This file describes changes in recent versions of SLURM. It primarily documents those changes that are of interest to users and admins. * Changes in SLURM 2.1.0-pre1 ============================= -- Slurmd notifies slurmctld of node boot time to better clean up after node reboots. -- Slurmd sends node registration information repeatedly until successful transmit. -- Change job_state in job structure to dedicate 8-bits to state flags. Added macros to get state information (IS_JOB_RUNNING(job_ptr), etc.) -- Added macros to get node state information (IS_NODE_DOWN(node_ptr), etc). -- Added support for Solaris. Patch from David Hoppner. -- Rename "slurm-aix-federation-.rpm" to just "slurm-aix-.rpm" (federation switch plugin may not be present). -- Eliminated the redundant squeue output format and sort options of "%o" and "%b". Use "%D" and "%S" formats respectively. Also eliminated "%X" and "%Y" and "%Z" formats. Use "%z" instead. -- Added mechanism for SPANK plugins to set environment variables for Prolog, Epilog, PrologSLurmctld and EpilogSlurmctld programs using the functions spank_get_job_env, spank_set_job_env, and spank_unset_job_env. See "man spank" for more information. -- Completed the work to begun in 2.0.0 to standardize on using '-Q' as the --quiet flag for all the commands. -- BLUEGENE - sinfo and sview now display correct cpu counts for partitions -- Cleaned up the cons_res plugin. It now uses a ptr to a part_record instead of having to do strcmp's to find the correct one. -- Pushed most all the plugin specific info in src/common/node_select.c into the respected plugin. -- BLUEGENE - closed some corner cases where a block could had been removed while a job was waiting for it to become ready because an underlying part of the block was put into an error state. -- Modify sbcast logic to prevent a user from moving files to nodes they have not been allocated (this would be possible in previous versions only by hacking the sbcast code). -- Add contribs/sjstat script (Perl tool to report job state information). Put into new RPM: sjstat. -- Add sched/wiki2 (Moab) JOBMODIFY command support for VARIABLELIST option to set supplemental environment variables for pending batch jobs. * Changes in SLURM 2.0.2 ======================== -- Fix, don't remove job details when a job is cancelled while pending. -- Do correct type for mktime so garbage isn't returned on 64bit systems for accounting archival. -- Better checking in sacctmgr to avoid infinite loops. -- fix minor memory leak in fake_slurm_step_layout_create() -- Fix node weight (scheduling priority) calculation for powered down nodes. Patch from Hongjia Cao, NUDT. -- Fix node suspend/resume rate calculations. Patch from Hongjia Cao, NUDT. -- Change calculations using ResumeRate and SuspendRate to provide higher resolution. -- Log the IP address for incoming messages having an invalid protocol version number. -- Fix for sacct to show jobs that start the same second as the sacct command is issued. -- BLUEGENE - Fix for -n option to work on correct cpu counts for each midplane instead of treating -n as a c-node count. -- salloc now sets SLURM_NTASKS_PER_NODE if --ntasks-per-node option is set. -- Fix select/linear to properly set a job's count of allocated processors (all processors on the allocated nodes). -- Fix select/cons_res to allocate proper CPU count when --ntasks-per-node option is used. -- Insure that no node is allocated to a job for which the CPU count is less than --ntasks-per-node * --cpus-per-task. -- Correct AllocProcs reported by "scontrol show node" when ThreadsPerCore is greater than 1 and select/cons_res is used. -- Fix scontrol show config for accounting information when values are not set in the slurm.conf. -- Added a set of SBATCH_CPU_BIND* and SBATCH_MEM_BIND* env variables to keep jobsteps launched from within a batch script from inheriting the CPU and memory affinity that was applied to the batch script. -- Ignore the extra processors on a node above configured size if either sched/gang or select/cons_res is configured. -- Fixed a race condition when writing labelled output with a file per task or per node, which potentially closed a file before all data was written. * Changes in SLURM 2.0.1 ======================== -- Fix, truncate time of start and end for job steps in sacct. -- Initialize all messages to slurmdbd. Previously uninitialized string could cause slurmctld to fail with invalid memory reference. -- BLUEGENE - Fix, for when trying to finish a torus on a block already visited. Even though this may be possible electrically this isn't valid in the under lying infrastructure. -- Fix, in mysql plugins change mediumints to int to support full 32bit numbers. -- Add sinfo node state filtering support for NO_RESPOND, POWER_SAVE, FAIL, MAINT, DRAINED and DRAINING states. The state filter of DRAIN still maps to any node in either DRAINED or DRAINING state. -- Fix reservation logic when job requests specific nodes that are already in some reservation the job can not use. -- Fix recomputation of a job's end time when allocated nodes which are being powered up. The end time would be set in the past if the job's time limit was INFINITE, resulting in it being prematurely terminated. -- Permit regular user to change the time limit of his pending jobs up to the partition's limit. -- Fix "-Q" (quiet) option for salloc and sbatch which was previously ignored. -- BLUEGENE - fix for finding odd shaped blocks in dynamic mode. -- Fix logic supporting SuspendRate and ResumeRate configuration parameters. Previous logic was changing state of one too many nodes per minute. -- Save new reservation state file on shutdown (even if no changes). -- Fix, when partitions are deleted the sched and select plugins are notified. -- Fix for slurmdbd to create wckeyid's when they don't exist -- Fix linking problem that prevented checkpoint/aix from working. * Changes in SLURM 2.0.0 ======================== -- Fix for bluegene systems to be able to create 32 node blocks with only 16 psets defined in dynamic layout mode. -- Improve srun_cr handling of child srun forking. Patch from Hongjia Cao, NUDT. -- Configuration parameter ResumeDelay replaced by SuspendTimeout and ResumeTimeout. -- BLUEGENE - sview/sinfo now displays correct cnode numbers for drained nodes or blocks in error state. -- Fix some batch job launch bugs when powering up suspended nodes. -- Added option '-T' for sacct to truncate time of start and end and set default of --starttime to Midnight of current day. * Changes in SLURM 2.0.0-rc2 ============================ -- Change fanout logic to start on calling node instead of first node in message nodelist. -- Fix bug so that smap builds properly on Sun Constellation system. -- Filter white-space out from node feature specification. -- Fixed issue with duration not being honored when updating start time in reservations. -- Fix bug in sched/wiki and sched/wiki2 plugins for reporting job resource allocation properly when node names are configured out of sort order with more than one numeric suffix (e.g. "tux10-1" is configured after "tux5-1"). -- Avoid re-use of job_id (if specified at submit time) when the existing job is in completing state (possible race condition with Moab). -- Added SLURM_DISTRIBUTION to env for salloc. -- Add support for "scontrol takeover" command for backup controller to assume control immediately. Patch from Matthieu Hautreux, CEA. -- If srun is unable to communicate with the slurmd tasks are now marked as failed with the controller. -- Fixed issues with requeued jobs not being accounted for correctly in the accounting. -- Clear node's POWER_SAVE flag if configuration changes to one lacking a ResumeProgram. -- Extend a job's time limit as appropriate due to delays powering up nodes. -- If sbatch is used to launch a job step within an existing allocation (as used by LSF) and the required node is powered down, print the message "Job step creation temporarily disabled, retrying", sleep, and retry. -- Configuration parameter ResumeDelay added to control how much time must after a node has been suspended before resume it (e.g. powering it back up). -- Fix CPU binding for batch program. Patch from Matthieu Hautreux, CEA. -- Fix for front end systems non-responding nodes now show up correctly in sinfo. * Changes in SLURM 2.0.0-rc1 ============================ -- Fix bug in preservation of advanced reservations when slurmctld restarts. -- Updated perlapi to match correctly with slurm.h structures -- Do not install the srun command on BlueGene systems (mpirun must be used to launch tasks). -- Corrections to scheduling logic for topology/tree in configurations where nodes are configured in multiple leaf switches. -- Patch from Matthieu Hautreux for backup mysql deamon support. -- Changed DbdBackup to DbdBackupHost for slurmdbd.conf file -- Add support for spank_strerror() function and improve error handling in general for SPANK plugins. -- Added configuration parameter SrunIOTimeout to optionally ping srun's tasks for better fault tolerance (e.g. killed and restarteed SLURM daemons on compute node). -- Add slurmctld and slurmd binding to appropriate communications address based upon NodeAddr, ControllerAddr and BackupAddr configuration parameters. Based upon patch from Matthieu Hautreux, CEA. NOTE: Fails when SlurmDBD is configured with some configurations. NOTE: You must define BIND_SPECIFIC_ADDR to enable this option. -- Avoid using powered down nodes when scheduling work if possible. Fix possible invalid memory reference in power save logic. * Changes in SLURM 1.4.0-pre13 ============================== -- Added new partition option AllocNodes which controls the hosts from which jobs can be submitted to this partition. From Matthieu Hautreux, CEA. -- Better support the --contiguous option for job allocations. -- Add new scontrol option: show topology (reports contents of topology.conf file via RPC if topology/tree plugin is configured). -- Add advanced reservation display to smap command. -- Replaced remaining references to SLURM_JOBID with SLURM_JOB_ID - except when needed for backwards compatibility. -- Fix logic to properly excise a DOWN node from the allocation of a job with the --no-kill option. -- The MySQL and PgSQL plugins for accounting storage and job completion are now only built if the underlying database libraries exists (previously the plugins were built to produce a fatal error when used). -- BLUEGENE - scontrol show config will now display bluegene.conf information. * Changes in SLURM 1.4.0-pre12 ============================== -- Added support for hard time limit by associations with added configuration option PriorityUsageResetPeriod. This specifies the interval at which to clear the record of time used. This is currently only available with the priority/multifactor plugin. -- Added SLURM_SUBMIT_DIR to sbatch's output environment variables. -- Backup slurmdbd support implemented. -- Update to checkpoint/xlch logic from Hongjia Cao, NUDT. -- Added configuration parameter AccountingStorageBackupHost. * Changes in SLURM 1.4.0-pre11 ============================== -- Fix slurm.spec file for RPM build. * Changes in SLURM 1.4.0-pre10 ============================== -- Critical bug fix in task/affinity when the CoresPerSocket is greater than the ThreadsPerCore (invalid memory reference). -- Add DebugFlag parameter of "Wiki" to log sched/wiki and wiki2 communications in greater detail. -- Add "-d " as an option to the slurmd daemon to specifying a non-stardard slurmstepd file, used for testing purposes. -- Minor cleanup to crypto/munge plugin. - Restrict uid allowed to decode job credentials in crypto/munge - Get slurm user id early in crypto/munge - Remove buggy error code handling in crypto/munge -- Added sprio command - works only with the priority/multifactor plugin -- Add real topology plugin infrastructure (it was initially added directly into slurmctld code). To specify topology information, set TopologyType=topology/tree and add configuration information to a new file called topology.conf. See "man topology.conf" or topology.html web page for details. -- Set "/proc/self/oom_adj" for slurmd and slurmstepd daemons based upon the values of SLURMD_OOM_ADJ and SLURMSTEPD_OOM_ADJ environment variables. This can be used to prevent daemons being killed when a node's memory is exhausted. Based upon patch by Hongjia Cao, NUDT. -- Fix several bugs in task/affinity: cpuset logic was broken and --cpus-per-task option not properly handled. -- Ensure slurmctld adopts SlurmUser GID as well as UID on startup. * Changes in SLURM 1.4.0-pre9 ============================= -- OpenMPI users only: Add srun logic to automatically recreate and re-launch a job step if the step fails with a reserved port conflict. -- Added TopologyPlugin configuration parameter. -- Added switch topology data structure to slurmctld (for use by select plugin) add load it based upon new slurm.conf parameters: SwitchName, Nodes, Switches and LinkSpeed. -- Modify select/linear and select/cons_res plugins to optimize resource allocation with respect to network topology. -- Added support for new configuration parameter EpilogSlurmctld (executed by slurmctld daemon). -- Added checkpoint/blcr plugin, SLURM now support job checkpoint/restart using BLCR. Patch from Hongjia Cao, NUDT, China. -- Made a variety of new environment variables available to PrologSlurmctld and EpilogSlurmctld. See the "Prolog and Epilog Scripts" section of the slurm.conf man page for details. -- NOTE: Cold-start (without preserving state) required for upgrade from version 1.4.0-pre8. * Changes in SLURM 1.4.0-pre8 ============================= -- In order to create a new partition using the scontrol command, use the "create" option rather than "update" (which will only operate upon partitions that already exist). -- Added environment variable SLURM_RESTART_COUNT to batch jobs to indicated the count of job restarts made. -- Added sacctmgr command "show config". -- Added the scancel option --nodelist to cancel any jobs running on a given list of nodes. -- Add partition-specific DefaultTime (default time limit for jobs, if not specified use MaxTime for the partition. Patch from Par Andersson, National Supercomputer Centre, Sweden. -- Add support for the scontrol command to be able change the Weight associated with nodes. Patch from Krishnakumar Ravi[KK] (HP). -- Add DebugFlag configuration option of "CPU_Bind" for detailed CPU binding information to be logged. -- Fix some significant bugs in task binding logic (possible infinite loops and memory corruption). -- Add new node state flag of NODE_STATE_MAINT indicating the node is in a reservation of type MAINT. -- Modified task/affinity plugin to automatically bind tasks to sockets, cores, or threads as appropriated based upon resource allocation and task count. User can override with srun's --cpu_bind option. -- Fix bug in backfill logic for select/cons_res plugin, resulted in error "cons_res:_rm_job_from_res: node_state mis-count". -- Add logic go bind a batch job to the resources allocated to that job. -- Add configuration parameter MpiParams for (future) OpenMPI port management. Add resv_port_cnt and resv_ports fields to the job step data structures. Add environment variable SLURM_STEP_RESV_PORTS to show what ports are reserved for a job step. -- Add support for SchedulerParameters=interval= to control the time interval between executions of the backfill scheduler logic. -- Preserve record of last job ID in use even when doing a cold-start unless there is no job state file or there is a change in its format (which only happens when there is a change in SLURM's major or minor version number: v1.3 -> v1.4). -- Added new configuration parameter KillOnBadExit to kill a job step as soon as any task of a job step exits with a non-zero exit code. Patch based on work from Eric Lin, Bull. -- Add spank plugin calls for use by salloc and sbatch command, see "man spank" for details. -- NOTE: Cold-start (without preserving state) required for upgrade from version 1.4.0-pre7. * Changes in SLURM 1.4.0-pre7 ============================= -- Bug fix for preemption with select/cons_res when there are no idle nodes. -- Bug fix for use of srun options --exclusive and --cpus-per-task together for job step resource allocation (tracking of cpus in use was bad). -- Added the srun option --preserve-env to pass the current values of environment variables SLURM_NNODES and SLURM_NPROCS through to the executable, rather than computing them from commandline parameters. -- For select/cons_res or sched/gang only: Validate a job's resource allocation socket and core count on each allocated node. If the node's configuration has been changed, then abort the job. -- For select/cons_res or sched/gang only: Disable updating a node's processor count if FastSchedule=0. Administrators must set a valid processor count although the memory and disk space configuration can be loaded from the compute node when it starts. -- Add configure option "--disable-iso8601" to disable SLURM use of ISO 8601 time format at the time of SLURM build. Default output for all commands is now ISO 8601 (yyyy-mm-ddThh:mm:ss). -- Add support for scontrol to explicity power a node up or down using the configured SuspendProg and ResumeProg programs. -- Fix book select/cons_res logic for tracking the number of allocated CPUs on a node when a partition's Shared value is YES or FORCE. -- Added configure options "--enable-cray-xt" and "--with-apbasil=PATH" for eventual support of Cray-XT systems. * Changes in SLURM 1.4.0-pre6 ============================= -- Fix job preemption when sched/gang and select/linear are configured with non-sharing partitions. -- In select/cons_res insure that required nodes have available resources. * Changes in SLURM 1.4.0-pre5 ============================= -- Correction in setting of SLURM_CPU_BIND environment variable. -- Rebuild slurmctld's job select_jobinfo->node_bitmap on restart/reconfigure of the daemon rather than restoring the bitmap since the nodes in a system can change (be added or removed). -- Add configuration option "--with-cpusetdir=PATH" for non-standard locations. -- Get new multi-core data structures working on BlueGene systems. -- Modify PMI_Get_clique_ranks() to return an array of integers rather than a char * to satisfy PMI standard. Correct logic in PMI_Get_clique_size() for when srun --overcommit option is used. -- Fix bug in select/cons_res, allocated a job all of the processors on a node when the --exclusive option is specified as a job submit option. -- Add NUMA cpu_bind support to the task affinity plugin. Binds tasks to a set of CPUs that belong NUMA locality domain with the appropriate --cpu-bind option (ldoms, rank_ldom, map_ldom, and mask_ldom), see "man srun" for more information. * Changes in SLURM 1.4.0-pre4 ============================= -- For task/affinity, force jobs to use a particular task binding by setting the TaskPluginParam configuration parameter rather than slurmd's SLURM_ENFORCED_CPU_BIND environment variable. -- Enable full preemption of jobs by partition with select/cons_res (cons_res_preempt.patch from Chris Holmes, HP). -- Add configuration parameter DebugFlags to provide detailed logging for specific subsystems (steps and triggers so far). -- srun's --no-kill option is passed to slurmctld so that a job step is killed even if the node where srun executes goes down (unless the --no-kill option is used, previous termination logic would fail if srun was not responding). -- Transfer a job step's core bitmap from the slurmctld to the slurmd within the job step credential. -- Add cpu_bind, cpu_bind_type, mem_bind and mem_bind_type to job allocation request and job_details structure in slurmctld. Add support to --cpu_bind and --mem_bind options from salloc and sbatch commands. * Changes in SLURM 1.4.0-pre3 ============================= -- Internal changes: CPUs per node changed from 32-bit to 16-bit size. Node count fields changed from 16-bit to 32-bit size in some structures. -- Remove select plugin functions select_p_get_extra_jobinfo(), select_p_step_begin() and select_p_step_fini(). -- Remove the following slurmctld job structure fields: num_cpu_groups, cpus_per_node, cpu_count_reps, alloc_lps_cnt, alloc_lps, and used_lps. Use equivalent fields in new "select_job" structure, which is filled in by the select plugins. -- Modify mem_per_task in job step request from 16-bit to 32-bit size. Use new "select_job" structure for the job step's memory management. -- Add core_bitmap_job to slurmctld's job step structure to identify which specific cores are allocated to the step. -- Add new configuration option OverTimeLimit to permit jobs to exceed their (soft) time limit by a configurable amount. Backfill scheduling will be based upon the soft time limit. -- Remove select_g_get_job_cores(). That data is now within the slurmctld's job structure. * Changes in SLURM 1.4.0-pre2 ============================= -- Remove srun's --ctrl-comm-ifhn-addr option (for PMI/MPICH2). It is no longer needed. -- Modify power save mode so that nodes can be powered off when idle. See https://computing.llnl.gov/linux/slurm/power_save.html or "man slurm.conf" (SuspendProgram and related parameters) for more information. -- Added configuration parameter PrologSlurmctld, which can be used to boot nodes into a particular state for each job. See "man slurm.conf" for details. -- Add configuration parameter CompleteTime to control how long to wait for a job's completion before allocating already released resources to pending jobs. This can be used to reduce fragmentation of resources. See "man slurm.conf" for details. -- Make default CryptoType=crypto/munge. OpenSSL is now completely optional. -- Make default AuthType=auth/munge rather than auth/none. -- Change output format of "sinfo -R" from "%35R %N" to "%50R %N". * Changes in SLURM 1.4.0-pre1 ============================= -- Save/restore a job's task_distribution option on slurmctld retart. NOTE: SLURM must be cold-started on converstion from version 1.3.x. -- Remove task_mem from job step credential (only job_mem is used now). -- Remove --task-mem and --job-mem options from salloc, sbatch and srun (use --mem-per-cpu or --mem instead). -- Remove DefMemPerTask from slurm.conf (use DefMemPerCPU or DefMemPerNode instead). -- Modify slurm_step_launch API call. Move launch host from function argument to element in the data structure slurm_step_launch_params_t, which is used as a function argument. -- Add state_reason_string to job state with optional details about why a job is pending. -- Make "scontrol show node" output match scontrol input for some fields ("Cores" changed to "CoresPerSocket", etc.). -- Add support for a new node state "FUTURE" in slurm.conf. These node records are created in SLURM tables for future use without a reboot of the SLURM daemons, but are not reported by any SLURM commands or APIs. * Changes in SLURM 1.3.17 ========================= -- Fix bug in configure script that can clear user specified LIBS. * Changes in SLURM 1.3.16 ========================= -- Fix memory leak in forward logic of tree message passing. -- Fix job exit code recorded for srun job allocation. -- Bluegene - Bug fix for too many parameters being passed to a debug statement -- Bluegene - Bug fix for systems running more than 8 in the X dim running Dynamic mode. * Changes in SLURM 1.3.15 ========================= -- Fix bug in squeue command with sort on job name ("-S j" option) for jobs that lack a name. Previously generated an invalid memory reference. -- Permit the TaskProlog to write to the job's standard output by writing a line containing the prefix "print " to it's standard output. -- Fix for making the slurmdbd agent thread start up correctly when stopped and then started again. -- Add squeue option to report jobs by account (-U or --account). Patch from Par Andersson, National Supercomputer Centre, Sweden. -- Add -DNUMA_VERSION1_COMPATIBILITY to Makefile CFLAGS for proper behavior when building with NUMA version 2 APIs. -- BLUEGENE - slurm works on a BGP system. -- BLUEGENE - slurm handles HTC blocks -- BLUEGENE - Added option DenyPassthrough in the bluegene.conf. Can be set to any combination of X,Y,Z to not allow passthroughs when running in dynamic layout mode. -- Fix bug in logic to remove a job's dependency, could result in abort. -- Add new error message to sched/wiki and sched/wiki2 (Maui and Moab) for STARTJOB request: "TASKLIST includes non-responsive nodes". -- Fix bug in select/linear when used with sched/gang that can result in a job's required or excluded node specification being ignored. -- Add logic to handle message connect timeouts (timed-out.patch from Chuck Clouston, Bull). -- BLUEGENE - CFLAGS=-m64 is no longer required in configure -- Update python-hostlist code from Kent Engström (NSC) to v1.5 - Add hostgrep utility to search for lines matching a hostlist. - Make each "-" on the command line count as one hostlist argument. If multiple hostslists are given on stdin they are combined to a union hostlist before being used in the way requested by the options. -- When using -j option in sacct no user restriction will applied unless specified with the -u option. -- For sched/wiki and sched/wiki2, change logging of wiki message traffic from debug() to debug2(). Only seen if SlurmctldDebug is configured to 6 or higher. -- Significant speed up for association based reports in sreport -- BLUEGENE - fix for checking if job can run with downed nodes. Previously sbatch etc would tell you node configuration not available now jobs are accepted but held until nodes are back up. -- Fix in accounting so if any nodes are removed from the system when they were previously down will be recorded correctly. -- For sched/wiki2 (Moab), add flag to note if job is restartable and prevent deadlock of job requeue fails. -- Modify squeue to return non-zero exit code on failure. Patch from Par Andersson (NSC). -- Correct logic in select/cons_res to allocate a job the maximum node count from a range rather than minimum (e.g. "sbatch -N1-4 my.sh"). -- In accounting_storage/filetxt and accounting_storage/pgsql fix possible invalid memory reference when a job lacks a name. -- Give srun command an exit code of 1 if the prolog fails. -- BLUEGENE - allows for checking nodecard states in the system instead of midplane state so as to not down an entire midplane if you don't have to. -- BLUEGENE - fix creation of MESH blocks -- BLUEGENE - on job cancellation we call jm_cancel_job and then wait until the system cleans up the job. Before we would send a SIGKILL right at the beginning. -- BLUEGENE - if a user specifies a node count that can not be met the job will be refused instead of before the plugin would search for the next larger size that could be created. This prevents users asking for things that can't be created, and then getting something back they might not be expecting. * Changes in SLURM 1.3.14 ========================= -- SECURITY BUG: Fix in sbcast logic that permits users to write files based upon supplimental groups of the slurmd daemon. Similar logic for event triggers if slurmctld is run as user root (not typical). * Changes in SLURM 1.3.13 ========================= -- Added ability for slurmdbd to archive and purge step and/or job records. -- Added DefaultQOS as an option to slurmdbd.conf for when clusters are added the default will be set to this if none is given in the sacctmgr line. -- Added configure option --enable-sun-const for Sun Constellation system with 3D torus interconnect. Supports proper smap and sview displays for 3-D topology. Node names are automatically put into Hilbert curve order given a one-line nodelist definition in slurm.conf (e.g. NodeNames=sun[000x533]). -- Fixed bug in parsing time for sacct and sreport to pick the correct year if none is specified. -- Provide better scheduling with overlapping partitions (when a job can not be scheduled due to insufficient resources, reserve specific the nodes associated with that partition rather than blocking all partitions with any overlapping nodes). -- Correct logic to log in a job's stderr that it was "CANCELLED DUE TO NODE FAILURE" rather than just "CANCELLED". -- Fix to crypto/openssl plugin that could result in job launch requests being spoofed through the use of an improperly formed credential. This bug could permit a user to launch tasks on compute nodes not allocated for their use, but will NOT permit them to run tasks as another user. For more information see http://www.ocert.org/advisories/ocert-2008-016.html * Changes in SLURM 1.3.12 ========================= -- Added support for Workload Characteristic Key (WCKey) in accounting. The WCkey is something that can be used in accounting to group associations together across clusters or within clusters that are not related. Use the --wckey option in srun, sbatch or salloc or set the SLURM_WCKEY env var to have this set. Use sreport with the wckey option to view reports. THIS CHANGES THE RPC LEVEL IN THE SLURMDBD. YOU MUST UPGRADE YOUR SLURMDBD BEFORE YOU UPGRADE THE REST OF YOUR CLUSTERS. THE NEW SLURMDBD WILL TALK TO OLDER VERSIONS OF SLURM FINE. -- Added configuration parameter BatchStartTimeout to control how long to allow for a batch job prolog and environment loading (for Moab) to run. Previously if job startup took too long, a batch job could be cancelled before fully starting with a SlurmctldLog message of "Master node lost JobId=#, killing it". See "man slurm.conf" for details. -- For a job step, add support for srun's --nodelist and --exclusive options to be used together. -- On slurmstepd failure, set node state to DRAIN rather than DOWN. -- Fix bug in select/cons_res that would incorrectly satify a tasks's --cpus-per-task specification by allocating the task CPUs on more than one node. -- Add support for hostlist expressions containing up to two numeric expressions (e.g. "rack[0-15]_blade[0-41]"). -- Fix bug in slurmd message forwarding which left file open in the case of some communication failures. -- Correction to sinfo node state information on BlueGene systems. DRAIN state was replaced with ALLOC or IDLE under some situations. -- For sched/wiki2 (Moab), strip quotes embedded within job names from the name reported. -- Fix bug in jobcomp/script that could cause the slurmctld daemon to exit upon reconfiguration ("scontrol reconfig" or SIGHUP). -- Fix to sinfo, don't print a node's memory size or tmp_disk space with suffix of "K" or "M" (thousands or millions of megabytes). -- Improve efficiency of scheduling jobs into partitions which do not overlap. -- Fixed sreport user top report to only display the limit specified instead of all users. * Changes in SLURM 1.3.11 ========================= -- Bluegene/P support added (minimally tested, but builds correctly). -- Fix infinite loop when using accounting_storage/mysql plugin either from the slurmctld or slurmdbd daemon. -- Added more thread safety for assoc_mgr in the controller. -- For sched/wiki2 (Moab), permit clearing of a job's dependencies with the JOB_MODIFY option "DEPEND=0". -- Do not set a running or pending job's EndTime when changing it's time limit. -- Fix bug in use of "include" parameter within the plugstack.conf file. -- Fix bug in the parsing of negative numeric values in configuration files. -- Propagate --cpus-per-task parameter from salloc or sbatch input line to the SLURM_CPUS_PER_TASK environment variable in the spawned shell for srun to use. -- Add support for srun --cpus-per-task=0. This can be used to spawn tasks without allocating resouces for the job step from the job's allocation when running multiple job steps with the --exclusive option. -- Remove registration messages from saved messages when bringing down cluster. Without causes deadlock if wrong cluster name is given. -- Correction to build for srun debugger (export symbols). -- sacct will now display more properly allocations made with salloc with only one step. -- Altered sacctmgr, sreport to look at complete option before applying. Before we would only look at the first determined significant characters. -- BLUGENE - in overlap mode marking a block to error state will now end jobs on overlapping blocks and free them. -- Give a batch job 20 minutes to start before considering it missing and killing it (long delay could result from slurmd being paged out). Changed the log message from "Master node lost JobId=%u, killing it" to "Batch JobId=%u missing from master node, killing it". -- Avoid "Invalid node id" error when a job step within an existing job allocation specifies a node count which is less than the node count allocated in order to satisfy the task count specification (e.g. "srun -n16 -N1 hostname" on allocation of 16 one-CPU nodes). -- For sched/wiki2 (Moab) disable changing a job's name after it has begun execution. * Changes in SLURM 1.3.10 ========================= -- Fix several bugs in the hostlist functions: - Fix hostset_insert_range() to do proper accounting of hl->nhosts (count). - Avoid assertion failure when callinsg hostset_create(NULL). - Fix return type of hostlist and hostset string functions from size_t to ssize_t. - Add check for NULL return from hostlist_create(). - Rewrite of hostrange_hn_within(), avoids reporting "tst0" in the hostlist "tst". -- Modify squeue to accept "--nodes=" rather than "--node=" and report all jobs with any allocated nodes from set of nodes specified. From Par Anderson, National Supercomputer Centre, Sweden. -- Fix bug preventing use of TotalView debugger with TaskProlog configured or or srun's --task-prolog option. -- Improve reliability of batch job requeue logic in the event that the slurmd daemon is temporarily non-responsive (for longer than the configured MessageTimeout value but less than the SlurmdTimeout value). -- In sched/wiki2 (Moab) report a job's MAXNODES (maximum number of permitted nodes). -- Fixed SLURM_TASKS_PER_NODE to live up more to it's name on an allocation. Will now contain the number of tasks per node instead of the number of CPUs per node. This is only for a resource allocation. Job steps already have the environment variable set correctly. -- Configuration parameter PropagateResourceLimits has new option of "NONE". -- User's --propagate options take precidence over PropagateResourceLimits configuration parameter in both srun and sbatch commands. -- When Moab is in use (salloc or sbatch is executed with the --get-user-env option to be more specific), load the user's default resource limits rather than propagating the Moab daemon's limits. -- Fix bug in slurmctld restart logic for recovery of batch jobs that are initiated as a job step rather than an independent job (used for LSF). -- Fix bug that can cause slurmctld restart to fail, bug introduced in SLURM version 1.3.9. From Eygene Ryabinkin, Kurchatov Institute, Russia. -- Permit slurmd configuration parameters to be set to new values from previously unset values. * Changes in SLURM 1.3.9 ======================== -- Fix jobs being cancelled by ctrl-C to have correct cancelled state in accounting. -- Slurmdbd will only cache user data, made for faster start up -- Improved support for job steps in FRONT_END systems -- Added support to dump and load association information in the controller on start up if slurmdbd is unresponsive -- BLUEGENE - Added support for sched/backfill plugin -- sched/backfill modified to initiate multiple jobs per cycle. -- Increase buffer size in srun to hold task list expressions. Critical for jobs with 16k tasks or more. -- Added support for eligible jobs and downed nodes to be sent to accounting from the controller the first time accounting is turned on. -- Correct srun logic to support --tasks-per-node option without task count. -- Logic in place to handle multiple versions of RPCs within the slurmdbd. THE SLURMDBD MUST BE UPGRADED TO THIS VERSION BEFORE UPGRADING THE SLURMCTLD OR THEY WILL NOT TALK. Older versions of the slurmctld will continue to talk to the new slurmdbd. -- Add support for new job dependency type: singleton. Only one job from a given user with a given name will execute with this dependency type. From Matthieu Hautreux, CEA. -- Updated contribs/python/hostlist to version 1.3: See "CHANGES" file in that directory for details. From Kent Engström, NSC. -- Add SLURM_JOB_NAME environment variable for jobs submitted using sbatch. In order to prevent the job steps from all having the same name as the batch job that spawned them, the SLURM_JOB_NAME environment variable is ignored when setting the name of a job step from within an existing resource allocation. -- For use with sched/wiki2 (Moab only), set salloc's default shell based upon the user who the job runs as rather than the user submitting the job (user root). -- Fix to sched/backfill when job specifies no time limit and the partition time limit is INFINITE. -- Validate a job's constraints (node features) at job submit or modification time. Major re-write of resource allocation logic to support more complex job feature requests. -- For sched/backfill, correct logic to support job constraint specification (e.g. node features). -- Correct power save logic to avoid trying to wake DOWN node. From Matthieu Hautreux, CEA. -- Cancel a job step when one of it's nodes goes DOWN based upon the job step's --no-kill option, by default the step is killed (previously the job step remained running even without the --no-kill option). -- Fix bug in logic to remove whitespace from plugstack.conf. -- Add new configuration parameter SallocDefaultCommand to control what shell that salloc launches by default. -- When enforcing PrivateData configuration parameter, failures return "Access/permission denied" rather than "Invalid user id". -- From sbatch and srun, if the --dependency option is specified then set the environment variable SLURM_JOB_DEPENDENCY to the same value. -- In plugin jobcomp/filetxt, use ISO8601 formats for time by default (e.g. YYYY-MM-DDTHH:MM:SS rather than MM/DD-HH:MM:SS). This restores the default behavior from Slurm version 1.2. Change the value of USE_ISO8601 in src/plusings/jobcomp/filetxt/jobcomp_filetxt.c to revert the behavior. -- Add support for configuration option of ReturnToService=2, which will return a DOWN to use if the node was previous set DOWN for any reason. -- Removed Gold accounting plugin. This plugin was to be used for accounting but has seen not been maintained and is no longer needed. If using this please contact slurm-dev@llnl.gov. -- When not enforcing associations and running accounting if a user submits a job to an account that does not have an association on the cluster the account will be changed to the default account to help avoid trash in the accounting system. If the users default account does not have an association on the cluster the requested account will be used. -- Add configuration parameter "--have-front-end" to define HAVE_FRONT_END in config.h and run slurmd only on a front end (suitable only for SLURM development and testing). * Changes in SLURM 1.3.8 ======================== -- Added PrivateData flags for Users, Usage, and Accounts to Accounting. If using slurmdbd, set in the slurmdbd.conf file. Otherwise set in the slurm.conf file. See "man slurm.conf" or "man slurmdbd.conf" for details. -- Reduce frequency of resending job kill RPCs. Helpful in the event of network problems or down nodes. -- Fix memory leak caused under heavy load when running with select/cons_res plus sched/backfill. -- For salloc, if no local command is specified, execute the user's default shell. -- BLUEGENE - patch to make sure when starting a job blocks required to be freed are checked to make sure no job is running on them. If one is found we will requeue the new job. No job will be lost. -- BLUEGENE - Set MPI environment variables from salloc. -- BLUEGENE - Fix threading issue for overlap mode -- Reject batch scripts containing DOS linebreaks. -- BLUEGENE - Added wait for block boot to salloc * Changes in SLURM 1.3.7 ======================== -- Add jobid/stepid to MESSAGE_TASK_EXIT to address race condition when a job step is cancelled, another is started immediately (before the first one completely terminates) and ports are reused. NOTE: This change requires that SLURM be updated on all nodes of the cluster at the same time. There will be no impact upon currently running jobs (they will ignore the jobid/stepid at the end of the message). -- Added Python module to process hostslists as used by SLURM. See contribs/python/hostlist. Supplied by Kent Engstrom, National Supercomputer Centre, Sweden. -- Report task termination due to signal (restored functionality present in slurm v1.2). -- Remove sbatch test for script size being no larger than 64k bytes. The current limit is 4GB. -- Disable FastSchedule=0 use with SchedulerType=sched/gang. Node configuration must be specified in slurm.conf for gang scheduling now. -- For sched/wiki and sched/wiki2 (Maui or Moab scheduler) disable the ability of a non-root user to change a job's comment field (used by Maui/Moab for storing scheduler state information). -- For sched/wiki (Maui) add pending job's future start time to the state info reported to Maui. -- Improve reliability of job requeue logic on node failure. -- Add logic to ping non-responsive nodes even if SlurmdTimeout=0. This permits the node to be returned to use when it starts responding rather than remaining in a non-usable state. -- Honor HealthCheckInterval values that are smaller than SlurmdTimeout. -- For non-responding nodes, log them all on a single line with a hostlist expression rather than one line per node. Frequency of log messages is dependent upon SlurmctldDebug value from 300 seconds at SlurmctldDebug<=3 to 1 second at SlurmctldDebug>=5. -- If a DOWN node is resumed, set its state to IDLE & NOT_RESPONDING and ping the node immediately to clear the NOT_RESPONDING flag. -- Log that a job's time limit is reached, but don't sent SIGXCPU. -- Fixed gid to be set in slurmstepd when run by root -- Changed getpwent to getpwent_r in the slurmctld and slurmd -- Increase timeout on most slurmdbd communications to 60 secs (time for substantial database updates). -- Treat srun option of --begin= with a value of now without a numeric component as a failure (e.g. "--begin=now+hours"). -- Eliminate a memory leak associated with notifying srun of allocated nodes having failed. -- Add scontrol shutdown option of "slurmctld" to just shutdown the slurmctld daemon and leave the slurmd daemons running. -- Do not require JobCredentialPrivateKey or JobCredentialPublicCertificate in slurm.conf if using CryptoType=crypto/munge. -- Remove SPANK support from sbatch. * Changes in SLURM 1.3.6 ======================== -- Add new function to get information for a single job rather than always getting information for all jobs. Improved performance of some commands. NOTE: This new RPC means that the slurmctld daemons should be updated before or at the same time as the compute nodes in order to process it. -- In salloc, sbatch, and srun replace --task-mem options with --mem-per-cpu (--task-mem will continue to be accepted for now, but is not documented). Replace DefMemPerTask and MaxMemPerTask with DefMemPerCPU, DefMemPerNode, MaxMemPerCPU and MaxMemPerNode in slurm.conf (old options still accepted for now, but mapped to "PerCPU" parameters and not documented). Allocate a job's memory memory at the same time that processors are allocated based upon the --mem or --mem-per-cpu option rather than when job steps are initiated. -- Altered QOS in accounting to be a list of admin defined states, an account or user can have multiple QOS's now. They need to be defined using 'sacctmgr add qos'. They are no longer an enum. If none are defined Normal will be the QOS for everything. Right now this is only for use with MOAB. Does nothing outside of that. -- Added spank_get_item support for field S_STEP_CPUS_PER_TASK. -- Make corrections in spank_get_item for field S_JOB_NCPUS, previously reported task count rather than CPU count. -- Convert configuration parameter PrivateData from on/off flag to have separate flags for job, partition, and node data. See "man slurm.conf" for details. -- Fix bug, failed to load DisableRootJobs configuration parameter. -- Altered sacctmgr to always return a non-zero exit code on error and send error messages to stderr. * Changes in SLURM 1.3.5 ======================== -- Fix processing of auth/munge authtentication key for messages originating in slurmdbd and sent to slurmctld. -- If srun is allocating resources (not within sbatch or salloc) and MaxWait is configured to a non-zero value then wait indefinitely for the resource allocation rather than aborting the request after MaxWait time. -- For Moab only: add logic to reap defunct "su" processes that are spawned by slurmd to load user's environment variables. -- Added more support for "dumping" account information to a flat file and read in again to protect data incase something bad happens to the database. -- Sacct will now report account names for job steps. -- For AIX: Remove MP_POERESTART_ENV environment variable, disabling poerestart command. User must explicitly set MP_POERESTART_ENV before executing poerestart. -- Put back notification that a job has been allocated resources when it was pending. * Changes in SLURM 1.3.4 ======================== -- Some updates to man page formatting from Gennaro Oliva, ICAR. -- Smarter loading of plugins (doesn't stat every file in the plugin dir) -- In sched/backfill avoid trying to schedule jobs on DOWN or DRAINED nodes. -- forward exit_code from step completion to slurmdbd -- Add retry logic to socket connect() call from client which can fail when the slurmctld is under heavy load. -- Fixed bug when adding associations to add correctly. -- Added support for associations for user root. -- For Moab, sbatch --get-user-env option processed by slurmd daemon rather than the sbatch command itself to permit faster response for Moab. -- IMPORTANT FIX: This only effects use of select/cons_res when allocating resources by core or socket, not by CPU (default for SelectTypeParameter). We are not saving a pending job's task distribution, so after restarting slurmctld, select/cons_res was over-allocating resources based upon an invalid task distribution value. Since we can't save the value without changing the state save file format, we'll just set it to the default value for now and save it in Slurm v1.4. This may result in a slight variation on how sockets and cores are allocated to jobs, but at least resources will not be over-allocated. -- Correct logic in accumulating resources by node weight when more than one job can run per node (select/cons_res or partition shared=yes|force). -- slurm.spec file updated to avoid creating empty RPMs. RPM now *must* be built with correct specification of which packages to build or not build. See the top of the slurm.spec file for information about how to control package building specification. -- Set SLURM_JOB_CPUS_PER_NODE for jobs allocated using the srun command. It was already set for salloc and sbatch commands. -- Fix to handle suspended jobs that were cancelled in accounting -- BLUEGENE - fix to only include bps given in a name from the bluegene.conf file. -- For select/cons_res: Fix record-keeping for core allocations when more than one partition uses a node or there is more than one socket per node. -- In output for "scontrol show job" change "StartTime" header to "EligibleTime" for pending jobs to accurately describe what is reported. -- Add more slurmdbd.conf paramters: ArchiveScript, ArchiveAge, JobPurge, and StepPurge (not fully implemented yet). -- Add slurm.conf parameter EnforcePartLimits to reject jobs which exceed a partition's size and/or time limits rather than leaving them queued for a later change in the partition's limits. NOTE: Not reported by "scontrol show config" to avoid changing RPCs. It will be reported in SLURM version 1.4. -- Added idea of coordinator to accounting. A coordinator can add associations between exsisting users to the account or any sub-account they are coordinator to. They can also add/remove other coordinators to those accounts. -- Add support for Hostname and NodeHostname in slurm.conf being fully qualified domain names (by Vijay Ramasubramanian, University of Maryland). For more information see "man slurm.conf". * Changes in SLURM 1.3.3 ======================== -- Add mpi_openmpi plugin to the main SLURM RPM. -- Prevent invalid memory reference when using srun's --cpu_bind=cores option (slurm-1.3.2-1.cea1.patch from Matthieu Hautreux, CEA). -- Task affinity plugin modified to support a particular cpu bind type: cores, sockets, threads, or none. Accomplished by setting an environment variable SLURM_ENFORCE_CPU_TYPE (slurm-1.3.2-1.cea2.patch from Matthieu Hautreux, CEA). -- For BlueGene only, log "Prolog failure" once per job not once per node. -- Reopen slurmctld log file after reconfigure or SIGHUP is received. -- In TaskPlugin=task/affinity, fix possible infinite loop for slurmd. -- Accounting rollup works for mysql plugin. Automatic rollup when using slurmdbd. -- Copied job stat logic out of sacct into sstat in the future sacct -stat will be deprecated. -- Correct sbatch processing of --nice option with negative values. -- Add squeue formatted print option %Q to print a job's integer priority. -- In sched/backfill, fix bug that was changing a pending job's shared value to zero (possibly changing a pending job's resource requirements from a processor on some node to the full node). * Changes in SLURM 1.3.2 ======================== -- Get --ntasks-per-node option working for sbatch command. -- BLUEGENE: Added logic to give back a best block on overlapped mode in test_only mode -- BLUEGENE: Updated debug info and man pages for better help with the numpsets option and to fail correctly with bad image request for building blocks. -- In sched/wiki and sched/wiki2 properly support Slurm license consumption (job state reported as "Hold" when required licenses are not available). -- In sched/wiki2 JobWillRun command, don't return an error code if the job(s) can not be started at that time. Just return an error message (from Doug Wightman, CRI). -- Fix bug if sched/wiki or sched/wiki2 are configured and no job comment is set. -- scontrol modified to report partition partition's "DisableRootJobs" value. -- Fix bug in setting host address for PMI communications (mpich2 only). -- Fix for memory size accounting on some architectures. -- In sbatch and salloc, change --dependency's one letter option from "-d" to "-P" (continue to accept "-d", but change the documentation). -- Only check that task_epilog and task_prolog are runable by the job's user, not as root. -- In sbatch, if specifying an alternate directory (--workdir/-D), then input, output and error files are in that directory rather than the directory from which the command is executed -- NOTE: Fully operational with Moab version 5.2.3+. Change SUBMITCMD in moab.cfg to be the location of sbatch rather than srun. Also set HostFormat=2 in SLURM's wiki.conf for improved performance. -- NOTE: We needed to change an RPC from version 1.3.1. You must upgrade all nodes in a cluster from v1.3.1 to v1.3.2 at the same time. -- Postgres plugin will work from job accounting, not for association management yet. -- For srun/sbatch --get-user-env option (Moab use only) look for "env" command in both /bin and /usr/sbin (for Suse Linux). -- Fix bug in processing job feature requests with node counts (could fail to schedule job if some nodes have not associated features). -- Added nodecnt and gid to jobcomp/script -- Insure that nodes select in "srun --will-run" command or the equivalent in sched/wiki2 are in the job's partition. -- BLUGENE - changed partition Min|MaxNodes to represent c-node counts instead of base partitions -- In sched/gang only, prevent possible invalid memory reference when slurmctld is reconfigured, e.g. "scontrol reconfig". -- In select/linear only, prevent invalid memory reference in log message when nodes are added to slurm.conf and then "scontrol reconfig" is executed. * Changes in SLURM 1.3.1 ======================== -- Correct logic for processing batch job's memory limit enforcement. -- Fix bug that was setting a job's requeue value on any update of the job using the "scontrol update" command. The invalid value of an updated job prevents it's recovery when slurmctld restarts. -- Add support for cluster-wide consumable resources. See "Licenses" parameter in slurm.conf man page and "--licenses" option in salloc, sbatch and srun man pages. -- Major changes in select/cons_res to support FastSchedule=2 with more resources configured than actually exist (useful for testing purposes). -- Modify srun --test-only response to include expected initiation time for a job as well as the nodes to be allocated and processor count (for use by Moab). -- Correct sched/backfill to properly honor job dependencies. -- Correct select/cons_res logic to allocate CPUs properly if there is more than one thread per core (previously failed to allocate all cores). -- Correct select/linear logic in shared job count (was off by 1). -- Add support for job preeption based upon partition priority (in sched/gang, preempt.patch from Chris Holmes, HP). -- Added much better logic for mysql accounting. -- Finished all basic functionality for sacctmgr. -- Added load file logic to sacctmgr for setting up a cluster in one step. -- NOTE: We needed to change an RPC from version 1.3.0. You must upgrade all nodes in a cluster from v1.3.0 to v1.3.1 at the same time. -- NOTE: Work is currently underway to improve placement of jobs for gang scheduling and preemption. -- NOTE: Work is underway to provide additional tools for reporting accounting information. * Changes in SLURM 1.3.0 ======================== -- In sched/wiki2, add processor count to JOBWILLRUN response. -- Add event trigger for node entering DRAINED state. -- Build properly without OpenSSL installed (OpenSSL is recommended, but not required). -- Added slurmdbd, and modified accounting_storage plugin to talk to it. Allowing multiple slurm systems to securly store and gather information not only about jobs, but the system also. See accounting web page for more information. * Changes in SLURM 1.3.0-pre11 ============================== -- Restructure the sbcast RPC to take advantage of larger buffers available in Slurm v1.3 RPCs. -- Fix several memory leaks. -- In scontrol, show job's Requeue value, permit change of Requeue and Comment values. -- In slurmctld job record, add QOS (quality of service) value for accounting purposes with Maui and Moab. -- Log to a job's stderr when it is being cancelled explicitly or upon reaching it's time limit. -- Only permit a job's account to be changed while that job is PENDING. -- Fix race condition in job suspend/resume (slurmd.sus_res.patch from HP). * Changes in SLURM 1.3.0-pre10 ============================== -- Add support for node-specific "arch" (architecture) and "os" (operating system) fields. These fields are set based upon values reported by the slurmd daemon on each compute node using SLURM_ARCH and SLURM_OS environment variables (if set, the uname function otherwise) and are intended to support changes in real time changes in operating system. These values are reported by "scontrol show node" plus the sched/wiki and sched/wiki2 plugins for Maui and Moab respectively. -- In sched/wiki and sched/wiki2: add HostFormat and HidePartitionJobs to "scontrol show config" SCHEDULER_CONF output. -- In sched/wiki2: accept hostname expression as input for GETNODES command. -- Add JobRequeue configuration parameter and --requeue option to the sbatch command. -- Add HealthCheckInterval and HealthCheckProgram configuration parameters. -- Add SlurmDbdAddr, SlurmDbdAuthInfo and SlurmDbdPort configuration parameters. -- Modify select/linear to achieve better load leveling with gang scheduler. -- Develop the sched/gang plugin to support select/linear and select/cons_res. If sched/gang is enabled and Shared=FORCE is configured for a partition, this plugin will gang-schedule or "timeslice" jobs that share common resources within the partition. Note that resources that are shared across partitions are not gang-scheduled. -- Add EpilogMsgTime configuration parameter. See "man slurm.conf" for details. -- Increase default MaxJobCount configuration parameter from 2000 to 5000. -- Move all database common files from src/common to new lib in src/database. -- Move sacct to src/accounting added sacctmgr for scontrol like operations to accounting infrastructure. -- Basic functions of sacctmgr in place to make for administration of accounting. -- Moved clusteracct_storage plugin to accounting_storage plugin, jobacct_storage is still it's own plugin for now. -- Added template for slurm php extention. -- Add infrastructure to support allocation of cluster-wide licenses to jobs. Full support will be added some time after version 1.3.0 is released. -- In sched/wiki2 with select/bluegene, add support for WILLRUN command to accept multiple jobs with start time specifications. * Changes in SLURM 1.3.0-pre9 ============================= -- Add spank support to sbatch. Note that spank_local_user() will be called with step_layout=NULL and gid=SLURM_BATCH_SCRIPT and spank_fini() will be called immediately afterwards. -- Made configure use mysql_config to find location of mysql database install Removed bluegene specific information from the general database tables. -- Re-write sched/backfill to utilize new will-run logic in the select plugins. It now supports select/cons_res and all job options (required nodes, excluded nodes, contiguous, etc.). -- Modify scheduling logic to better support overlapping partitions. -- Add --task-mem option and remove --job-mem option from srun, salloc, and sbatch commands. Enforce step memory limit, if specified and there is no job memory limit specified (--mem). Also see DefMemPerTask and MaxMemPerTask in "man slurm.conf". Enforcement is dependent upon job accounting being enabled with non-zero value for JoabAcctGatherFrequency. -- Change default node tmp_disk size to zero (for diskless nodes). * Changes in SLURM 1.3.0-pre8 ============================= -- Modify how strings are packed in the RPCs, Maximum string size increased from 64KB (16-bit size field) to 4GB (32-bit size field). -- Fix bug that prevented time value of "INFINITE" from being processed. -- Added new srun/sbatch option "--open-mode" to control how output/error files are opened ("t" for truncate, "a" for append). -- Added checkpoint/xlch plugin for use with XLCH (Hongjia Cao, NUDT). -- Added srun option --checkpoint-path for use with XLCH (Hongjia Cao, NUDT). -- Added new srun/salloc/sbatch option "--acctg-freq" for user control over accounting data collection polling interval. -- In sched/wiki2 add support for hostlist expression use in GETNODES command with HostFormat=2 in the wiki.conf file. -- Added new scontrol option "setdebug" that can change the slurmctld daemons debug level at any time (Hongjia Cao, NUDT). -- Track total total suspend time for jobs and steps for accounting purposes. -- Add version information to partition state file. -- Added 'will-run' functionality to all of the select plugins (bluegene, linear, and cons_res) to return node list and time job can start based on other jobs running. -- Major restructuring of node selection logic. select/linear now supports partition max_share parameter and tries to match like size jobs on the same nodes to improve gang scheduling performance. Also supports treating memory as consumable resource for job preemption and gang scheduling if SelectTypeParameter=CR_Memory in slurm.conf. -- BLUEGENE: Reorganized bluegene plugin for maintainability sake. -- Major restructuring of data structures in select/cons_res. -- Support job, node and partition names of arbitrary size. -- Fix bug that caused slurmd to hang when using select/linear with task/affinity. * Changes in SLURM 1.3.0-pre7 ============================= -- Fix a bug in the processing of srun's --exclusive option for a job step. * Changes in SLURM 1.3.0-pre6 ============================= -- Add support for configurable number of jobs to share resources using the partition Shared parameter in slurm.conf (e.g. "Shared=FORCE:3" for two jobs to share the resources). From Chris Holmes, HP. -- Made salloc use api instead of local code for message handling. * Changes in SLURM 1.3.0-pre5 ============================= -- Add select_g_reconfigure() function to node changes in slurmctld configuration that can impact node scheduling. -- scontrol to set/get partition's MaxTime and job's Timelimit in minutes plus new formats: min:sec, hr:min:sec, days-hr:min:sec, days-hr, etc. -- scontrol "notify" command added to send message to stdout of srun for specified job id. -- For BlueGene, make alpha part of node location specification be case insensitive. -- Report scheduler-plugin specific configuration information with the "scontrol show configuration" command on the SCHEDULER_CONF line. This information is not found in the "slurm.conf" file, but a scheduler plugin specific configuration (e.g. "wiki.conf"). -- sview partition information reported now includes partition priority. -- Expand job dependency specification to support concurrent execution, testing of job exit status and multiple job IDs. * Changes in SLURM 1.3.0-pre4 ============================= -- Job step launch in srun is now done from the slurm api's all further modifications to job launch should be done there. -- Add new partition configuration parameter Priority. Add job count to Shared parameter. -- Add new configuration parameters DefMemPerTask, MaxMemPerTask, and SchedulerTimeSlice. -- In sched/wiki2, return REJMESSAGE with details on why a job was requeued (e.g. what node failed). * Changes in SLURM 1.3.0-pre3 ============================= -- Remove slaunch command -- Added srun option "--checkpoint=time" for job step to automatically be checkpointed on a period basis. -- Change behavior of "scancel -s KILL " to send SIGKILL to all job steps rather than cancelling the job. This now matches the behavior of all other signals. "scancel " still cancels the job and all steps. -- Add support for new job step options --exclusive and --immediate. Permit job steps to be queued when resources are not available within an existing job allocation to dedicate the resources to the job step. Useful for executing simultaneous job steps. Provides resource management both at the level of jobs and job steps. -- Add support for feature count in job constraints, for example srun --nodes=16 --constraint=graphics*4 ... Based upon work by Kumar Krishna (HP, India). -- Add multi-core options to salloc and sbatch commands (sbatch.patch and cleanup.patch from Chris Holmes, HP). -- In select/cons_res properly release resources allocated to job being suspended (rmbreak.patch, from Chris Holmes, HP). -- Removed database and jobacct plugin replaced with jobacct_storage and jobacct_gather for easier hooks for further expansion of the jobacct plugin. * Changes in SLURM 1.3.0-pre2 ============================= -- Added new srun option --pty to start job with pseudo terminal attached to task 0 (all other tasks have I/O discarded) -- Disable user specifying jobid when sched/wiki2 configured (needed for Moab releases until early 2007). -- Report command, args and working directory for batch jobs with "scontrol show job". * Changes in SLURM 1.3.0-pre1 ============================= -- !!! SRUN CHANGES !!! The srun options -A/--allocate, -b/--batch, and -a/--attach have been removed! That functionality is now available in the separate commands salloc, sbatch, and sattach, respectively. -- Add new node state FAILING plus trigger for when node enters that state. -- Add new configuration paramter "PrivateData". This can be used to prevent a user from seeing jobs or job steps belonging to other users. -- Added configuration parameters for node power save mode: ResumeProgram ResumeRate, SuspendExcNodes, SuspendExcParts, SuspendProgram and SuspendRate. -- Slurmctld maintains the IP address (rather than hostname) for srun communications. This fixes some possible network routing issues. -- Added global database plugin. Job accounting and Job completion are the first to use it. Follow documentation to add more to the plugin. -- Removed no-longer-needed jobacct/common/common_slurmctld.c since that is replaced by the database plugin. -- Added new configuration parameter: CryptoType. Moved existing digital signature logic into new plugin: crypto/openssl. Added new support for crypto/munge (available with GPL license). * Changes in SLURM 1.2.36 ========================= -- For spank_get_item(S_JOB_ARGV) for batch job with script input via STDIN, set argc value to 1 (rather than 2, argv[0] still set to path of generated script). -- sacct will now display more properly allocations made with salloc with only one step. * Changes in SLURM 1.2.35 ========================= -- Permit SPANK plugins to dynamically register options at runtime base upon configuration or other runtime checks. -- Add "include" keywork to SPANK plugstack.conf file to optionally include other configuration files or directories of configuration files. -- Srun to wait indefinitely for resource allocation to be made. Used to abort after two minutes. * Changes in SLURM 1.2.34 ========================= -- Permit the cancellation of a job that is in the process of being requeued. -- Ignore the show_flag when getting job, step, node or partition information for user root. -- Convert some functions to thread-safe versions: getpwnam, getpwuid, getgrnam, and getgrgid to similar functions with "_r" suffix. While no failures have been observed, a race condition would in the worst case permit a user access to a partition not normally allowed due to the AllowGroup specification or the wrong user identified in an accounting record. The job would NOT be run as the wrong user. -- For PMI only (MPICH2/MVAPICH2) base address to send messages to (the srun) upon the address from which slurmd gets the task launch request rather then "hostname" where srun executes. -- Make test for StateSaveLocation directory more comprehensive. -- For jobcomp/script plugin, PROCS environment variable is now the actual count of allocated processors rather than the count of processes to be started. * Changes in SLURM 1.2.33 ========================= -- Cancelled or Failed jobs will now report their job and step id on exit -- Add SPANK items available to get: SLURM_VERSION, SLURM_VERSION_MAJOR, SLURM_VERISON_MINOR and SLURM_VERSION_MICRO. -- Fixed handling of SIGPIPE in srun. Abort job. -- Fix bug introduced to MVAPICH plugin preventing use of TotalView debugger. -- Modify slurmctld to get srun/salloc network address based upon the incoming message rather than hostname set by the user command (backport of logic in SLURM v1.3). * Changes in SLURM 1.2.32 ========================= -- LSF only: Enable scancel of job in RootOnly partition by the job's owner. -- Add support for sbatch --distribution and --network options. -- Correct pending job's wait reason to "Priority" rather than "Resources" if required resources are being held in reserve for a higher priority job. -- In sched/wiki2 (Moab) report a node's state as "Drained" rather than "Draining" if it has no allocated work (An undocumented Moab wiki option, see CRI ticket #2394). -- Log to job's output when it is cancelled or reaches it's time limit (ported from existing code in slurm v1.3). -- Add support in salloc and sbatch commands for --network option. -- Add support for user environment variables that include '\n' (e.g. bash functions). -- Partial rewrite of mpi/mvapich plugin for improved scalability. * Changes in SLURM 1.2.31 ========================= -- For Moab only: If GetEnvTimeout=0 in slurm.conf then do not run "su" to get the user's environment, only use the cache file. -- For sched/wiki2 (Moab), treat the lack of a wiki.conf file or the lack of a configured AuthKey as a fatal error (lacks effective security). -- For sched/wiki and sched/wiki2 (Maui or Moab) report a node's state as Busy rather than Running when allocated if SelectType=select/linear. Moab was trying to schedule job's on nodes that were already allocated to jobs that were hidden from it via the HidePartitionJobs in Slurm's wiki.conf. -- In select/cons_res improve the resource selection when a job has specified a processor count along with a maximum node count. -- For an srun command with --ntasks-per-node option and *no* --ntasks count, spawn a task count equal to the number of nodes selected multiplied by the --ntasks-per-node value. -- In jobcomp/script: Set TZ if set in slurmctld's environment. -- In srun with --verbose option properly format CPU allocation information logged for clusters with 1000+ nodes and 10+ CPUs per node. -- Process a job's --mail_type=end option on any type of job termination, not just normal completion (e.g. all failure modes too). * Changes in SLURM 1.2.30 ========================= -- Fix for gold not to print out 720 error messages since they are potentally harmful. -- In sched/wiki2 (Moab), permit changes to a pending job's required features: CMD=CHANGEJOB ARG= RFEATURES= -- Fix for not aborting when node selection doesn't load, fatal error instead -- In sched/wiki and sched/wiki2 DO NOT report a job's state as "Hold" if it's dependencies have not been satisfied. This reverses a changed made in SLURM version 1.2.29 (which was requested by Cluster Resources, but places jobs in a HELD state indefinitely). * Changes in SLURM 1.2.29 ========================= -- Modified global configuration option "DisableRootJobs" from number (0 or 1) to boolean (YES or NO) to match partition parameter. -- Set "DisableRootJobs" for a partition to match the global parameters value for newly created partitions. -- In sched/wiki and sched/wiki2 report a node's updated features if changed after startup using "scontrol update ..." command. -- In sched/wiki and sched/wiki2 report a job's state as "Hold" if it's dependencies have not been satisfied. -- In sched/wiki and sched/wiki2 do not process incoming requests until slurm configuration is completely loaded. -- In sched/wiki and sched/wiki2 do not report a job's node count after it has completed (slurm decrements the allocated node count when the nodes transition from completing to idle state). -- If job prolog or epilog fail, log the program's exit code. -- In jobacct/gold map job names containing any non-alphanumeric characters to '_' to avoid MySQL parsing problems. -- In jobacct/linux correct parsing if command name contains spaces. -- In sched/wiki and sched/wiki2 report make job info TASK count reflect the actual task allocation (not requested tasks) even after job terminates. Useful for accounting purposes only. * Changes in SLURM 1.2.28 ========================= -- Added configuration option "DisableRootJobs" for parameter "PartitionName". See "man slurm.conf" for details. -- Fix for faking a large system to correctly handle node_id in the task afffinity plugin for ia64 systems. * Changes in SLURM 1.2.27 ========================= -- Record job eligible time in accounting database (for jobacct/gold only). -- Prevent user root from executing a job step within a job allocation belonging to another user. -- Fixed limiting issue for strings larger than 4096 in xstrfmtcat -- Fix bug in how Slurm reports job state to Maui/Moab when a job is requeued due to a node failure, but we can't terminate the job's spawned processes. Job was being reported as PENDING when it was really still COMPLETING. -- Added patch from Jerry Smith for qstat -a output -- Fixed looking at the correct perl path for Slurm.pm in torque wrappers. -- Enhance job requeue on node failure to be more robust. -- Added configuration parameter "DisableRootJobs". See "man slurm.conf" for details. -- Fixed issue with account = NULL in Gold job accounting plugin * Changes in SLURM 1.2.26 ========================= -- Correct number of sockets/cores/threads reported by slurmd (from Par Andersson, National Supercomputer Centre, Sweden). -- Update libpmi linking so that libslurm is not required for PMI use (from Steven McDougal, SiCortex). -- In srun and sbatch, do not check the PATH env var if an absolute pathname of the program is specified (previously reported an error if no PATH). -- Correct output of "sinfo -o %C" (CPU counts by node state). * Changes in SLURM 1.2.25 ========================= -- Bug fix for setting exit code in accounting for batch script. -- Add salloc option, --no-shell (for LSF). -- Added new options for sacct output -- mvapich: Ensure MPIRUN_ID is unique for all job steps within a job. (Fixes crashes when running multiple job steps within a job on one node) -- Prevent "scontrol show job" from failing with buffer overflow when a job has a very long Comment field. -- Make certain that a job step is purged when a job has been completed. Previous versions could have the job step persist if an allocated node went DOWN and the slurmctld restarted. -- Fix bug in sbcast that can cause communication problems for large files. -- Add sbcast option -t/--timeout and SBCAST_TIMEOUT environment variable to control message timeout. -- Add threaded agent to manage a queue of Gold update requests for performance reasons. -- Add salloc options --chdir and --get-user-env (for Moab). -- Modify scontrol update to support job comment changes. -- Do not clear a DRAINED node's reason field when slurmctld restarts. -- Do not cancel a pending job if Moab or Maui try to start it on unusable nodes. Leave the job queued. -- Add --requeue option to srun and sbatch (these undocumented options have no effect in slurm v1.2, but are legitimate options in slurm v1.3). * Changes in SLURM 1.2.24 ========================= -- In sched/wiki and sched/wiki2, support non-zero UPDATE_TIME specification for GETNODES and GETJOBS commands. -- Bug fix for sending accounting information multiple times for same info. patch from Hongjia Cao (NUDT). -- BLUEGENE - try FILE pointer rotation logic to avoid core dump on bridge log rotate -- Spread out in time the EPILOG_COMPLETE messages from slurmd to slurmctld to avoid message congestions and retransmission. * Changes in SLURM 1.2.23 ========================= -- Fix for libpmi to not export unneeded variables like xstr* -- BLUEGENE - added per partition dynamic block creation -- fix infinite loop bug in sview when there were multiple partitions -- Send message to srun command when a job is requeued due to node failure. Note this will be overwritten in the output file unless JobFileAppend is set in slurm.conf. In slurm version 1.3, srun's --open-mode=append option will offer this control for each job. -- Change a node's default TmpDisk from 1MB to 0MB and change job's default disk space requirement from 1MB to 0MB. -- In sched/wiki (Maui scheduler) specify a QOS (quality of service) by specifying an account of the form "qos-name". -- In select/linear, fix bug in scheduling required nodes that already have a job running on them (req.load.patch from Chris Holmes, HP). -- For use with Moab only: change timeout for srun/sbatch --get-user-env option to 2 secs, don't get DISPLAY environment variables, but explicitly set ENVIRONMENT=BATCH and HOSTNAME to the execution host of the batch script. -- Add configuration parameter GetEnvTimeout for use with Moab. See "man slurm.conf" for details. -- Modify salloc and sbatch to accept both "--tasks" and "--ntasks" as equivalent options for compatibility with srun. -- If a partition's node list contains space separators, replace them with commas for easier parsing. -- BLUEGENE - fixed bug in geometry specs when creating a block. -- Add support for Moab and Maui to start jobs with select/cons_res plugin and jobs requiring more than one CPU per task. * Changes in SLURM 1.2.22 ========================= -- In sched/wiki2, add support for MODIFYJOB option "MINSTARTTIME=