I use a 10GB glusterfs mirrored fs for the slurmctl saved state between the Primary and Backup. Seems to work well enough, for our smallish 45 node cluster. Super easy to setup as well.
On Fri, Jul 29, 2016 at 6:52 AM, jupiter <[email protected]> wrote: > Hi, > > Does slurm have the mechanism to synchronise scheduler jobs and states > between the ControlMachine and BackupController? If not and if it is admin's > responsibility to take care about it, there could be following two options: > > 1. Set up a single node for mysql and slurmdb, point the variables of > StateSaveLocation and SlurmdSpoolDir to the remote single db node. The > question is, how can the values in StateSaveLocation and SlurmdSpoolDir be > set up to a remote location? Or it is not possible? > > 2. Set the variables of StateSaveLocation and SlurmdSpoolDir to a share > location which mount to both StateSaveLocation and SlurmdSpoolDir. > > What is your recommendation? > > Thank you. > > Kind regards, > > - j
