burzio@mmlai.UUCP (Tony Burzio) (06/05/89)
I noticed something strange with the new bdf command. When you enter the command from the ROOTSERVER, you get the correct list: $ bdf Filesystem kbytes used avail capacity Mounted on /dev/dsk/0s0 241215 183478 33615 85% / /dev/dsk/2s0 290754 99450 162228 38% /disk2 /dev/dsk/1s0 290754 140697 120981 54% /disk1 mmlai:/users 291606 251770 10676 96% /mmlai/users However, when you are on a diskless cnode (with swap disk) you get the following: $ bdf Filesystem kbytes used avail capacity Mounted on /dev/dsk/0s0 -637229 -514041 1312557-21364% / /dev/dsk/2s0 290754 92459 169219 35% /disk2 /dev/dsk/1s0 290754 140695 120983 54% /disk1 mmlai:/users 291606 251770 10676 96% /mmlai/users Most strange. My guess is that since my swap disk on the diskless node is also 0s0, bdf is getting confused. The df command works on both. Is this a "feature"? While I'm on the topic, anybody at HP know when you will be able to put a user disk on a satellite node? ********************************************************************* Tony Burzio * A' swatting we will go... Martin Marietta Labs * Take that you bug! Hee'yah! mmlai!burzio@uunet.uu.net * *********************************************************************
johnson@hpcilzb.HP.COM (Phil Johnson) (06/06/89)
> I noticed something strange with the new bdf command. ... > > However, when you are on a diskless cnode (with swap disk) you get > the following: > > $ bdf > Filesystem kbytes used avail capacity Mounted on > /dev/dsk/0s0 -637229 -514041 1312557-21364% / > /dev/dsk/2s0 290754 92459 169219 35% /disk2 > /dev/dsk/1s0 290754 140695 120983 54% /disk1 > mmlai:/users 291606 251770 10676 96% /mmlai/users I believe this to be a bug with bdf. I formally reported it to the HP-UX development group a while back so I expect it will be resolved in a future release. You are correct that this problem only occurs on cnodes with local swap space, when the /dev/dsk/ directory is present on the cnode. Try deleting the cnode's /dev/dsk/ directory. You probably don't refer to this directory anywhere anyway, because /etc/checklist isn't necessary to find the swap disk on the cnode. This has solved the problem here. The /dev/rdsk/ directory doesn't seem to cause any problems. Phil
rml@hpfcdc.HP.COM (Bob Lenk) (06/07/89)
> Try deleting the cnode's /dev/dsk/ directory. I think that is sufficient but not necessary. All that appears to be necessary is to avoid any name collisions between device files on the client and those naming mounted file systems on the server. Thus the case of the posted output, the device file on the client could be named anything other than /dev/dsk/[012]s0. Another workaround is to use remsh to execute bdf on the server. > While I'm on the topic, anybody at HP know when you will be able to > put a user disk on a satellite node? This is being considered for a future release, but there are no committed plans yet. Bob Lenk rml@hpfcla.hp.com hplabs!hpfcla!rml