I never cared much about GPFS disk names so all of our disks are named like "gpfs1200nsd" thus i let GPFS determine the final disk name.
But now the questions comes up: Must be a GPFS disk name unique in a multi cluster env. ?
Meaning a nsd named: gpfs1200nsd is not allowed to exist on 2 different cluster in case they build a multi cluster env. where both cluster mount the others GPFS ?
NOTICE: developerWorks Community will be offline May 29-30, 2015 while we upgrade to the latest version of IBM Connections. For more information, read our upgrade FAQ.
This topic has been locked.
7 replies Latest Post - 2013-02-11T12:02:14Z by SystemAdmin
Pinned topic GPFS disk names in a multi cluster environment ( must be unique or not )
Answered question This question has been answered.
Unanswered question This question has not been answered yet.
Updated on 2013-02-11T12:02:14Z at 2013-02-11T12:02:14Z by SystemAdmin
dlmcnabb 120000P4JT1012 Posts
Re: GPFS disk names in a multi cluster environment ( must be unique or not )2013-02-11T08:48:30Z in response to dlmcnabbthx for the clarification.
GPFS uses "gpfs" as the prefix for its NSD disks - is it possible to set a certain variable that GPFS will use another prefix ? Since i see no need to maintain the disk number but i could create "unique" disk names even across clusters
SystemAdmin 110000D4XK2092 PostsACCEPTED ANSWER
Re: GPFS disk names in a multi cluster environment ( must be unique or not )2013-02-11T09:28:31Z in response to HajoEhlersI don't think there is a parameter to change the generic prefix GPFS uses. But you can supply your own names when creating NSDs. I routinely name NSDs myself with names reflecting architecture of the system:
root@host# mmlsdisk gpfsFS1
disk driver sector failure holds holds storage
name type size group metadata data status availability pool
R1_GPFS_01 nsd 512 1 yes yes ready up system
R1_GPFS_02 nsd 512 1 yes yes ready up system
R1_GPFS_03 nsd 512 1 yes yes ready up system
R1_GPFS_04 nsd 512 1 yes yes ready up system
R2_GPFS_01 nsd 512 2 yes yes ready up system
R2_GPFS_02 nsd 512 2 yes yes ready up system
R2_GPFS_03 nsd 512 2 yes yes ready up system
R2_GPFS_04 nsd 512 2 yes yes ready up system
Q1_GPFS_01 nsd 512 3 no no ready up system
R1 = Storage subsystem in rack 2
R2 = Storage subsystem in rack 2
Q1 = Local quorum / tie breaker disk on quorum node
I do name the LUNs the same in the storage controller (if you can name them). For example DS8000 controllers assign an unique number to a LUN, so I'd call a NSD something like DSxx_abcde, where ABCDE is the LUN id.
I find it important that it is easy to identify LUNs.
Re: GPFS disk names in a multi cluster environment ( must be unique or not )2013-02-11T10:50:08Z in response to SystemAdminHi Markus,
> I find it important that it is easy to identify LUNs.
Me too, for this reason i wrote a script that gives me all information i am looking for.
I am not a fan of "speaking" names since in case something changes in the background the name is not valid anymore.
For example we moved luns on our EMC storage from one EMC box to another. In case the NSD name contains the storage array we are out of sync.
Another example: We moved a LUN on a given EMC from FC storage to SATA. In case the NSD name contains the storage type we are out of sync.
For this reason i wrote a small script which gives me all required information about my luns (AIX).
Also nice to find not assigend resources ;-)
disk vg/gpfs nsd(gpfs) type pvid size(GB) lun comment hdisk0 rootvg n/a scsd 0006ba4aea392253 68G n/a hdisk79 gpfs1 gpfs1196nsd nexsan 00c07045b0f8e75d 20G 0x0 hdisk163 gpfs2 gpfs1197nsd nexsan 00c07045b0f8cd50 20G 0x1000000000000 hdisk343 None n/a nexsan 00c07045b0f8dd69 20G 0x9000000000000 hdisk347 FREE gpfs1201nsd nexsan 00c07045b0f8e445 5120G 0xd000000000000 hdiskpower0 gpfs3 gpfs1096nsd power 00589a3d30ce6e7a 978G 0x0 EMC,tc_tccore6_sg,tcgpfsc1_1_LUN0
Having a selfdefined prefix would be nice for a future case of merging clusters.
In case i could have a self defined prefix , the prefix would be related in my case to the gpfs cluster name. Our cluster names or GPFS are not starting with GPFS so the probability of having 2 disk with the same name is very unlikely.
Srikanth.S 060000PX4P3 PostsACCEPTED ANSWER
Re: GPFS disk names in a multi cluster environment ( must be unique or not )2013-02-11T09:47:35Z in response to HajoEhlersYou need to specify a unique disk name that confirms to the rules to have a nsd name of choice. If its not specified gpfs will use the default naming convention 'gpfsNNNsd'. This convention cannot be modified (nor can a prefix of choice be specified). Please refer the following link for a more detailed explanation on mmcrnsd command and its options.
Specify the name you desire for the NSD to be created. This name must not already be used as another GPFS disk name, and it must not begin with the reserved string 'gpfs'.
Note: This name can contain only the following characters: 'A' through 'Z', 'a' through 'z', '0' through '9', or '_' (the underscore). All other characters are not valid.
If this clause is not specified, GPFS will generate a unique name for the disk and will add the appropriate nsd=NsdName clause to the stanza file. The NSD is assigned a name according to the convention:
where NN is a unique nonnegative integer not used in any prior NSD.
Re: GPFS disk names in a multi cluster environment ( must be unique or not )2013-02-11T10:23:40Z in response to Srikanth.SThx, I am aware of the man page.
> ... nor can a prefix of choice be specified.
That's not in the man page....
A quick look in the /usr/lpp/mmfs/bin/mmcrnsd shows that the prefix is hardcoded.
Would be nice if this limitation could be removed. ( No, i do not open a DCR )
SystemAdmin 110000D4XK2092 PostsACCEPTED ANSWER
Re: GPFS disk names in a multi cluster environment ( must be unique or not )2013-02-11T12:02:14Z in response to HajoEhlersHajo, I agree that overly descriptive names have the nasty tendency drift away from reality. So I like simple things like storage device number and lun number.
Besides the limitation that you can not give your own prefix (with the workaround that you can specify the full name) I found it more annoying that you are not allowed to specify a name starting with gpfs. This means after moving luns around a lot you'll have 'holes' in your numbering scheme you can not fill because the automatic number just increase and you can not manually specify a name starting with 'gpfs'.