Merge pull request #425 from rug-cit-hpc/develop

Critical bug fix for slurm prolog should be added to master for new release too.

group_vars/talos_cluster/vars.yml

@@ -35,23 +35,23 @@ additional_etc_hosts:
 vcompute_hostnames: "{{ stack_prefix }}-vcompute[01-03]"
 vcompute_sockets: 4
 vcompute_cores_per_socket: 1
-vcompute_real_memory: 7821
+vcompute_real_memory: 7820
 vcompute_max_cpus_per_node: "{{ vcompute_sockets * vcompute_cores_per_socket - 2 }}"
 vcompute_max_mem_per_node: "{{ vcompute_real_memory - vcompute_sockets * vcompute_cores_per_socket * 512 }}"
 vcompute_local_disk: 270000
 vcompute_features: 'tmp08'
 vcompute_ethernet_interfaces:
-  - 'eth0'
-  - 'eth1'
+  - 'vlan983'
+  - 'vlan985.isilon'
 ui_hostnames: "{{ slurm_cluster_name }}"
 ui_sockets: 4
 ui_cores_per_socket: 1
-ui_real_memory: 7821
+ui_real_memory: 7820
 ui_local_disk: 0
 ui_features: 'prm08,tmp08'
 ui_ethernet_interfaces:
-  - 'eth0'
-  - 'eth1'
+  - 'vlan983'
+  - 'vlan985.isilon'
 ssh_host_signer_ca_private_key: "{{ ssh_host_signer_ca_keypair_dir }}/umcg-hpc-development-ca"
 use_ldap: yes
 create_ldap: no

roles/slurm_management/files/slurm.prolog

@@ -3,9 +3,6 @@
 if [[ -z "${SLURM_JOB_ID}" ]]; then
     logger -s "FATAL: SLURM_JOB_ID is empty or unset in SLURM prolog."
     exit 1
-elif [[ -z "${SLURM_JOB_QOS}" ]]; then
-    logger -s "FATAL: SLURM_JOB_QOS is empty or unset in SLURM prolog."
-    exit 1
 #else
 #    logger -s "DEBUG: Found SLURM_JOB_ID ${SLURM_JOB_ID} and SLURM_JOB_QOS ${SLURM_JOB_QOS} in SLURM prolog."
 fi
@@ -18,23 +15,7 @@ LOCAL_SCRATCH_DIR='/local'
 # Check if local scratch dir is mountpoint and hence not a dir on the system disk.
 #
 if [[ $(stat -c '%d' "${LOCAL_SCRATCH_DIR}") -eq $(stat -c '%d' "${LOCAL_SCRATCH_DIR}/..") ]]; then
-    if [[ "${SLURM_JOB_QOS}" =~ ^ds.* ]]; then
-        #
-        # For the data staging QoS "ds", which executes jobs only on the UI,
-        # a dedicated tmp dir per job may be absent as not all UIs have a /local mount.
-        #
-        logger -s "WARN: local scratch disk (${LOCAL_SCRATCH_DIR}) is not mounted."
-    else
-        #
-        # Make sure we can create tmp dirs in /local on compute nodes.
-        # When this fails the job must not continue as SLURM will default to /tmp,
-        # which is not suitable for heavy random IO nor large data sets.
-        # Hammering /tmp may effectively result in the node going down.
-        # When the prolog fails the node will be set to state=DRAIN instead.
-        #
-        logger -s "FATAL: local scratch disk (${LOCAL_SCRATCH_DIR}) is not mounted."
-        exit 1
-    fi
+    logger -s "WARN: local scratch disk (${LOCAL_SCRATCH_DIR}) for Slurm jobs is not mounted/available."
 else
     #
     # Create dedicated tmp dir for this job.