Skip to content
Snippets Groups Projects
NEWS 129 KiB
Newer Older
Christopher J. Morrone's avatar
Christopher J. Morrone committed
This file describes changes in recent versions of SLURM. It primarily
documents those changes that are of interest to users and admins.
Morris Jette's avatar
Morris Jette committed
* Changes in SLURM 2.4.0.pre4
=============================
 -- Add logic to cache GPU file information (bitmap index mapping to device
    file number) in the slurmd daemon and transfer that information to the
    slurmstepd whenever a job step is initiated. This is needed to set the
    appropriate CUDA_VISIBLE_DEVICES environment variable value when the
    devices are not in strict numeric order (e.g. some GPUs are skipped).
    Based upon work by Nicolas Bigaouette.
* Changes in SLURM 2.4.0.pre3
=============================
 -- Let a job be submitted even if it exceeds a QOS limit. Job will be left
    in a pending state until the QOS limit or job parameters change. Patch by
    Phil Eckert, LLNL.
Morris Jette's avatar
Morris Jette committed
 -- Add sacct support for the option "--name". Work by Yuri D'Elia, Center for
    Biomedicine, EURAC Research, Italy.
Danny Auble's avatar
Danny Auble committed
 -- BGQ - handle preemption.
 -- Add an srun shepard process to cancel a job and/or step of the srun process
    is killed abnormally (e.g. SIGKILL).
 -- BGQ - handle deadlock issue when a nodeboard goes into an error state.
 -- BGQ - more thorough handling of blocks with multiple jobs running on them.
 -- Fix man2html process to compile in the build directory instead of the
    source dir.
 -- Behavior of srun --multi-prog modified so that any program arguments
    specified on the command line will be appended to the program arguments
    specified in the program configuration file.
jette's avatar
jette committed
 -- Add new command, sdiag, which reports a variety of job scheduling
    statistics. Based upon work by Alejandro Lucero Palau, BSC.
 -- BLUEGENE - Added DefaultConnType to the bluegene.conf file.  This makes it
    so you can specify any connection type you would like (TORUS or MESH) as
    the default in dynamic mode.  Previously it always defaulted to TORUS.
 -- Made squeue -n and -w options more consistent with salloc, sbatch, srun,
    and scancel. Patch by Don Lipari, LLNL.
 -- Have sacctmgr remove user records when no associations exist for that user.
jette's avatar
jette committed
 -- Several header file changes for clean build with NetBSD. Patches from
    Aleksej Saushev.
 -- Fix for possible deadlock in accounting logic: Avoid calling
    jobacct_gather_g_getinfo() until there is data to read from the socket.
 -- Fix race condition that could generate "job_cnt_comp underflow" errors on
    front-end architectures.
 -- BGQ - Fix issue where a system with missing cables could cause core dump.
Morris Jette's avatar
Morris Jette committed
* Changes in SLURM 2.4.0.pre2
=============================
 -- CRAY - Add support for GPU memory allocation using SLURM GRES (Generic
    RESource) support. Work by Steve Trofinoff, CSCS.
 -- Add support for job allocations with multiple job constraint counts. For
    example: salloc -C "[rack1*2&rack2*4]" ... will allocate the job 2 nodes
    from rack1 and 4 nodes from rack2. Support for only a single constraint
    name been added to job step support.
 -- BGQ - Remove old method for marking cnodes down.
 -- BGQ - Remove BGP images from view in sview.
 -- BGQ - print out failed cnodes in scontrol show nodes.
 -- BGQ - Add srun option of "--runjob-opts" to pass options to the runjob
    command.
 -- FRONTEND - handle step launch failure better.
 -- BGQ - Added a mutex to protect the now changing ba_system pointers.
 -- BGQ - added new functionality for sub-block allocations - no preemption
    for this yet though.
 -- Add --name option to squeue to filter output by job name. Patch from Yuri
    D'Elia.
 -- BGQ - Added linking to runjob client libary which gives support to totalview
    to use srun instead of runjob.
 -- Add numeric range checks to scontrol update options. Patch from Phil
    Eckert, LLNL.
 -- Add ReconfigFlags configuration option to control actions of "scontrol
    reconfig". Patch from Don Albert, Bull.
 -- BGQ - handle reboots with multiple jobs running on a block.
 -- BGQ - Add message handler thread to forward signals to runjob process.
* Changes in SLURM 2.4.0.pre1
=============================
 -- BGQ - use the ba_geo_tables to figure out the blocks instead of the old
    algorithm.  The improves timing in the worst cases and simplifies the code
    greatly.
 -- BLUEGENE - Change to output tools labels from BP to Midplane
    (i.e. BP List -> MidplaneList).
 -- BLUEGENE - read MPs and BPs from the bluegene.conf
 -- Modify srun's SIGINT handling logic timer (two SIGINTs within one second) to
    be based microsecond rather than second timer.
 -- Modify advance reservation to accept multiple specific block sizes rather
    than a single node count.
 -- Permit administrator to change a job's QOS to any value without validating
    the job's owner has permission to use that QOS. Based upon patch by Phil
    Eckert (LLNL).
 -- Add trigger flag for a permanent trigger. The trigger will NOT be purged
    after an event occurs, but only when explicitly deleted.
 -- Interpret a reservation with Nodes=ALL and a Partition specification as
    reserving all nodes within the specified partition rather than all nodes
    on the system. Based upon patch by Phil Eckert (LLNL).
 -- Add the ability to reboot all compute nodes after they become idle. The
    RebootProgram configuration parameter must be set and an authorized user
    must execute the command "scontrol reboot_nodes". Patch from Andriy
    Grytsenko (Massive Solutions Limited).
 -- Modify slurmdbd.conf parsing to accept DebugLevel strings (quiet, fatal,
    info, etc.) in addition to numeric values. The parsing of slurm.conf was
    modified in the same fashion for SlurmctldDebug and SlurmdDebug values.
    The output of sview and "scontrol show config" was also modified to report
    those values as strings rather than numeric values.
 -- Changed default value of StateSaveLocation configuration parameter from
    /tmp to /var/spool.
 -- Prevent associations from being deleted if it has any jobs in running,
    pending or suspended state. Previous code prevented this only for running
    jobs.
 -- If a job can not run due to QOS or association limits, then do not cancel
    the job, but leave it pending in a system held state (priority = 1). The
    job will run when its limits or the QOS/association limits change. Based
    upon a patch by Phil Ekcert (LLNL).
 -- BGQ - Added logic to keep track of cnodes in an error state inside of a
    booted block.
 -- Added the ability to update a node's NodeAddr and NodeHostName with
    scontrol. Also enable setting a node's state to "future" using scontrol.
 -- Add a node state flag of CLOUD and save/restore NodeAddr and NodeHostName
    information for nodes with a flag of CLOUD.
 -- Cray: Add support for job reservations with node IDs that are not in
    numeric order. Fix for Bugzilla #5.
 -- BGQ - Fix issue with smap -R
 -- Fix association limit support for jobs queued for multiple partitions.
 -- BLUEGENE - fix issue for sub-midplane systems to create a full system
    block correctly.
 -- BLUEGENE - Added option to the bluegene.conf to tell you are running on
    a sub midplane system.
 -- Added the UserID of the user issuing the RPC to the job_submit/lua
    functions.
 -- Fixed issue where if a job ended with ESLURMD_UID_NOT_FOUND and
    ESLURMD_GID_NOT_FOUND where slurm would be a little over zealous
    in treating missing a GID or UID as a fatal error.
 -- If job time limit exceeds partition maximum, but job's minimum time limit
    does not, set job's time limit to partition maximum at allocation time.
Morris Jette's avatar
Morris Jette committed
* Changes in SLURM 2.3.4
========================
 -- Set DEFAULT flag in partition structure when slurmctld reads the
    configuration file. Patch from Rémi Palancher.
 -- Fix for possible deadlock in accounting logic: Avoid calling
    jobacct_gather_g_getinfo() until there is data to read from the socket.
 -- Fix typo in accounting when using reservations. Patch from Alejandro
    Lucero Palau.
 -- Fix to the multifactor priority plugin to calculate effective usage earlier
    to give a correct priority on the first decay cycle after a restart of the
    slurmctld. Patch from Martin Perry, Bull.
 -- Permit user root to run a job step for any job as any user. Patch from
    Didier Gazen, Laboratoire d'Aerologie.
 -- BLUEGENE - fix for not allowing jobs if all midplanes are drained and all
    blocks are in an error state.
 -- Avoid slurmctld abort due to bad pointer when setting an advanced
    reservation MAINT flag if it contains no nodes (only licenses).
Morris Jette's avatar
Morris Jette committed
 -- Fix bug when requeued batch job is scheduled to run on a different node
    zero, but attemts job launch on old node zero.
Morris Jette's avatar
Morris Jette committed

* Changes in SLURM 2.3.3
========================
 -- Fix task/cgroup plugin error when used with GRES. Patch by Alexander
    Bersenev (Institute of Mathematics and Mechanics, Russia).
 -- Permit pending job exceeding a partition limit to run if its QOS flag is
    modified to permit the partition limit to be exceeded. Patch from Bill
    Brophy, Bull.
 -- BLUEGENE - Fixed preemption issue.
 -- sacct search for jobs using filtering was ignoring wckey filter.
 -- Fixed issue with QOS preemption when adding new QOS.
 -- Fixed issue with comment field being used in a job finishing before it
    starts in accounting.
 -- Add slashes in front of derived exit code when modifying a job.
 -- Handle numeric suffix of "T" for terabyte units. Patch from John Thiltges,
    University of Nebraska-Lincoln.
 -- Prevent resetting a held job's priority when updating other job parameters.
    Patch from Alejandro Lucero Palau, BSC.
Morris Jette's avatar
Morris Jette committed
 -- Improve logic to import a user's environment. Needed with --get-user-env
    option used with Moab. Patch from Mark Grondona, LLNL.
 -- Fix bug in sview layout if node count less than configured grid_x_width.
 -- Modify PAM module to prefer to use SLURM library with same major release
    number that it was built with.
 -- Permit gres count configuration of zero.
 -- Fix race condition where sbcast command can result in deadlock of slurmd
    daemon. Patch by Don Albert, Bull.
 -- Fix bug in srun --multi-prog configuration file to avoid printing duplicate
    record error when "*" is used at the end of the file for the task ID.
 -- Let operators see reservation data even if "PrivateData=reservations" flag
    is set in slurm.conf. Patch from Don Albert, Bull.
 -- Added new sbatch option "--export-file" as needed for latest version of
    Moab. Patch from Phil Eckert, LLNL.
 -- Fix for sacct printing CPUTime(RAW) where the the is greater than a 32 bit
    number.
 -- Fix bug in --switch option with topology resulting in bad switch count use.
    Patch from Alejandro Lucero Palau (Barcelona Supercomputer Center).
 -- Fix PrivateFlags bug when using Priority Multifactor plugin.  If using sprio
    all jobs would be returned even if the flag was set.
    Patch from Bill Brophy, Bull.
 -- Fix for possible invalid memory reference in slurmctld in job dependency
    logic. Patch from Carles Fenoy (Barcelona Supercomputer Center).
* Changes in SLURM 2.3.2
========================
 -- Add configure option of "--without-rpath" which builds SLURM tools without
    the rpath option, which will work if Munge and BlueGene libraries are in
    the default library search path and make system updates easier.
 -- Fixed issue where if a job ended with ESLURMD_UID_NOT_FOUND and
Loading
Loading full blame...