On Fri, Oct 23, 2009 at 2:27 PM, Nils Goroll <slink at schokola.de> wrote: > Hi, > > today, I've come across a configuration which I would consider at least > "unusual": Someone has configured a metaset and is using a global/pxfs mount > on a slice of the (only) disk in the set: > > nodea # metastat -s share -p > nodea # metaset -s share > > Set name = share, Set number = 1 > > Host ? ? ? ? ? ? ? ?Owner > ?nodea ? ? ? ? ? ? Yes > ?nodeb > > Driv Dbase > > d5 ? Yes > > nodea # grep shared /etc/vfstab > /dev/did/dsk/d5s0 /dev/did/rdsk/d5s0 /shared ufs 5 yes global,logging > > d5 is not the quorum disk (at least...). > > I am tempted to argue that this is unsupported because, conceptually, the > metaset owns its disks, so they should not be accessed from outside SVM. > Also, an administrator could assume the set to be unused and create an SVM > object on the disk which is already in use - and SVM has no knowledge of the > file system on it. > > But despite my concerns, this configuration actually seems to work - or at > least I have not yet found how the case during normal cluster operation > where it fails.
I suppose if you start creating metadevices/volumes on the metaset there will be trouble. You are essentially bypassing the SVM layer by accessing a component disk directly. What SVM would put where on each disk cannot predicted by UFS (on the individual disk). cheers Binu > Can anyone make a definite statement on this configuration? > > Thanks, Nils > > _______________________________________________ > ha-clusters-discuss mailing list > ha-clusters-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/ha-clusters-discuss >
