Solaris 10 11/06 on an Ultra 60 with 2x360Mhz cpus, pretty much a server install, not running much of anything yet.
Tried to newfs a 4GB 10k rpm disk on the same bus as the boot drive, which is the same kind of disk. Newfs just hangs, and takes about 30 minutes to finish (date; newfs /dev/rdsk/c0t1d0s0;date -- I answered "y" to the prompt from newfs right away) .
There happens to be a ZFS raidz volumn on drives on the other SCSI bus, not doing anything. Network interface is "down", it's not plugged in to anything.
Anyone have any clues why this should be? Bad enough the bid delays in the format command after selecting a disk, but this is ridiculous!
More info -- when there is no network connection, only! When I ifconfig hme0 down, newfs hangs. When I ifconfig hme0 up, newfs starts doing it's thing right away, no delay. Hme0 is plugged into an otherwise empty switch, nothing else on this system's network.
Question is, what network dependency causes this problem with newfs? I cannot think of any good reason why newfs should care whether the network interface is up or not.
Somebody must know something... please advise.
is NFS running? or is the machine maybe running in a NIS domain?
What I'm trying to get at is that it sounds like some external naming or maybe authentication service is being consulted successfully and returning when the link is up, but something else is consulted internally which is not responding when you are not hitting the network.
what does your /etc/nsswitch.conf look like?
how about looking at /var/adm/messages and
to inspect if you're getting any hardware errors that might be causing timeouts
sorry to have done a reply, to a reply, to a xx month old thread - normally, that's one of my dislikes...
Message was edited by: Jeffrey.s
Do you guys use mpxio and Hitachi arrays? Then maybe you hitting the bug below:
Bug ID: 6546175
Synopsis: Format is taking a long time in ioctl() when mpxio is enabled on Solaris 10
newfs also uses ioctl calls so it's relevant.
Basically, round robin load balancing scheme doesn't work and customization are required in scsi_vhci.conf. You can try the workaround but unfortunately it didn't work for me.