Commits · 718fe3d7a2767e842a653038b5f78aaac02bf06d · tud-zih-energy / Slurm

Oct 03, 2003
- Note system updates for 0.2.20 release. · 718fe3d7
  Moe Jette authored 21 years ago
  
  718fe3d7
- Resend KILL_JOB message if a completing job hangs. This can recover from the · 5d9fa3f6
  Moe Jette authored 21 years ago
  
  lost of an EPILOG_COMPLETE message.
  5d9fa3f6
- Check for consistent plugin version number before decoding contents. · d3966576
  Moe Jette authored 21 years ago
  
  d3966576
- Identify some default parameter values per user feedback. · 4e01e511
  Moe Jette authored 21 years ago
  
  4e01e511
- Reject job requests explicitly requesting more than 8 sets of consecutive · ebeb27bd
  Moe Jette authored 21 years ago
  
  nodes. This is a temporary patch until DPCS is upgraded in mid-October.
  ebeb27bd
- Set log prefix for jobid.stepid to use %u format instead of %d. · e60b231f
  Moe Jette authored 21 years ago
  
  e60b231f
Oct 02, 2003
- Modify popt use to conform with spec (and not fail under RedHat 9). · f8481775
  Moe Jette authored 21 years ago
  
  f8481775
- Fix use of popt for proper operation under RH9. · 5af828d7
  Moe Jette authored 21 years ago
  
  5af828d7
- o fix for problem with SIGCHLD and sigwait on RH9-based systems. · 60a563ac
  Mark Grondona authored 21 years ago
  
  Install a dummy handler for SIGCHLD so our process is actually delivered said signal.
  60a563ac
- Remove defunct definitions to eliminate warnings in RH9 system. · b8f7993d
  Moe Jette authored 21 years ago
  
  b8f7993d
Oct 01, 2003
- Replace defunct sigprocmask() with xsignal_sigset_create() and xsignal_block(). · 6a6913f9
  Moe Jette authored 21 years ago
  
  6a6913f9
- Update slurm_receive_msg() for new argument. · 87a38457
  Moe Jette authored 21 years ago
  
  87a38457
- o update for bug-fix release 0.2.19 · b66fa1cb
  Mark Grondona authored 21 years ago
  
  b66fa1cb
- o further fixes for invalid writes in hostlist_ranged_string · 42d1c6f5
  Mark Grondona authored 21 years ago
  
  42d1c6f5
- Correct default PluginDir value, /usr/local/lib/slurm. · 8c6ecb85
  Moe Jette authored 21 years ago
  
  8c6ecb85
Sep 30, 2003
- Establish META file for 0.2.18 release. · 23d1eefe
  Moe Jette authored 21 years ago
  
  23d1eefe
- Move job priority=1 setting all into the function schedule(). · b780206c
  Moe Jette authored 21 years ago
  
  b780206c
- Increase the initial buffer size for constructing a job's hostlist string. · 223b1708
  Moe Jette authored 21 years ago
  
  223b1708
- *** empty log message *** · ab43ebb0
  Mark Grondona authored 21 years ago
  
  ab43ebb0
- o add new --jobid option · f9f996d7
  Mark Grondona authored 21 years ago
  
  f9f996d7
- o print error message if slurmd session manager is killed by a signal · c3b8b6c1
  Mark Grondona authored 21 years ago
  
  c3b8b6c1
- Fix bug where building of hostlist string went off the end of its allocated · 3744aa68
  Moe Jette authored 21 years ago
  
  buffer and corrupted memory.
  3744aa68
- o Do not close input stream when output buffer is full (gnats:294) · 3ceeec27
  Mark Grondona authored 21 years ago
  
  3ceeec27
Sep 29, 2003
- Identify updates since previous slurm release. · fea6a2eb
  Moe Jette authored 21 years ago
  
  fea6a2eb
- If a job is submitted with explicit nodes selected (like DPCS), then · ba88d8cd
  Moe Jette authored 21 years ago
  
  let that job be scheduled even if there are jobs in COMPLETING state.
  ba88d8cd
- Convert fatal() calls to error()/abort() calls or xassert() calls. · c59d18d5
  Moe Jette authored 21 years ago
  
  Fatal() no longer calls abort(), but terminates job using exit(1).
  c59d18d5
- Add missing job_state_string_compact for node_fail (nf). · ff29479e
  Moe Jette authored 21 years ago
  
  ff29479e
- List the possible job states. · 4adf8800
  Moe Jette authored 21 years ago
  
  4adf8800
- o do not set SLURM_NNODES SLURM_JOBID SLURM_NPROCS within session manger · 28f72600
  Mark Grondona authored 21 years ago
  
  28f72600
- o fatal() no longer calls abort(). Use error(); abort() instead. · d8fec52e
  Mark Grondona authored 21 years ago
  
  d8fec52e
- Prevent the scheduling of jobs while one is in the process of completing · 94bd5639
  Moe Jette authored 21 years ago
  
  (state is completing and end time is within KillWait (+2 seconds).
  94bd5639
- Add missing node state save on epilog completion failure and job state · be1abd0e
  Moe Jette authored 21 years ago
  
  save on job_step completion. This should reduce the likelyhood of information being lost on a slurmctld restart.
  be1abd0e
- Fix allignment of code. No changes in logic. · f1592773
  Moe Jette authored 21 years ago
  
  f1592773
- Increase interval of periodic schedule() calls from 30 to 60 seconds. · e5e5749c
  Moe Jette authored 21 years ago
  
  e5e5749c
Sep 27, 2003
- o don't use shm_step_signal when signalling from kill_job · cb487d5d
  Mark Grondona authored 21 years ago
  
  cb487d5d
- o rename static reconfig/shutdown variables to have preceding `_' · 0a178a6b
  Mark Grondona authored 21 years ago
  
  0a178a6b
- Plug some possible memory leaks on various error conditions. · 0b05bda4
  Moe Jette authored 21 years ago
  
  0b05bda4
- Eliminate possible memory leak on g_slurm_auth_pack error. · 3cf37512
  Moe Jette authored 21 years ago
  
  3cf37512
Sep 26, 2003
- End cases fixes for building hostlist string in buffer that is too small. · f0495e31
  Moe Jette authored 21 years ago
  
  Was overwritting data off of buffer passed as argument.
  f0495e31
Sep 25, 2003

o Reset timeouts and max_threads if running under TotalView (gnats:287) · 283a1665

Mark Grondona authored 21 years ago

 o Use hostname instead of ip addr string in MPIR_proctable to allow
   TotalView bulk launch capability.
 o Add --jobid option to srun

283a1665