I recently built a new server for the staff here to use. Previously we'd been running a PII/333 with a handful of SCSI disks, running Samba and NFS to export it's disks to Win NT/9x machines and Solaris machines. It was based on Debian 1.3 and kernel 2.0.36. A very stable machine, but it had always had problems with the NFS to the Solaris machines. So much so that we stuck an external RAID server on one of Solaris machines to act as filespace for the Suns.
I recently built a replacement for it based on a twin Xeon III box, lots of memory, dual channel UW SCSI, Debian 2.1 and kernel 2.2.6. It's using software RAID 5 over 4 spindles. It still seems to suffer from the same old NFS problems with the Suns though - directories going read-only, symbolic links vanishing. Has anyone else ever seen this? Is there a cure? NFS to other Linux workstations seems to be ok, it's just to & from the Suns that seems to be a problem (And running Linux on the Suns isn't an option, alas!!!) Another problems I've had is with ypbind. Twice now in the 2 weeks it's been running, ypbind has done something really strange. It runs as the NIS master for the network, so when it croaks, it has a bad knock-on effect to the rest of the system. Re-starting it doesn't seem to help, but reboot does. Anyone else seen this? I never saw any problems on the old server. Very strange. Any help or advice would be appreciated! Thanks, Gordon
