lockd woes on Solaris 2.4

lockd woes on Solaris 2.4

Post by Simon Bur » Wed, 20 Sep 1995 04:00:00

We've got a Sparc 1000 running Solaris 2.4 as an NFS server, and about
100 clients running Ultrix 4.4.  The application we use creates about
20 to 30 NFS locks (maybe more, maybe less) per workstation.

Quite often we see messages like the following in /var/adm/messages:

Message 1:
lockd[211]: nlm1_reply: RPC: Unknown host
lockd[211]: create_client: no name for inet address 0xC0A80C67

Message 2:
unix: WARNING: File ../../common/os/flock_ops.c, line=787: assertion
failed: candidate->start - lock->end == 1

Message 1 seems only to affect the given workstation.  Doing a reverse
lookup for the IP address works ok.  We've been told that lockd times
out waiting for DNS to respond, and issues the above message.

When we get message 2 however, things seem to have fallen totally
apart, and we need to reboot the Sun.

We were originally using DNS (to an Ultrix nameserver on the same
subnet) for host name and address lookups.  Sun suggested we use a full
/etc/inet/hosts file with all of the NFS clients, but if anything this
had made the situation worse.  We are not running a named on the Sun,
but are considering doing so to see if this makes a difference.

Has anyone else seen problems like this before?  Sun have told us that
2.5 has completely re-architected kernel locking, but this really
doesn't help us now...


UNIX Support, CPR Project, Telstra.                     +61 3 9634 3974
I've got a can opener at home.  Sometimes it works, and other times it doesn't.
Sometimes it only goes half-way around, and then does a lap of honor - Jimeoin.