- 07 Sep, 2018 1 commit
-
-
Morris Jette authored
The GRES information now contains socket binding information, necessitating this change to the parsing function
-
- 06 Sep, 2018 8 commits
-
-
Alejandro Sanchez authored
Bug 4479
-
Brian Christiansen authored
"Argument with 'nonnull' attribute passed null"
-
Morris Jette authored
If the Cores count in gres.conf is bad, slurmstepd was generating a fatal error at time of step launch. Rather than quitting then, this change logs the error and ignores the Cores value.
-
Morris Jette authored
Coverity CID 187778
-
Morris Jette authored
Coverity CID 187927
-
Morris Jette authored
Coverity CID 187926
-
Morris Jette authored
-
Morris Jette authored
This reverts some logic added in commit 01336077 for debugging
-
- 05 Sep, 2018 6 commits
-
-
Danny Auble authored
"no_send_gids to go back to the original behavior. Bug 5620
-
Danny Auble authored
LaunchParameters=send_gids. Bug 5620
-
Tim Wickberg authored
sizeof(char) is defined to be 1, and Slurm is already heavily reliant on that throughout. Stop pretending Slurm would work on a hypothetical system in which this was not true. (C99 Standard - 6.5.3.4.3.) Strip off unnecessary casts while here as well.
-
Tim Wickberg authored
Replace with xstrfmtcat(), and resolve this overrun warning.
-
Tim Wickberg authored
Cannot return NULL, it will abort() before then. Remove the unnecessary cast while here.
-
Morris Jette authored
Add configuration parameter "GpuFreqDef" to control a job's default GPU frequency. Also Coverity CID 187908
-
- 04 Sep, 2018 8 commits
-
-
Morris Jette authored
-
Morris Jette authored
Bug 5659
-
Danny Auble authored
This reverts commit 35285e40.
-
Danny Auble authored
Bug 5659
-
Morris Jette authored
Bug 5663
-
Morris Jette authored
-
Morris Jette authored
-
Morris Jette authored
-
- 31 Aug, 2018 17 commits
-
-
Morris Jette authored
-
Tim Wickberg authored
-
Tim Wickberg authored
-
Tim Wickberg authored
-
Tim Wickberg authored
-
Tim Wickberg authored
-
Tim Wickberg authored
-
Marshall Garey authored
Spaces need to be built into each key definition, otherwise it'll all run together. Bug 5657.
-
Tim Wickberg authored
Otherwise assoc_mgr_fill_in_user() will abort on a failed lock annotation. Convert to designated initializer format while fixing this. Bug 5653.
-
Tim Wickberg authored
-
Tim Wickberg authored
-
Tim Wickberg authored
Further work will be needed to track these independently. Bug 5256.
-
Tim Wickberg authored
Move the trigger_backup_ctld_fail() call into ping_controllers(), and use the ping_all boolean as a standin for whether we're the active controller or not. Bug 5256.
-
Tim Wickberg authored
The handling for response_time was a bit odd, as care was needed to make the timestamp values match up. Meaning they're not really time values, just a complicated way of signaling whether the remote controller is responding. So, simplify this by just indicating through a bool whether a given controller is responding or not. Bug 5256.
-
Tim Wickberg authored
If true, ping everyone except outselves. If false, continue to ping only the higher priority controllers. Also move the ctld_ping xmalloc and xfree in here so we don't need to recreate them in the controller code later. Bug 5256.
-
Tim Wickberg authored
Both the primary and backup need to provide similar ping capabilities. This implementation - with multiple threads launched in parallel - is the more advanced one, so start reworking it to be able to use it in the primary controller as well instead of _ping_backup_controller(). Add a boolean option to indicate whether this is running in the active controller or the backup. This will control whether all other controllers are pinged (needed by the primary) or only the higher priority controllers (as done in the backups). Bug 5256.
-
Morris Jette authored
with some configurations test39.1 would fail. This was due to not considering use of GRES that are non-local to available CPUs, even without enforced binding. This was due to always trying to enforce binding on the first pass. If that fails, then consider use of non-local GRES. The problem in the logic is it would never try non-local GRES if avail_res_array->avail_res_cnt was zero. This patch sets its value to the node's CPU count, which may not be ideal, but is a reasonable option.
-