Merge pull request #312 from casparvl/adapt_arch_target_map

Adapt arch target map

Author: bedroge

app.cfg.example

@@ -305,19 +305,59 @@ signing =
 
 
 [architecturetargets]
-# defines for which architectures the bot will build and what job submission
-#   parameters shall be used to allocate a compute node with the correct
-arch_target_map = {
-    "linux/x86_64/generic": "--partition x86-64-generic-node",
-    "linux/x86_64/amd/zen2": "--partition x86-64-amd-zen2-node" }
+# arch_target_map has been replaced by node_type_map
+# arch_target_map = {
+# }
 
+# Each entry in the node_type_map dictionary describes a build node type. The key is a (descriptive) name for this build node, and its value is a dictionary containing the following build node properties as key-value pairs:
+  - os: its operating system (os)
+  - cpu_subdir: its CPU architecture
+  - slurm_params: the SLURM parameters that need to be passed to submit jobs to it
+  - repo_targets: supported repository targets for this node type
+  - accel (optional): which accelerators this node has
+# All are strings, except repo_targets, which is a list of strings.
+# Note that the Slurm parameters should typically be chosen such that a single type of node (with one specific type of
+# CPU and one specific type of GPU) should be allocated.
+# Below is an example configuration for a system that contains 4 types of nodes: zen2 CPU nodes, zen4 CPU nodes,
+# GPU nodes with an icelake CPU and A100 GPU, GPU nodes with a zen4 CPU and an H100 GPU.
+# The 'on:' argument to the bot build command determines which node type will be allocated for the build job,
+# e.g. 'bot:build on:arch=zen4,accel=nvidia/cc90 for:...' will match the gpu_h100 node type below.
+# If no 'on:' argument is passed to the build command, the 'for:' argument is used instead,
+# e.g. 'bot:build for:arch=icelake,accel=nvidia/cc80' will match the gpu_a100 node type below.
+node_type_map = {
+    "cpu_zen2": {
+        "os": "linux",
+        "cpu_subdir": "x86_64/amd/zen2",
+        "slurm_params": "-p rome --nodes 1 --ntasks-per-node 16 --cpus-per-task 1",
+        "repo_targets": ["eessi.io-2023.06-compat","eessi.io-2023.06-software"]
+    },
+    "cpu_zen4": {
+        "os": "linux",
+        "cpu_subdir": "x86_64/amd/zen4",
+        "accel": "None",
+        "slurm_params": "-p genoa --nodes 1 --ntasks-per-node 24 --cpus-per-task 1",
+        "repo_targets": ["eessi.io-2023.06-compat","eessi.io-2023.06-software"]
+    },
+    "gpu_a100": {
+        "os": "linux",
+        "cpu_subdir": "x86_64/intel/icelake",
+        "accel": "nvidia/cc80",
+        "slurm_params": "-p gpu_a100 --nodes 1 --tasks-per-node 18 --cpus-per-task 1 --gpus-per-node 1",
+        "repo_targets": ["eessi.io-2023.06-compat","eessi.io-2023.06-software"]
+    },
+    "gpu_h100": {
+        "os": "linux",
+        "cpu_subdir": "x86_64/amd/zen4",
+        "accel": "nvidia/cc90",
+        "slurm_params": "-p gpu_h100 --nodes 1 --tasks-per-node 16 --cpus-per-task 1 --gpus-per-node 1",
+        "repo_targets": ["eessi.io-2023.06-compat","eessi.io-2023.06-software"]
+    }}
 
 [repo_targets]
-# defines for which repository a arch_target should be build for
-#
-# EESSI/2023.06 and EESSI/2025.06
-repo_target_map = {
-    "linux/x86_64/amd/zen2" : ["eessi.io-2023.06-software","eessi.io-2025.06-software"] }
+
+# No longer used, repo targets are now specified per node type in the node_type_map
+# repo_target_map = {
+#     "linux/x86_64/amd/zen2" : ["eessi.io-2023.06-software","eessi.io-2025.06-software"] }
 
 # points to definition of repositories (default repository defined by build container)
 repos_cfg_dir = PATH_TO_SHARED_DIRECTORY/repos
@@ -360,8 +400,12 @@ scontrol_command = /usr/bin/scontrol
 # awaits_release = job id `{job_id}` awaits release by job manager
 awaits_release_delayed_begin_msg = job id `{job_id}` will be eligible to start in about {delay_seconds} seconds
 awaits_release_hold_release_msg = job id `{job_id}` awaits release by job manager
-initial_comment = New job on instance `{app_name}` for CPU micro-architecture `{arch_name}`{accelerator_spec} for repository `{repo_id}` in job dir `{symlink}`
+new_job_instance_repo = New job on instance `{app_name}` for repository `{repo_id}`
+build_on_arch = Building on: `{on_arch}`{on_accelerator}
+build_for_arch = Building for: `{for_arch}`{for_accelerator}
+jobdir = Job dir: `{symlink}`
 with_accelerator =  and accelerator `{accelerator}`
+# initial_comment = New job on instance `{app_name}` for repository `{repo_id}`\nBuilding on: `{on_arch}`{on_accelerator}\nBuilding for: `{for_arch}`{for_accelerator}\nJob dir: `{symlink}`  # no longer used
 
 
 [new_job_comments]

eessi_bot_event_handler.py

@@ -29,8 +29,8 @@
 
 # Local application imports (anything from EESSI/eessi-bot-software-layer)
 from connections import github
-from tasks.build import check_build_permission, get_architecture_targets, get_repo_cfg, \
-    request_bot_build_issue_comments, submit_build_jobs
+from tasks.build import check_build_permission, get_node_types, request_bot_build_issue_comments, \
+    submit_build_jobs
 from tasks.deploy import deploy_built_artefacts, determine_job_dirs
 from tasks.clean_up import move_to_trash_bin
 from tools import config
@@ -43,7 +43,7 @@
 
 REQUIRED_CONFIG = {
     config.SECTION_ARCHITECTURETARGETS: [
-        config.ARCHITECTURETARGETS_SETTING_ARCH_TARGET_MAP],       # required
+        config.NODE_TYPE_MAP],       # required
     config.SECTION_BOT_CONTROL: [
         # config.BOT_CONTROL_SETTING_CHATLEVEL,                      # optional
         config.BOT_CONTROL_SETTING_COMMAND_PERMISSION,             # required
@@ -104,10 +104,12 @@
     config.SECTION_JOB_MANAGER: [
         config.JOB_MANAGER_SETTING_POLL_INTERVAL],                 # required
     config.SECTION_REPO_TARGETS: [
-        config.REPO_TARGETS_SETTING_REPO_TARGET_MAP,               # required
         config.REPO_TARGETS_SETTING_REPOS_CFG_DIR],                # required
     config.SECTION_SUBMITTED_JOB_COMMENTS: [
-        config.SUBMITTED_JOB_COMMENTS_SETTING_INITIAL_COMMENT,     # required
+        config.SUBMITTED_JOB_COMMENTS_SETTING_INSTANCE_REPO,       # required
+        config.SUBMITTED_JOB_COMMENTS_SETTING_BUILD_ON_ARCH,       # required
+        config.SUBMITTED_JOB_COMMENTS_SETTING_BUILD_FOR_ARCH,      # required
+        config.SUBMITTED_JOB_COMMENTS_SETTING_JOBDIR,              # required
         # config.SUBMITTED_JOB_COMMENTS_SETTING_AWAITS_RELEASE,      # optional
         config.SUBMITTED_JOB_COMMENTS_SETTING_AWAITS_RELEASE_DELAYED_BEGIN_MSG,  # required
         config.SUBMITTED_JOB_COMMENTS_SETTING_AWAITS_RELEASE_HOLD_RELEASE_MSG,   # required
@@ -411,23 +413,21 @@ def handle_pull_request_opened_event(self, event_info, pr, req_chatlevel=ChatLev
         app_name = self.cfg[config.SECTION_GITHUB][config.GITHUB_SETTING_APP_NAME]
         # TODO check if PR already has a comment with arch targets and
         # repositories
-        arch_map = get_architecture_targets(self.cfg)
-        repo_cfg = get_repo_cfg(self.cfg)
-
-        comment = f"Instance `{app_name}` is configured to build for:"
-        architectures = ['/'.join(arch.split('/')[1:]) for arch in arch_map.keys()]
-        comment += "\n- architectures: "
-        if len(architectures) > 0:
-            comment += f"{', '.join([f'`{arch}`' for arch in architectures])}"
-        else:
-            comment += "none"
-        repositories = list(set([repo_id for repo_ids in repo_cfg[config.REPO_TARGETS_SETTING_REPO_TARGET_MAP].values()
-                            for repo_id in repo_ids]))
-        comment += "\n- repositories: "
-        if len(repositories) > 0:
-            comment += f"{', '.join([f'`{repo_id}`' for repo_id in repositories])}"
-        else:
-            comment += "none"
+        node_map = get_node_types(self.cfg)
+
+        comment = f"Instance `{app_name}` is configured to build on:"
+        for node in node_map:
+            comment += f"\n- Node type `{node}`:"
+            current_node_type = node_map[node]
+            if "os" in current_node_type:
+                comment += f"\n  - OS: `{current_node_type['os']}`"
+            if "cpu_subdir" in current_node_type:
+                comment += f"\n  - CPU architecture: `{current_node_type['cpu_subdir']}`"
+            if "repo_targets" in current_node_type:
+                comment += f"\n  - Repositories: `{current_node_type['repo_targets']}`"
+            if "accel" in current_node_type:
+                comment += f"\n  - Accelerators: `{current_node_type['accel']}`"
+            comment += "\n"
 
         self.log(f"PR opened: comment '{comment}'")
 
@@ -532,7 +532,7 @@ def handle_bot_command_build(self, event_info, bot_command):
         build_msg = ''
         if check_build_permission(pr, event_info):
             # use filter from command
-            submitted_jobs = submit_build_jobs(pr, event_info, bot_command.action_filters)
+            submitted_jobs = submit_build_jobs(pr, event_info, bot_command.action_filters, bot_command.build_params)
             if submitted_jobs is None or len(submitted_jobs) == 0:
                 build_msg = "\n  - no jobs were submitted"
             else:
@@ -578,8 +578,8 @@ def handle_bot_command_status(self, event_info, bot_command):
             bot_command (EESSIBotCommand): command to be handled
 
         Returns:
-            github.IssueComment.IssueComment (note, github refers to
-                 PyGithub, not the github from the internal connections module)
+            (string): list item with a link to the issue comment that was created
+                containing the status overview
         """
         self.log("processing bot command 'status'")
         repo_name = event_info['raw_request_body']['repository']['full_name']
@@ -588,18 +588,23 @@ def handle_bot_command_status(self, event_info, bot_command):
 
         comment_status = ''
         comment_status += "\nThis is the status of all the `bot: build` commands:"
-        comment_status += "\n|arch|result|date|status|url|"
-        comment_status += "\n|----|------|----|------|---|"
+        comment_status += "\n|on|for|repo|result|date|status|url|"
+        comment_status += "\n|----|----|----|------|----|------|---|"
         for x in range(0, len(status_table['date'])):
-            comment_status += f"\n|{status_table['arch'][x]}|"
+            comment_status += f"\n|{status_table['on arch'][x]}|"
+            comment_status += f"{status_table['for arch'][x]}|"
+            comment_status += f"{status_table['for repo'][x]}|"
             comment_status += f"{status_table['result'][x]}|"
             comment_status += f"{status_table['date'][x]}|"
             comment_status += f"{status_table['status'][x]}|"
             comment_status += f"{status_table['url'][x]}|"
 
         self.log(f"Overview of finished builds: comment '{comment_status}'")
         issue_comment = create_comment(repo_name, pr_number, comment_status, ChatLevels.MINIMAL)
-        return issue_comment
+        if issue_comment:
+            return f"\n  - added status comment {issue_comment.html_url}"
+        else:
+            return "\n  - failed to create status comment"
 
     def start(self, app, port=3000):
         """
@@ -692,7 +697,7 @@ def main():
     opts = event_handler_parse()
 
     # config is read and checked for settings to raise an exception early when the event_handler starts.
-    if config.check_required_cfg_settings(REQUIRED_CONFIG):
+    if config.check_cfg_settings(REQUIRED_CONFIG):
         print("Configuration check: PASSED")
     else:
         print("Configuration check: FAILED")

eessi_bot_job_manager.py

@@ -623,7 +623,7 @@ def main():
 
     # config is read and checked for settings to raise an exception early when
     # the job_manager runs
-    if config.check_required_cfg_settings(REQUIRED_CONFIG):
+    if config.check_cfg_settings(REQUIRED_CONFIG):
         print("Configuration check: PASSED")
     else:
         print("Configuration check: FAILED")