Global infrastructure information
This section present the global ‘stanard’ infrastructure configuration for the cluster. This need no specific knowledge about clustering to understand and manage those configuration.
Storage Infrastructure
For now, we have only one storage server available. The other 7 servers need to review the cooling system and power system on the room.
The current storage is done through NFS, the mount part is done via autoFS which allow mount “on-demand” on worker node. This avoid sequential reboot (NFS server not need to be UP when worker boot) and useless overload on storage system.
The NFS server is t3ps001 and there are 2 main export
/usr/local (2TB): This export will simplify the management of some specific software.
/data (~ 40TB): This will export user home will be the main storage for data analysis.
The main file on t3ps server is /etc/export:
/usr/local 10.0.0.0/255.0.0.0(rw,async,insecure,no_root_squash)
/data 10.0.0.0/255.0.0.0(rw,async,insecure,no_root_squash)
On the worker nodes, the configuration is done throught autofs
yum install -y autofs
The configuration involve 2 differents file:
/etc/auto.master.d/nfs.autofs: which define a autofs entry point
/etc/auto.nfs: which define the different mount point we will have
/etc/auto.master.d/nfs.autofs:
/nfs /etc/auto.nfs
/etc/auto.nfs:
local t3ps001:/usr/local
data t3ps001:/data
Machine will see the data on /nfs/local (for /usr/local on t3ps001) and /nfs/data (for /data on t3ps001).