HAStorage or HAStoragePlus not needed with SC3.1, naking NFS HA?
Greetings experts,
some time ago I participated in a SC3.0 (yes, that's quite some time ago) training. Besides other, we learnt how to make NFS HA. In brief we constructed a triple ("a resource group) consisting of 3 resources: the logical host, the NFS application and the HAStorage Resource:
scrgadm -a -g <rg> -j <has-res> -t SUNW.HAStorage -x ServicePaths=... -x AffinityOn=true
This week I setup an SC3.1 NFS HA failover service, (SC3.1 8/05 / S10). I worked using this document: http://docs.sun.com/app/docs?q=817-6999-10 . Pages 21 and 22 give an example how to setup and configure Sun Cluster HA for NFS. Interestingly there isno mention of HAStorage or HAStoragePlus on these pages. Only p. 25 and p. 26 speak then of HAStoragePlus.
Therefore I wonder if anyone could comment upon these two questions:
1. Isn't it a must anymore to include HAStorage / HAStoragePlus anymore into such a resource group?
2. If it is still a must (what I believe) can you provide an example for HAStoragePlus (such as the one I provided for HAStorage) please? It appears as if HAStoragePlus does not know about i.e. "ServicePaths"
Thanks
[1197 byte] By [
hontoya] at [2007-11-27 7:30:07]
# 1
Hi,
as long as you want to create a failover file system, the HAS+ resource is a must.
With a global file system (pxfs) HAS+ is a can, but it is highly advisable to create a HAS+resource. If your resource group with the NFS server will fail over, the HAS+ resource will take the device group to the other node. Without it, every file access from th NFS server will travel through the interconnect to devicegroup primary and you will see a performance impact.
You should use the HAS+ istead of HAS resources
On http://docs.sun.com/app/docs/doc/819-0703/6n343k6i3?a=view There is a full fledged example for HANFS, instead of ServicePaths, you use Filesytemmountpoints
scrgadm -a -j nfs-hastp-rs -g nfs-rg -t SUNW.HAStoragePlus \
-x FilesystemMountPoints=/global/local-fs/nfs \
-x AffinityOn=True
Detlef
# 2
Hi Detlev,
many thanks for your feedback.
- Do you have any idea (or maybe insight) of why the document I referred to explains how to setup the cluster without HAStoragePlus (or HAStorage, respectively)?
-And taking the example you referred to into account: Am I right to understand that "global/local-fs/nfs" within -x FilesystemMountPoints=/global/local-fs/nfs is that what is contained in the vfstab?
- There is obviously also the possibility to refer to GlobalDevicePaths: do I here need to specify my metaset name?
- And in general: is FilesystemMountPoints implicitly implicating that I am using a filesystem to run I/O while GlobalDevicePaths implicitly implicating that I am using raw I/O?
Thanks Kind regards hontoy
# 3
Hi,
Tehre is the general possibility to set it up withot, althoug it is defienitly not advisable.
You are right /global/local-fs/nfs should be in the vfstab.
In th eglobal devicepath you have to put in the metaset name which is the same as the deviecgroup name.
This implication is not true, I personally would use it in this way, because my resource parameters will be checked against the vfstab if you use Filesysttemmountpouints.
Kind Regards
Detlef
