DAOS-9883 downgrade libfabric to 1.12 and varia (#16)

Downgrade libfabric from 1.14 to 1.12 to address performance issue.
Adjust a few yaml parameter for 2.0
Set first core to 0.
Use single thread dfuse for io500.
Bump default aggregation threshold to meet DAOS 1.2 values.

Signed-off-by: Johann Lombardi <johann.lombardi@intel.com>

images/configs/daos_server.yml

@@ -5,13 +5,13 @@ provider: ofi+tcp;ofi_rxm
 disable_vfio: true
 crt_timeout: 300
 nr_hugepages: 4096
-control_log_file: /var/daos/daos_server.log
+control_log_file: /var/daos/server.log
 
 engines:
 -
    targets: 8
    nr_xs_helpers: 0
-   first_core: 1
+   first_core: 0
    bypass_health_chk: true
    fabric_iface: eth0
    fabric_iface_port: 31316
@@ -20,6 +20,8 @@ engines:
 
    env_vars:
       - FI_OFI_RXM_DEF_TCP_WAIT_OBJ=pollfd
+      - DTX_AGG_THD_CNT=16777216
+      - DTX_AGG_THD_AGE=700
 
    storage:
    -

images/scripts/install_daos.sh

@@ -223,6 +223,17 @@ install_daos() {
   fi
 }
 
+downgrade_libfabric() {
+  if [[ ${DAOS_VERSION%%.*} == "2" ]]; then
+    log "Downgrading libfabric to v1.12 - see https://daosio.atlassian.net/browse/DAOS-9883"
+    wget https://packages.daos.io/v1.2/CentOS7/packages/x86_64/libfabric-1.12.0-1.el7.x86_64.rpm
+    rpm -i --force ./libfabric-1.12.0-1.el7.x86_64.rpm
+    rpm --erase --nodeps  libfabric-1.14.0
+    echo "exclude=libfabric" >> /etc/yum.repos.d/daos.repo
+    rm -f ./libfabric-1.12.0-1.el7.x86_64.rpm
+  fi
+}
+
 install_additional_pkgs() {
   yum install -y clustershell curl git jq patch pdsh rsync wget
 }
@@ -234,8 +245,9 @@ main() {
   verify_version
   add_repo
   install_epel
-  install_daos
   install_additional_pkgs
+  install_daos
+  downgrade_libfabric
   printf "\n%s\n\n" "DONE! DAOS v${DAOS_VERSION} installed"
 }
 

terraform/examples/io500/run_io500-sc21.sh

@@ -144,7 +144,7 @@ mount_dfuse() {
       "mkdir -p '${IO500_RESULTS_DFUSE_DIR}'"
 
     clush --hostfile=hosts_clients --dsh \
-      "dfuse --pool='${DAOS_POOL_LABEL}' --container='${DAOS_CONT_LABEL}' --mountpoint='${IO500_RESULTS_DFUSE_DIR}'"
+      "dfuse -S --pool='${DAOS_POOL_LABEL}' --container='${DAOS_CONT_LABEL}' --mountpoint='${IO500_RESULTS_DFUSE_DIR}'"
 
     sleep 10
 
@@ -159,7 +159,7 @@ io500_prepare() {
   source /opt/intel/oneapi/setvars.sh
 
   export PATH=$PATH:${IO500_DIR}/bin
-  export LD_LIBRARY_PATH=/usr/local/mpifileutils/install/lib64/
+  export LD_LIBRARY_PATH=/usr/local/mpifileutils/install/lib64/:$LD_LIBRARY_PATH
 
   log "Prepare config file 'temp.ini' for IO500"