ["ibmcloud_appid_auth"]
 ibmcloud_appid_auth = []
+job_cleanup = []
 
 [dependencies]
 direct-access-api = { workspace = true }

commands/qrun/qrun/src/main.rs

@@ -9,6 +9,7 @@
 // copyright notice, and modified files need to carry a notice indicating
 // that they have been altered from the originals.
 
+#![allow(unused_imports)]
 use std::fs::File;
 use std::fs::OpenOptions;
 use std::io::prelude::*;
@@ -61,6 +62,7 @@ struct Args {
 }
 
 // Handle signals, and cancel QPU job if SIGTERM is received.
+#[cfg(feature = "job_cleanup")]
 async fn handle_signals(mut signals: Signals, job: PrimitiveJob) {
     while let Some(signal) = signals.next().await {
         // To cancel a job, invoke scancel without --signal option. This will send
@@ -205,7 +207,9 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
         .unwrap();
 
     // scancel related signals
+    #[cfg(feature = "job_cleanup")]
     let signals = Signals::new([SIGTERM, SIGCONT])?;
+    #[cfg(feature = "job_cleanup")]
     let handle = signals.handle();
 
     let f = File::open(args.input).expect("file not found");
@@ -225,6 +229,7 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
         )
         .await?;
 
+    #[cfg(feature = "job_cleanup")]
     let signals_task = tokio::spawn(handle_signals(signals, primitive_job.clone()));
 
     let mut succeeded: bool = true;
@@ -250,6 +255,7 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
                 "Error occurred while waiting for final state: {:?}",
                 e.to_string()
             );
+            #[cfg(feature = "job_cleanup")]
             let _ = primitive_job.cancel(false).await;
             succeeded = false;
         }
@@ -289,10 +295,13 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
         }
     }
 
-    client.delete_job(&primitive_job.job_id).await?;
+    #[cfg(feature = "job_cleanup")]
+    {
+        client.delete_job(&primitive_job.job_id).await?;
 
-    handle.close();
-    signals_task.await?;
+        handle.close();
+        signals_task.await?;
+    }
 
     Ok(())
 }

demo/qrun/etc/cloud_secrets/directaccess_service_crn

@@ -0,0 +1 @@
+YOUR_PROVISIONED_INSTANCE

demo/qrun/etc/cloud_secrets/ibmcloud_iam_apikey

@@ -0,0 +1 @@
+YOUR_IAM_APIKEY

demo/qrun/etc/slurm/prolog.d/set_cloud_env.sh

@@ -0,0 +1,29 @@
+#!/bin/bash
+# /etc/slurm/prolog.d/set_cloud_env.sh
+
+if [ -n $SLURM_JOB_PARTITION ]
+then
+  if [ $SLURM_JOB_PARTITION == "staging" ]
+  then
+    echo "export IBMQRUN_AWS_ACCESS_KEY_ID=$(cat /etc/cloud_secrets/aws_access_key)"
+    echo "export IBMQRUN_AWS_SECRET_ACCESS_KEY=$(cat /etc/cloud_secrets/aws_secret_access_key)"
+    echo "export IBMQRUN_IAM_APIKEY=$(cat /etc/cloud_secrets/ibmcloud_iam_apikey)"
+    echo "export IBMQRUN_SERVICE_CRN=$(cat /etc/cloud_secrets/directaccess_service_crn)"
+    echo "export IBMQRUN_IAM_ENDPOINT=https://iam.test.cloud.ibm.com"
+    echo "export IBMQRUN_DAAPI_ENDPOINT=http://127.0.0.1:8080"
+    echo "export IBMQRUN_S3_ENDPOINT=https://s3.us-east.cloud-object-storage.appdomain.cloud"
+    echo "export IBMQRUN_S3_BUCKET=<your bucket name>"
+    echo "export IBMQRUN_S3_REGION=us-east"
+  elif [ $SLURM_JOB_PARTITION == "normal" ]
+  then
+    echo "export IBMQRUN_AWS_ACCESS_KEY_ID=minioadmin"
+    echo "export IBMQRUN_AWS_SECRET_ACCESS_KEY=minioadmin"
+    echo "export IBMQRUN_IAM_APIKEY=demoapikey1"
+    echo "export IBMQRUN_SERVICE_CRN=crn:v1:local:daa_sim"
+    echo "export IBMQRUN_IAM_ENDPOINT=http://daapi:8290"
+    echo "export IBMQRUN_DAAPI_ENDPOINT=http://daapi:8290"
+    echo "export IBMQRUN_S3_ENDPOINT=http://minio:9000"
+    echo "export IBMQRUN_S3_BUCKET=test"
+    echo "export IBMQRUN_S3_REGION=us-east"
+  fi
+fi

demo/qrun/etc/slurm/slurm.conf

@@ -36,7 +36,7 @@ ReturnToService=0
 #SrunProlog=
 #SrunEpilog=
 TaskProlog=/etc/slurm/prolog.d/set_cloud_env.sh
-TaskEpilog=/etc/slurm/epilog.d/unset_cloud_env.sh
+#TaskEpilog=
 #TaskPlugin=
 #TrackWCKey=no
 #TreeWidth=50

plugins/spank_ibm_qrun/CMakeLists.txt

@@ -8,6 +8,12 @@ project (spank_ibm_qrun C)
 include(CheckIncludeFiles)
 include(FindPackageHandleStandardArgs)
 
+set(TARGET_NAME spank_ibm_qrun)
+set(DAAPI_C_LIB
+  -L${CMAKE_CURRENT_SOURCE_DIR}/../../commands/qrun/target/release
+  -ldirect_access_capi
+)
+
 find_path(SLURM_INCLUDE_DIR NAMES slurm/slurm.h)
 find_library(SLURM_LIBRARY NAMES libslurm.so)
 find_package_handle_standard_args(SLURM DEFAULT_MSG SLURM_LIBRARY SLURM_INCLUDE_DIR)
@@ -23,6 +29,13 @@ else (NOT SLURM_FOUND)
 endif (NOT SLURM_FOUND)
 mark_as_advanced (SLURM_LIBRARIES SLURM_INCLUDE_DIRS)
 
-add_library (spank_ibm_qrun MODULE spank_ibm_qrun.c)
+add_library (${TARGET_NAME} MODULE spank_ibm_qrun.c)
 include_directories (BEFORE ${SLURM_INCLUDE_DIRS})
-set_target_properties (spank_ibm_qrun PROPERTIES PREFIX "" SUFFIX "" OUTPUT_NAME "spank_ibm_qrun.so")
+target_include_directories(${TARGET_NAME} PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/../../commands/qrun/daapi_c/)
+target_link_libraries(${TARGET_NAME} PRIVATE ${DAAPI_C_LIB})
+target_link_libraries(${TARGET_NAME} PUBLIC "-lssl")
+target_link_libraries(${TARGET_NAME} PUBLIC "-lcrypto")
+target_link_libraries(${TARGET_NAME} PUBLIC "-ldl")
+target_link_libraries(${TARGET_NAME} PUBLIC "-lm")
+target_link_libraries(${TARGET_NAME} PUBLIC "-lpthread")
+set_target_properties (${TARGET_NAME} PROPERTIES PREFIX "" SUFFIX "" OUTPUT_NAME "spank_ibm_qrun.so")