File System naming Standard

[Tony Williams]

Does anyone have a way to name file sytems
so that they are completely portable among different servers?

We would like to start naming our file systems so that in the event of a disaster, we may recover the file systems to a different remote located server. However today most of the file systems have the same name among different servers. e.g. /opt/db01, /opt/db02, /opt/db03
is a common naming standard for Oracle database file systems. It would be hard to mount /opt/db01 to a different server.

you can't use hostname, or serial number in the file system name, becuase they change frequently. You can't assign the project name because it may change (/opt/oracle-apps/db01).

Anyone have any ideas on this?

[Patrick Wallek]

At my previous job, one of the sys admins responsible for some of our Oracle DB machines kept a list of filesystem names and went down the list whenever a new filesystem was created on ANY machine. All filesystems had a unique number and it was not duplicated on ANY other machine in the company.

So machine A may have: /opt/db01, db02, db03 and db12

Machineb: db04, db05, db06, db07 and db13

Machinec: db08, db09, db10, db11, db14

The file systems do not have to be in order on each server. db12 on machinea was created after machineb and machinec were setup with their filesystems.

Hopefully you get the idea.

[Pete Randall]

Tony,

How are you going to be moving these filesystems? If you're using vgexport/vgimport you can rename them during the import. Or, if they're mounted, mount them to a different mount point.

Just a thought,
Pete

[Christopher Caldwell]

We assign IPs to our MC/ServiceGuard packages - IPs/packages are named {somewhat} similar to the function they perform - perhaps with a serial number.

So we end up with things like

/opt/ns-home.pkg1
/opt/ns-home.pkg2

where the suffix is guaranteed to be unique among all of our systems.

We've done similar things with databases, webservers, and web filesystems.

[Sajid_1]

hello,

there is a standard naming convention for logical volumes (lvol1, lvol2 ..etc), but not for the file systems. it's all depend on your need and wish. The file system name (actual mount point) doesn't matter when you move the data from one system to another. But in this case the LV names matters. If you want to keep the old LV names, you need to have a map file. Otherwise during a vgimport/export, the new system will accept only standard naming conventions for LV. In either case you can have any name for the file systems. I keep my file system name as:

/db1, db2 ..et

[Tony Williams]

Thanks,

I like the idea. We have /mnt/data1
... /mnt/data2 as a standard file system name. We could just continue on and make /mnt/data1
... /mnt/data50.

We could also make a standard /mnt/data
and then for each server create a /mnt/data/a01
/mnt/data/b01, /mnt/data/c01.

[Martin Johnson]

We attach a numeric suffix on the file name. So the first install of pgp would go to /var/pgp01. The 2nd to /var/pgp02, etc. We also name the lv as the file name lv_pgp01 would be mounted on /var/pgp01.

HTH
Marty

[Michael Tully]

I think that logical volumes and filesystems can be named anything you wish as long as they are easily identified or documented and this includes entries in /etc/fstab.
For example if you using raw logical volumes like we do, we have a spreadsheet for each LUN that is used and for what group.
For mounted filesystems we use a convention of who uses what particularly for development systems.

/dev/apps/weblogic mounted on /weblogic for weblogic
/dev/apps/informix mounted on /informix " "

/dev/data1/rootdbs
/dev/data1/tempdbs

etc for raw volumes.