Another Words: if any of the process open a file in write mode
no other process will able to write it at the time.
PARTICLE EXAMPLE: if same file open by the two process,when second process open file you will get warning message like below in read mode.
GFS > Uses Lock Manager > Synchronize access to the file system.
CLVM > Uses Lock Manager > Synchronize update to the LVM and volume group.
Rgmanager > Uses Lock Manager > Synchronize the service states.
Synchronized access to the shared resources.
Advantage: Avoid inconsistency.
CLVM & CMAN will interact background for sharing cluster resources & information.
Comparison Between LVM, High Availability LVM , Cluster LVM
| LVM | HA- LVM | Cluster LVM |
LVM -> Available only on single node | LVM in Clustering failover mode. Active Passive mode. | LVM -> Available across the multiple cluster nodes. |
| No other service need to be run. | No other service need to be run. | Clvmd -> Service should run across all the cluster nodes. It will distribute the metadata in the cluster and provide the same logical volume to all the cluster nodes. NOTE: All the functionality like create/reduce similar to the LVM except clvmd service. |
/etc/lvm.conf Locking_type=1 (File Based Locking) | /etc/lvm.conf Locking_type=1 (File Based Locking) | /etc/lvm.conf Locking_type=3 (Cluster Locking) |
Comparison Between GFS and Lustre.
| GFS(Global File system) | LUSTRE |
| 1)Shared File system | Distributed File system |
| 2) Maintain the High availability(or) Load Balancing more than two nodes. | High Performance Computing Environment. |
| 3) Allow all nodes to have concurrent access to the same shared block storage.(no client or server models,All node in GFS cluster function as peers). | Have server/client Communication setup in the cluster. |
Global File System Installation
Step By Step CLVM Configuration
1)[root@node11 ~]# /etc/init.d/clvmd start
Starting clvmd:
Activating VG(s): 0 logical volume(s) in volume group "vgcluster" now active
2 logical volume(s) in volume group "vgcluster3" now active
[ OK ]
2)[root@node11 ~]# pvcreate /dev/mapper/1IET_00020001p1
Physical volume "/dev/mapper/1IET_00020001p1" successfully created
3)[root@node11 ~]# pvs
PV VG Fmt Attr PSize PFree
/dev/mapper/1IET_00020001p1 lvm2 a-- 2.00g 2.00g
Whatever create physical it will automatically reflect in the node10 (Remaining Cluster node) also.
How it is happens because of the clvmd service.
4)[root@node11 ~]# vgcreate vgcluster /dev/mapper/1IET_00020001p1
Clustered volume group "vgcluster" successfully created
5)[root@node11 ~]# vgs
VG #PV #LV #SN Attr VSize VFree
vgcluster 1 0 0 wz--nc 2.00g 2.00g
6)[root@node11 ~]# lvcreate -L +500M -n lvcluster vgcluster
Logical volume "lvcluster" create
7)[root@node10 ~]# mkfs.gfs2 -j 2 -plock_dlm -t MY_CLUSTER2:mygfs2 /dev/vgcluster/lvcluster-gfs
-j <Journling>
-p <name of the locking protocol>
This will destroy any data on /dev/vgcluster/lvcluster-gfs.
It appears to contain: symbolic link to `../dm-4'
Are you sure you want to proceed? [y/n] y
Device: /dev/vgcluster/lvcluster-gfs
Blocksize: 4096
Device Size 0.49 GB (128000 blocks)
Filesystem Size: 0.49 GB (127997 blocks)
Journals: 2
Resource Groups: 2
Locking Protocol: "lock_dlm"
Lock Table: "MY_CLUSTER2:mygfs2"
UUID: afd06add-4c31-640e-488e-db057d2b42e4
Solution And Error Message
# mkfs.gfs2 -j 2 -p lock_dlm -t MY_CLUSTER2:mygfs2 /dev/vgcluster/lvcluster-gfs
This will destroy any data on /dev/vgcluster/lvcluster-gfs.
It appears to contain: symbolic link to `../dm-5'
Are you sure you want to proceed? [y/n] y
Not enough space available on device
Solution: While Creating the lvm Give around 500MB then the problem got solved.
0 Comments