Merge branch 'master' into amd/pre-read

Features: dfuse

Required-githooks: true

.rpmignore

@@ -3,28 +3,22 @@
 # but should not be included in the current release
 #
 
-centos7/daos-client-tests-openmpi*.rpm
-centos7/daos-firmware*.rpm
-centos7/daos-mofed*.rpm
-centos7/daos-serialize*.rpm
-centos7/daos-server-tests-openmpi*.rpm
-centos7/daos-tests-internal*.rpm
-centos7/ucx*.rpm
-
-el8/daos-client-tests-openmpi*.rpm
+el8/daos-*tests*.rpm
 el8/daos-firmware*.rpm
 el8/daos-mofed*.rpm
 el8/daos-serialize*.rpm
-el8/daos-server-tests-openmpi*.rpm
-el8/daos-tests-internal*.rpm
 el8/ucx*.rpm
 
-leap15/daos-client-tests-openmpi*.rpm
+el9/daos-*tests*.rpm
+el9/daos-firmware*.rpm
+el9/daos-mofed*.rpm
+el9/daos-serialize*.rpm
+el9/ucx*.rpm
+
+leap15/daos-*tests*.rpm
 leap15/daos-firmware*.rpm
 leap15/daos-mofed*.rpm
 leap15/daos-serialize*.rpm
-leap15/daos-server-tests-openmpi*.rpm
-leap15/daos-tests-internal*.rpm
 leap15/openucx*.rpm
 leap15/ucx*.rpm
 leap15/*protobuf-c*.rpm

Jenkinsfile

@@ -74,6 +74,11 @@ void job_step_update(def value) {
 Map nlt_test() {
     // groovylint-disable-next-line NoJavaUtilDate
     Date startDate = new Date()
+    try {
+        unstash('nltr')
+    } catch (e) {
+        print 'Unstash failed, results from NLT stage will not be included'
+    }
     sh label: 'Fault injection testing using NLT',
        script: './ci/docker_nlt.sh --class-name el8.fault-injection fi'
     List filesList = []
@@ -1104,7 +1109,6 @@ pipeline {
                             sconsBuild(parallel_build: true,
                                        scons_args: 'PREFIX=/opt/daos TARGET_TYPE=release BUILD_TYPE=debug',
                                        build_deps: 'no'))
-                        unstash('nltr')
                         job_step_update(nlt_test())
                         recordCoverage(tools: [[parser: 'COBERTURA', pattern:'nltr.xml']],
                                        skipPublishingChecks: true,

SConstruct

@@ -363,7 +363,7 @@ MINIMAL_ENV = ('HOME', 'TERM', 'SSH_AUTH_SOCK', 'http_proxy', 'https_proxy', 'PK
 
 # Environment variables that are also kept when LD_PRELOAD is set.
 PRELOAD_ENV = ('LD_PRELOAD', 'D_LOG_FILE', 'DAOS_AGENT_DRPC_DIR', 'D_LOG_MASK', 'DD_MASK',
-               'DD_SUBSYS')
+               'DD_SUBSYS', 'D_IL_MAX_EQ')
 
 
 def scons():

ci/rpm/build_unsuccessful.sh

@@ -8,7 +8,7 @@ mydir="$(cd "$(dirname "${BASH_SOURCE[0]}")" >/dev/null 2>&1 && pwd)"
 ci_envs="$mydir/../parse_ci_envs.sh"
 if [ -e "${ci_envs}" ]; then
   # at some point we want to use: shellcheck source=ci/parse_ci_envs.sh
-  # shellcheck disable=SC1091
+  # shellcheck disable=SC1091,SC1090
   source "${ci_envs}"
 fi
 
@@ -25,6 +25,9 @@ if [ -d /var/cache/pbuilder/ ]; then
     exit 0
 fi
 
+rpm -q mock
+mock --debug-config
+
 mockroot="/var/lib/mock/$CHROOT_NAME"
 cat "$mockroot"/result/{root,build}.log 2>/dev/null || true
 

mkdocs.yml

@@ -1,5 +1,5 @@
 # Project Information
-site_name: DAOS v2.5 - master
+site_name: DAOS v2.5
 site_description: Distributed Asynchronous Object Storage
 site_author: DAOS Project
 

site_scons/components/__init__.py

@@ -58,7 +58,7 @@ def check(self, name):
             self.installed.append(name)
             return True
 
-        if not GetOption('help'):
+        if not GetOption('help') and not GetOption('silent'):
             print(f'Using build version of {name}')
         self.not_installed.append(name)
         return False
@@ -90,12 +90,14 @@ def check(reqs, name, built_str, installed_str=""):
 
 def ofi_config(config):
     """Check ofi version"""
-    print('Checking for libfabric > 1.11...', end=' ')
+    if not GetOption('silent'):
+        print('Checking for libfabric > 1.11...', end=' ')
     code = """#include <rdma/fabric.h>
 _Static_assert(FI_MAJOR_VERSION == 1 && FI_MINOR_VERSION >= 11,
                "libfabric must be >= 1.11");"""
     rc = config.TryCompile(code, ".c")
-    print('yes' if rc else 'no')
+    if not GetOption('silent'):
+        print('yes' if rc else 'no')
     return rc
 
 
@@ -132,6 +134,7 @@ def define_mercury(reqs):
                 libs=['fabric'],
                 config_cb=ofi_config,
                 headers=['rdma/fabric.h'],
+                pkgconfig='libfabric',
                 package='libfabric-devel' if inst(reqs, 'ofi') else None,
                 patch_rpath=['lib'],
                 build_env={'CFLAGS': "-fstack-usage"})
@@ -184,12 +187,6 @@ def define_mercury(reqs):
     else:
         mercury_build.append('-DMERCURY_ENABLE_DEBUG:BOOL=OFF')
 
-    mercury_build.extend(check(reqs,
-                               'ofi',
-                               ['-DOFI_INCLUDE_DIR:PATH=$OFI_PREFIX/include',
-                                '-DOFI_LIBRARY:FILEPATH=$OFI_PREFIX/lib/libfabric.so'],
-                               []))
-
     reqs.define('mercury',
                 retriever=GitRepoRetriever('https://github.com/mercury-hpc/mercury.git', True),
                 commands=[mercury_build,

site_scons/env_modules.py

@@ -36,14 +36,15 @@ class _env_module():  # pylint: disable=invalid-name
                 "openmpi": ['mpi/mlnx_openmpi-x86_64', 'mpi/openmpi3-x86_64',
                             'gnu-openmpi', 'mpi/openmpi-x86_64']}
 
-    def __init__(self):
+    def __init__(self, silent=False):
         """Load Modules for initializing environment variables"""
         # Leap 15's lmod-lua doesn't include the usual module path
         # in it's MODULEPATH, for some unknown reason
         os.environ["MODULEPATH"] = ":".join([os.path.join(os.sep, "usr", "share", "modules"),
                                              os.path.join(os.sep, "usr", "share", "modulefiles"),
                                              os.path.join(os.sep, "etc", "modulefiles")]
                                             + os.environ.get("MODULEPATH", "").split(":"))
+        self._silent = silent
         self._module_load = self._init_mpi_module()
 
     def _module_func(self, command, *arguments):  # pylint: disable=no-self-use
@@ -56,7 +57,8 @@ def _module_func(self, command, *arguments):  # pylint: disable=no-self-use
 
         # pylint: disable=consider-using-with
         try:
-            print(f"Going to run {cmd}")
+            if not self._silent:
+                print(' '.join(cmd))
             proc = Popen(cmd, stdout=PIPE, stderr=PIPE)
         except OSError as error:
             if error.errno == errno.ENOENT:
@@ -81,11 +83,11 @@ def _module_func(self, command, *arguments):  # pylint: disable=no-self-use
         # return _mlstatus, stderr.decode() # pylint: disable=undefined-variable
 
     def _init_mpi_module(self):
-        """init mpi module function"""
+        """Init mpi module function"""
         return self._mpi_module
 
     def _mpi_module(self, mpi):
-        """attempt to load the requested module"""
+        """Attempt to load the requested module"""
         load = []
         unload = []
 
@@ -109,16 +111,17 @@ def _mpi_module(self, mpi):
                 self._module_func('unload', to_unload)
 
         for to_load in load:
-            print(f"Trying to load {to_load}")
-            if self._module_func('is-avail', to_load)[0] and \
-               self._module_func('load', to_load)[0]:
-                print(f'Loaded {to_load}')
+            if not self._silent:
+                print(f"Trying to load {to_load}")
+            if self._module_func('is-avail', to_load)[0] and self._module_func('load', to_load)[0]:
+                if not self._silent:
+                    print(f'Loaded {to_load}')
                 return True
 
         return False
 
     def _mpi_module_old(self, mpi):
-        """attempt to load the requested module"""
+        """Attempt to load the requested module"""
         load = []
         for key, value in self._mpi_map.items():
             if key == mpi:
@@ -162,7 +165,7 @@ def load_mpi(self, mpi):
         return True
 
     def show_avail(self):
-        """list available modules"""
+        """List available modules"""
         try:
             status, output = self._module_func('avail')
             if not status:
@@ -172,12 +175,12 @@ def show_avail(self):
         return output
 
     def get_map(self, key):
-        """return the mpi map"""
+        """Return the mpi map"""
         return self._mpi_map[key]
 
 
-def load_mpi(mpi):
-    """global function to load MPI into os.environ"""
+def load_mpi(mpi, silent=False):
+    """Global function to load MPI into os.environ"""
     # On Ubuntu, MPI stacks use alternatives and need root to change their
     # pointer, so just verify that the desired MPI is loaded
     if distro.id() == "ubuntu":
@@ -201,19 +204,19 @@ def load_mpi(mpi):
         return False
 
     if _env_module.env_module_init is None:
-        _env_module.env_module_init = _env_module()
+        _env_module.env_module_init = _env_module(silent)
     return _env_module.env_module_init.load_mpi(mpi)
 
 
 def show_avail():
-    """global function to show the available modules"""
+    """Global function to show the available modules"""
     if _env_module.env_module_init is None:
         _env_module.env_module_init = _env_module()
     return _env_module.env_module_init.show_avail()
 
 
 def get_module_list(key):
-    """global function to show the modules that map to a key"""
+    """Global function to show the modules that map to a key"""
     if _env_module.env_module_init is None:
         _env_module.env_module_init = _env_module()
     return _env_module.env_module_init.get_map(key)

site_scons/prereq_tools/base.py

@@ -1125,6 +1125,11 @@ def _parse_config(self, env, opts):
 
         return
 
+    def _print(self, msg):
+        if GetOption('silent'):
+            return
+        print(msg)
+
     def has_missing_targets(self, env):
         """Check for expected build targets (e.g. libraries or headers)"""
         # pylint: disable=too-many-return-statements
@@ -1151,7 +1156,7 @@ def has_missing_targets(self, env):
             print('help set')
             return True
 
-        print(f"Checking targets for component '{self.name}'")
+        self._print(f"Checking targets for component '{self.name}'")
 
         config = env.Configure()
         config_cb = self.key_words.get("config_cb", None)
@@ -1244,7 +1249,6 @@ def configure(self):
 
     def set_environment(self, env, needed_libs):
         """Modify the specified construction environment to build with the external component"""
-
         if self.skip_arch:
             return
 

site_scons/site_tools/compiler_setup.py

@@ -39,8 +39,9 @@ def _base_setup(env):
     compiler = env['CC']
 
     build_type = env['BUILD_TYPE']
-    print(f'Setting up compile environment for {compiler}')
-    print(f"Build type is '{build_type}'")
+    if not GetOption('silent'):
+        print(f'Setting up compile environment for {compiler}')
+        print(f"Build type is '{build_type}'")
 
     prev_compiler = env.get('BSETUP', False)
     if prev_compiler:

site_scons/site_tools/daos_builder.py

@@ -224,6 +224,10 @@ def _configure_mpi(self):
     if GetOption('help'):
         return None
 
+    def _print(msg):
+        if not GetOption('silent'):
+            print(msg)
+
     env = self.Clone()
 
     env['CXX'] = None
@@ -233,13 +237,13 @@ def _configure_mpi(self):
         return env
 
     for mpi in ['openmpi', 'mpich']:
-        if not load_mpi(mpi):
+        if not load_mpi(mpi, GetOption('silent')):
             continue
         if _find_mpicc(env):
-            print(f'{mpi} is installed')
+            _print(f'{mpi} is installed')
             return env
-        print(f'No {mpi} installed and/or loaded')
-    print("No MPI installed")
+        _print(f'No {mpi} installed and/or loaded')
+    _print("No MPI installed")
     return None
 
 

src/SConscript

@@ -49,7 +49,8 @@ def read_and_save_version(env):
                       '@Template for @': ''}
 
         out = env.Substfile(tmpl_hdr_in, SUBST_DICT=subst_dict)
-        print(f'generated daos version header file: {out[0].abspath}')
+        if not GetOption('silent'):
+            print(f'generated daos version header file: {out[0].abspath}')
 
         return version
 

src/cart/README.env

@@ -80,6 +80,11 @@ This file lists the environment variables used in CaRT.
    by default, and can be specified using DD_SUBSYS, for example:
    "DD_SUBSYS=RPC,BULK,CORPC,GRP,LM,HG,PMIX,ST,IV" or also "DD_SUBSYS=all".
 
+ . D_MEMORY_TRACK
+   User can enable memory track for daos engine by D_MEMORY_TRACK=1. With the
+   environment, all of allocations inside DAOS (by D_ALLOC) will be tracked, and
+   total allocated bytes per xstream can be shown through metrics.
+
  . CRT_TIMEOUT
    Set it as integer in the range of (0, 3600] to set the global timeout value
    of all RPCs (second). Without setting it or set it as any other value will

src/cart/crt_hg.c

@@ -617,8 +617,9 @@ crt_provider_get_ctx_idx(bool primary, int provider)
 		}
 	}
 
-	D_ERROR("ctx_num %d, will exceed CRT_SRV_CONTEXT_NUM (%d) if create more context.\n",
-		prov_data->cpg_ctx_num, CRT_SRV_CONTEXT_NUM);
+	D_DEBUG(DB_ALL, "provider:%d allowed context limit = %d exceeded\n",
+		provider, CRT_SRV_CONTEXT_NUM);
+
 	return -1;
 }
 

src/client/api/event.c

@@ -484,8 +484,13 @@ daos_event_complete(struct daos_event *ev, int rc)
 	}
 
 	if (evx->evx_status == DAOS_EVS_READY || evx->evx_status == DAOS_EVS_COMPLETED ||
-	    evx->evx_status == DAOS_EVS_ABORTED)
+	    evx->evx_status == DAOS_EVS_ABORTED) {
+		if (evx->is_errno)
+			ev->ev_error = daos_der2errno(rc);
+		else
+			ev->ev_error = rc;
 		goto out;
+	}
 
 	D_ASSERT(evx->evx_status == DAOS_EVS_RUNNING);
 
@@ -830,7 +835,7 @@ daos_eq_destroy(daos_handle_t eqh, int flags)
 
 	eqx = daos_eq_lookup(eqh);
 	if (eqx == NULL) {
-		D_ERROR("eqh nonexist.\n");
+		D_ERROR("daos_eq_lookup() failed: "DF_RC"\n", DP_RC(-DER_NONEXIST));
 		return -DER_NONEXIST;
 	}
 
@@ -862,8 +867,7 @@ daos_eq_destroy(daos_handle_t eqh, int flags)
 	if (eqx->eqx_ctx != NULL) {
 		rc = crt_context_flush(eqx->eqx_ctx, 0);
 		if (rc != 0) {
-			D_ERROR("failed to flush client context: "DF_RC"\n",
-				DP_RC(rc));
+			D_ERROR("failed to flush client context: "DF_RC"\n", DP_RC(rc));
 			return rc;
 		}
 	}