Copy build log and artifacts to permanents locations after a failure

Packages are be built in some selected build path (--buildpath), and
the logs of successful compilation are then concentrated to some other
location for permanent storage (--logfile-format). Logs of failed builds
remain in the build path location so that they can be inspected.

However, this setup is problematic when building software in HPC jobs. Quite
often in HPC systems the build path is set to some fast storage local to
the node, like NVME raid mounted on `/tmp` or `/dev/shm` (as suggested
in the documentation: https://docs.easybuild.io/configuration/#buildpath).
The node storage is often wiped out after the end of a job, so the log
files and the artifacts are no longer available after the termination of
the job.

This commit adds options (--log-error-path and --artifact-error-path) to
accumulate error logs and artifacts in some more permanent locations, so
that the can be easily inspected after a failed build.

Author: gkaf89

easybuild/framework/easyblock.py

@@ -76,15 +76,17 @@
 from easybuild.tools.config import EASYBUILD_SOURCES_URL # noqa
 from easybuild.tools.config import build_option, build_path, get_log_filename, get_repository, get_repositorypath
 from easybuild.tools.config import install_path, log_path, package_path, source_paths
+from easybuild.tools.config import get_log_error_path, get_artifact_error_path
 from easybuild.tools.environment import restore_env, sanitize_env
 from easybuild.tools.filetools import CHECKSUM_TYPE_MD5, CHECKSUM_TYPE_SHA256
 from easybuild.tools.filetools import adjust_permissions, apply_patch, back_up_file, change_dir, check_lock
-from easybuild.tools.filetools import compute_checksum, convert_name, copy_file, create_lock, create_patch_info
-from easybuild.tools.filetools import derive_alt_pypi_url, diff_files, dir_contains_files, download_file
-from easybuild.tools.filetools import encode_class_name, extract_file
-from easybuild.tools.filetools import find_backup_name_candidate, get_source_tarball_from_git, is_alt_pypi_url
-from easybuild.tools.filetools import is_binary, is_sha256_checksum, mkdir, move_file, move_logs, read_file, remove_dir
-from easybuild.tools.filetools import remove_file, remove_lock, verify_checksum, weld_paths, write_file, symlink
+from easybuild.tools.filetools import convert_name, copy_file, copy_dir, create_lock, create_patch_info
+from easybuild.tools.filetools import create_unused_dirs, derive_alt_pypi_url, diff_files, dir_contains_files
+from easybuild.tools.filetools import download_file, encode_class_name, extract_file, compute_checksum
+from easybuild.tools.filetools import find_backup_name_candidate, get_source_tarball_from_git, det_common_path_prefix
+from easybuild.tools.filetools import is_alt_pypi_url,is_binary, is_predecesor, is_readable, is_sha256_checksum, mkdir
+from easybuild.tools.filetools import move_file, move_logs, read_file, remove_dir, remove_file, remove_lock
+from easybuild.tools.filetools import verify_checksum, weld_paths, write_file, symlink
 from easybuild.tools.hooks import BUILD_STEP, CLEANUP_STEP, CONFIGURE_STEP, EXTENSIONS_STEP, FETCH_STEP, INSTALL_STEP
 from easybuild.tools.hooks import MODULE_STEP, MODULE_WRITE, PACKAGE_STEP, PATCH_STEP, PERMISSIONS_STEP, POSTITER_STEP
 from easybuild.tools.hooks import POSTPROC_STEP, PREPARE_STEP, READY_STEP, SANITYCHECK_STEP, SOURCE_STEP
@@ -1045,16 +1047,23 @@ def moduleGenerator(self):
     #
     # DIRECTORY UTILITY FUNCTIONS
     #
-    def gen_builddir(self):
-        """Generate the (unique) name for the builddir"""
+    def get_relative_builddir_base_path(self):
+        """Generate builddir base name relative to build_path"""
         clean_name = remove_unwanted_chars(self.name)
 
         # if a toolchain version starts with a -, remove the - so prevent a -- in the path name
         tc = self.cfg['toolchain']
         tcversion = tc['version'].lstrip('-')
         lastdir = "%s%s-%s%s" % (self.cfg['versionprefix'], tc['name'], tcversion, self.cfg['versionsuffix'])
 
-        builddir = os.path.join(os.path.abspath(build_path()), clean_name, self.version, lastdir)
+        relative_builddir = os.path.join(clean_name, self.version, lastdir)
+
+        return relative_builddir
+
+    def gen_builddir(self):
+        """Generate the (unique) name for the builddir"""
+        relative_builddir = self.get_relative_builddir_base_path()
+        builddir = os.path.join(os.path.abspath(build_path()), relative_builddir)
 
         # make sure build dir is unique if cleanupoldbuild is False or not set
         if not self.cfg.get('cleanupoldbuild', False):
@@ -4198,6 +4207,88 @@ def print_dry_run_note(loc, silent=True):
     dry_run_msg(msg, silent=silent)
 
 
+def create_persistence_paths(operation_args):
+    persistence_paths = []
+    for op in operation_args:
+        _, _, target_path, _ = op
+        persistence_paths.append(target_path)
+
+    persistence_paths = create_unused_dirs(persistence_paths)
+
+    for i in range(len(operation_args)):
+        op = operation_args[i]
+        operation_args[i] = (op[0], op[1], persistence_paths[i], op[3])
+
+    return operation_args
+
+def execute_persistence_operation(operation, source_paths, target_dir, msg, silent):
+    for p in source_paths:
+        operation(p, target_dir)
+    print_msg(msg, log=_log, silent=silent)
+
+def persist_failed_compilation_log_and_artifacts(build_successful, application_log, silent, app, easyconfig):
+    if not application_log:
+        return
+
+    # there may be multiple log files, or the file name may be different due to zipping
+    logs = glob.glob('%s*' % application_log)
+    print_msg(
+        "Results of the build can be found in the temporary log file(s) %s" % ', '.join(logs),
+        log=_log,
+        silent=silent
+    )
+
+    if build_successful:
+        return
+
+    datetime_stamp = time.strftime("%Y%m%d") + '-' + time.strftime("%H%M%S")
+    operation_args = []
+
+    log_error_path = get_log_error_path(ec=easyconfig)
+    if log_error_path is not None:
+        log_error_path = os.path.join(log_error_path, datetime_stamp)
+
+        if not is_predecesor(app.builddir, log_error_path):
+            operation_args.append(
+                (
+                    copy_file,
+                    logs,
+                    log_error_path,
+                    "Logs of failed build copied to permanent storage: %s" % log_error_path
+                )
+            )
+        else:
+            print_msg(
+                "Persistent log directory is subdirectory of build directory; not copying logs.",
+                log=_log,
+                silent=silent
+            )
+
+    artifact_error_path = get_artifact_error_path(ec=easyconfig)
+    if artifact_error_path is not None:
+        artifact_error_path = os.path.join(artifact_error_path, datetime_stamp)
+
+        if not is_predecesor(app.builddir, artifact_error_path):
+            operation_args.append(
+                (
+                    lambda source, destination : copy_dir(source, destination, dirs_exist_ok=True),
+                    [app.builddir],
+                    artifact_error_path,
+                    "Artifacts of failed build copied to permanent storage: %s" % artifact_error_path
+                )
+            )
+        else:
+            print_msg(
+                "Persistent artifact directory is subdirectory of build directory; not copying artifacts.",
+                log=_log,
+                silent=silent
+            )
+
+    operation_args = create_persistence_paths(operation_args)
+    for op in operation_args:
+        execute_persistence_operation(*op, silent=silent)
+
+
 def build_and_install_one(ecdict, init_env):
     """
     Build the software
@@ -4456,10 +4547,7 @@ def ensure_writable_log_dir(log_dir):
         else:
             dry_run_msg("(no ignored errors during dry run)\n", silent=silent)
 
-    if application_log:
-        # there may be multiple log files, or the file name may be different due to zipping
-        logs = glob.glob('%s*' % application_log)
-        print_msg("Results of the build can be found in the log file(s) %s" % ', '.join(logs), log=_log, silent=silent)
+    persist_failed_compilation_log_and_artifacts(success, application_log, silent, app, ecdict['ec'])
 
     del app
 

easybuild/tools/config.py

@@ -104,9 +104,11 @@
 DEFAULT_MODULE_SYNTAX = 'Lua'
 DEFAULT_MODULES_TOOL = 'Lmod'
 DEFAULT_PATH_SUBDIRS = {
+    'artifact_error_path' : 'error_artifacts',
     'buildpath': 'build',
     'containerpath': 'containers',
     'installpath': '',
+    'log_error_path': 'error_logs',
     'packagepath': 'packages',
     'repositorypath': 'ebfiles_repo',
     'sourcepath': 'sources',
@@ -475,13 +477,15 @@ class ConfigurationVariables(BaseConfigurationVariables):
 
     # list of known/required keys
     REQUIRED = [
+        'artifact_error_path',
         'buildpath',
         'config',
         'containerpath',
         'installpath',
         'installpath_modules',
         'installpath_software',
         'job_backend',
+        'log_error_path',
         'logfile_format',
         'moduleclasses',
         'module_naming_scheme',
@@ -843,6 +847,58 @@ def log_path(ec=None):
     return log_file_format(return_directory=True, ec=ec, date=date, timestamp=timestamp)
 
 
+def get_log_error_path(ec=None):
+    """
+    Return the 'log_error_path', the location where logs are copied in case of failure
+
+    :param ec:  dict-like value with at least the keys 'name' and 'version' defined
+    """
+    log_error_path = ConfigurationVariables()['log_error_path']
+
+    if not log_error_path:
+        return None
+
+    if ec is None:
+        raise EasyBuildError("Easyconfig cannot be empty.")
+
+    name, version = ec.get('name'), ec.get('version')
+
+    if name is None:
+        raise EasyBuildError("The 'name' key is not defined.")
+    if version is None:
+        raise EasyBuildError("The 'version' key is not defined.")
+
+    path = os.path.join(log_error_path, name + '-' + version)
+
+    return path
+
+
+def get_artifact_error_path(ec=None):
+    """
+    Return the 'artifact_error_path', the location where build directories are copied in case of failure
+
+    :param ec:  dict-like value with at least the keys 'name' and 'version' defined
+    """
+    artifact_error_path = ConfigurationVariables()['artifact_error_path']
+
+    if not artifact_error_path:
+        return None
+
+    if ec is None:
+        raise EasyBuildError("Easyconfig cannot be empty.")
+
+    name, version = ec.get('name'), ec.get('version')
+
+    if name is None:
+        raise EasyBuildError("The 'name' key is not defined.")
+    if version is None:
+        raise EasyBuildError("The 'version' key is not defined.")
+
+    path = os.path.join(artifact_error_path, name + '-' + version)
+
+    return path
+
+
 def get_build_log_path():
     """
     Return (temporary) directory for build log

easybuild/tools/options.py

@@ -559,6 +559,9 @@ def config_options(self):
         descr = ("Configuration options", "Configure EasyBuild behavior.")
 
         opts = OrderedDict({
+            'artifact-error-path': ("Location where artifacts are copied in case of an error, "
+                                    "leave empty to avoid copying artifacts",
+                                    None, 'store', mk_full_default_path('artifact_error_path')),
             'avail-module-naming-schemes': ("Show all supported module naming schemes",
                                             None, 'store_true', False,),
             'avail-modules-tools': ("Show all supported module tools",
@@ -589,6 +592,9 @@ def config_options(self):
                                      None, 'store', None),
             'job-backend': ("Backend to use for submitting jobs", 'choice', 'store',
                             DEFAULT_JOB_BACKEND, sorted(avail_job_backends().keys())),
+            'log-error-path': ("Location where logs and artifacts are copied in case of an error, "
+                               "leave empty to avoid copying logs",
+                               None, 'store', mk_full_default_path('log_error_path')),
             # purposely take a copy for the default logfile format
             'logfile-format': ("Directory name and format of the log file",
                                'strtuple', 'store', DEFAULT_LOGFILE_FORMAT[:], {'metavar': 'DIR,FORMAT'}),
@@ -1153,8 +1159,8 @@ def _postprocess_config(self):
         # - the <path> could also specify the location of a *remote* (Git( repository,
         #   which can be done in variety of formats (git@<url>:<org>/<repo>), https://<url>, etc.)
         #   (see also https://github.com/easybuilders/easybuild-framework/issues/3892);
-        path_opt_names = ['buildpath', 'containerpath', 'git_working_dirs_path', 'installpath',
-                          'installpath_modules', 'installpath_software', 'prefix', 'packagepath',
+        path_opt_names = ['artifact_error_path', 'buildpath', 'containerpath', 'git_working_dirs_path', 'installpath',
+                          'installpath_modules', 'installpath_software', 'log_error_path', 'prefix', 'packagepath',
                           'robot_paths', 'sourcepath']
 
         for opt_name in path_opt_names:
@@ -1163,8 +1169,8 @@ def _postprocess_config(self):
         if self.options.prefix is not None:
             # prefix applies to all paths, and repository has to be reinitialised to take new repositorypath in account
             # in the legacy-style configuration, repository is initialised in configuration file itself
-            path_opts = ['buildpath', 'containerpath', 'installpath', 'packagepath', 'repository', 'repositorypath',
-                         'sourcepath']
+            path_opts = ['artifact_error_path', 'buildpath', 'containerpath', 'installpath', 'log_error_path',
+                         'packagepath', 'repository', 'repositorypath', 'sourcepath']
             for dest in path_opts:
                 if not self.options._action_taken.get(dest, False):
                     if dest == 'repository':

test/framework/options.py

@@ -5283,7 +5283,16 @@ def test_prefix_option(self):
 
         regex = re.compile(r"(?P<cfg_opt>\S*).*%s.*" % self.test_prefix, re.M)
 
-        expected = ['buildpath', 'containerpath', 'installpath', 'packagepath', 'prefix', 'repositorypath']
+        expected = [
+            'artifact-error-path',
+            'buildpath',
+            'containerpath',
+            'installpath',
+            'log-error-path',
+            'packagepath',
+            'prefix',
+            'repositorypath',
+        ]
         self.assertEqual(sorted(regex.findall(txt)), expected)
 
     def test_dump_env_script(self):

test/framework/sandbox/sources/toy/toy-0.0_add-bug.patch

@@ -0,0 +1,10 @@
+--- a/toy-0.0.orig/toy.source	2014-03-06 18:48:16.000000000 +0100
++++ b/toy-0.0/toy.source	2020-08-18 12:19:35.000000000 +0200
+@@ -2,6 +2,6 @@
+ 
+ int main(int argc, char* argv[]){
+ 
+-    printf("I'm a toy, and proud of it.\n");
++    printf("I'm a toy, and proud of it.\n")
+     return 0;
+ }