cancel
Showing results for 
Search instead for 
Did you mean: 

linux cluste

Prashanth Waugh
Esteemed Contributor

linux cluste

Dear all,

we are building two node cluster using oracle linux. On one node I have craeted the cluster file systems.Now How to make it visible to another node like in HP service gurad we will do vgexport/import the map file.

Thanks,
Prashant
For success, attitude is equally as important as ability
8 REPLIES
Matti_Kurkela
Honored Contributor

Re: linux cluste

In Linux, there is no such thing as a map file.

First, make sure both nodes can see the actual disks/LUNs at the hardware level.

If necessary, see RedHat's Online Storage Reconfiguration Guide form making the disks/LUNs visible without a reboot. Most of the techniques within it are applicable to all Linux distributions, not just RedHat.

https://www.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/html/Online_Storage_Reconfiguration_Guide/index.html

If you used traditional PC partition tables, use "partprobe" (if available) or "sfdisk -R /dev/" to make the kernel detect the partitions.

If you used LVM, run "vgscan". Then see what is required by your clustering software.

Non-cluster filesystems like ext3 will become corrupted if mounted on two or more nodes at the same time, so a clustering software will usually have some sort of a locking mechanism to make sure it does not happen by accident.

Cluster filesystems like OCFS or GFS can be mounted by multiple nodes simultaneously, but the nodes will usually need to communicate with each other over the network to coordinate their operations. If you're mounting a cluster filesystem, make sure this communication works before actually mounting the filesystem on the 2nd node.

MK
MK
Prashanth Waugh
Esteemed Contributor

Re: linux cluste

Hi Matti,

In hp unix we will do the vgexport/import on adotive node . How we need to do it linux so that file system will be available on adoptive node when failover occur.

Regards,
Prashant
For success, attitude is equally as important as ability
Matti_Kurkela
Honored Contributor

Re: linux cluste

Run "vgscan -vv" and read the output to verify that the system sees the necessary VG(s). If that indicates the VG is visible and there are no errors, the VG is ready for activation.

You might also use the commands "pvs", "vgs" and "lvs" to verify the PV/VG/LV visibility.

In Linux with 2.6.* kernel, the device files for the LVs will be created automatically when you activate the VG, not before.

MK
MK
Ivan Ferreira
Honored Contributor

Re: linux cluste

>>> we are building two node cluster using oracle linux

Is this a RAC Cluster? If so, probably would be better if you use ASM or OCFS2. In this case, LVM should be avoided.

If not, ¿what cluster software are you using?
Por que hacerlo dificil si es posible hacerlo facil? - Why do it the hard way, when you can do it the easy way?
Prashanth Waugh
Esteemed Contributor

Re: linux cluste

Hi ,

we are using the oracle clusterware.

Regards,
Prashant
For success, attitude is equally as important as ability
Ivan Ferreira
Honored Contributor

Re: linux cluste

If you are using the oracle clusterware, then you should go for using ASM/OCFS2 as repository for your database. ASM/OCFS2 allows you simultaneous access from both nodes. ASM/OCFS2 don't use LVM.
Por que hacerlo dificil si es posible hacerlo facil? - Why do it the hard way, when you can do it the easy way?
Prashanth Waugh
Esteemed Contributor

Re: linux cluste

Hi Ivan,

For creating the ocfs2 first we need to create the
1) pv
2) vg
3)lv
then need to use ocfs2 file sytem for creating the file system.
I think in linux no need to do the export/import. after doing vgscan on secondary node the vgs will be visible, Is it correct.
-----------
Using ASM, Could you please share the the document if you have related to form the cluster using ASM

Thanks,
Prashant
For success, attitude is equally as important as ability
Ivan Ferreira
Honored Contributor

Re: linux cluste

For creating the ocfs2 first we need to create the
1) pv
2) vg
3)lv

That is incorrect, you create the OCFS file system direct over the "RAW" device. This should be a persistent name defined with udev for example. Both nodes should have the same device name. You won't create logical volumes on these devices, just a partition and then the file system.

ASM is a DB administrator duty. You only give to him the device names (No lvm again).
Por que hacerlo dificil si es posible hacerlo facil? - Why do it the hard way, when you can do it the easy way?