- 24 May, 2017 11 commits
-
-
Danny Auble authored
-
Danny Auble authored
# Conflicts: # src/common/slurm_acct_gather.c # src/plugins/sched/backfill/backfill.c # src/slurmctld/node_scheduler.c
-
Danny Auble authored
done in that commit, but only in the main scheduler instead of backfill.
-
Tim Wickberg authored
'scontrol update reservationname=foo duration=unlimited' sets INFINITE as the duration, which needs to be translated to a year as is done elsewhere. Otherwise it'll convert to 49710 days, which is definitely wrong. Bug 3836.
-
Brian Christiansen authored
-
Brian Christiansen authored
And handle memory leak. Continuation of d3ee28fa
-
Brian Christiansen authored
-
Alejandro Sanchez authored
build_all_frontend_info() could be called twice unintentionally.
-
Alejandro Sanchez authored
Bug 3443.
-
Alejandro Sanchez authored
_get_begin_next_month() may be called twice in succession as a side-effect of the MIN macro. Store in a local variable first.
-
Alejandro Sanchez authored
MAX() will re-evaluate the higher value argument; if this is a function is may be called twice over, leading to unintended side effects or a crash. Bug 3443.
-
- 23 May, 2017 15 commits
-
-
Brian Christiansen authored
fed.weight was removed in ae0c71d0.
-
Brian Christiansen authored
fed.weight was removed in ae0c71d0 but the packing of it wasn't removed in the case of a null slurmdb_cluster_rec_t.
-
Danny Auble authored
Otherwise the recently newly added '#include <lua.h>' condition will break the build on systems without any Lua installation. Follow on to commit 7cc4d0d8. Bug 3681.
-
Danny Auble authored
This also fixes the fed_mgr on the backup slurmctld to start backup correctly if the backup takes control more than once. Bug 3827
-
Danny Auble authored
-
Brian Christiansen authored
Continuation of a6e19307 CID:169658
-
Brian Christiansen authored
-
Isaac Hartung authored
-
Isaac Hartung authored
-
Isaac Hartung authored
-
Isaac Hartung authored
Since the cluster records are merged into one cluster record in a federation, the total time needs to account for each cluster's time.
-
Alejandro Sanchez authored
It will result in the job being held if it is requeued at a later time. Bug 3751.
-
Tim Shaw authored
Previously, incorrect partitions and duplicated partition names could be shown. The array needs to be incremented by two, not one, as each element is a start-end pair. Bug 3793.
-
Brian Christiansen authored
-
Brian Christiansen authored
-
- 22 May, 2017 12 commits
-
-
Brian Christiansen authored
when configured with memory-leak-debug
-
Brian Christiansen authored
-
Brian Christiansen authored
-
Brian Christiansen authored
-
Brian Christiansen authored
Made obsolete by moving the federation to an async model.
-
Brian Christiansen authored
Don't need it anymore. Was used to hosw communications between sibling clusters.
-
Brian Christiansen authored
Previously it was only opening the connections at startup and then if had something to send to a sibling that it didn't have a connection to. It's better to have them establish the connection when the cluster is added to the federation so that it can detect if the cluster is down later and the clusters can sync jobs when joing the federation.
-
Brian Christiansen authored
-
Brian Christiansen authored
Since moving to a asynchronous model for submitting jobs in the federation, LLC doesn't apply anymore. And since LLC was the only flag, don't display the federation flags in outputs by default anymore. Leaving the variable in the struct in case it is needed later.
-
Brian Christiansen authored
Doesn't apply anymore now that federation communication is asynchronous and jobs are submitted to all siblings without regards to a cluster's weight.
-
Tim Wickberg authored
I cannot fathom why this was added - this would only have influenced something if listen() was implemented as an inline function or a macro, which I don't believe has ever been the case. Regardless, on modern Linux systems that limit is set dynamically through /proc/sys/net/core/somaxconn.
-
Tim Wickberg authored
While not POSIX compliant, BSD and Linux platforms have shipped this function for a long time, and slurmctld/slurmdbd/slurmd have actually been using the version from libc instead of this local redefinition.
-
- 20 May, 2017 2 commits
-
-
Tim Wickberg authored
-
Tim Wickberg authored
Zero cleanup functions exist within the source, remove all this dead infrastructure.
-