home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.hp
- Path: sparky!uunet!charon.amdahl.com!pacbell.com!sgiblab!sdd.hp.com!scd.hp.com!hpscdm!cupnews0.cup.hp.com!pasq
- From: pasq@cup.hp.com (Mark Di Pasquale)
- Subject: Re: Mail in 710 cluster with NFS fails
- Sender: news@cupnews0.cup.hp.com
- Message-ID: <BxBKGw.3Mn@cup.hp.com>
- Date: Sat, 7 Nov 1992 00:10:08 GMT
- References: <1992Oct22.012738.25751@state.systems.sa.gov.au>
- Organization: Hewlett-Packard
- X-Newsreader: TIN [version 1.1.4 PL6]
- Lines: 81
-
- enisaxg@state.systems.sa.gov.au wrote:
- : Problem with Mail under a common NFS mounted /usr/mail.
- : ...
- :
- : We have no problem running any mailers between the CDC, SUN, and HP 847.
- : We do, however, have 5 HP 710's running as a cluster off of one server.
- :
- : On the HP 710 *server*, there is no problem accessing the NFS common maildrop.
- :
- : On *clients* of the HP 710, there is no way we can access the NFS common
- : maildrop! This appears to be related to the fairly proprietary nature of the
- : cluster concept. The errors we get from the clients are:
-
- This may have something to do with "context dependent files". If the
- clients are not gaining access to some resource the server has, it may
- be that the "resource" does not have a "context" for the clients. For
- example, on your server system, /etc/inittab is a context-dependent file
- (CDF). If you were to "cd /etc/inittab+" and do an ls(1), you should
- see the names of all of your clients, and the server's name (or
- something like "localroot", which means "the rootserver"). In this
- case, each "name" listed is a unique file, one for each machine in the
- cluster. But the name is important!! For example, if you have a client
- named hptm1 listed in /etc/inittab+, and you were to "mv
- /etc/inittab+/hptm1 /tmp", then client hptm1 would no longer be able to
- see /etc/inittab:
-
- hptm1: ls /etc/inittab
- /etc/inittab not found
- hptm1:
-
- Now, if the problem you are experiencing with /usr/mail relates to
- context, then you have a CDF file or directory that does not have the
- names of the clients listed (or the key-word "remoteroot" which means
- "all machines in the cluster that are not the rootserver").
-
- More specifically, when you are accessing the "common mail drop" from
- your server, I assume you have a directory/file on the server machine
- that "points" to the common drop? For example, you may have done an NFS
- mount from the server system like this:
-
- CDC:/usr/mail /usr/mail nfs ro,soft 0 0
-
- Check to see if /usr/mail (or some other involved file) is a CDF. You
- can do this by: "cd /usr; ls -H mail". If you see mail listed as
- "mail+", then change directories into mail+ and list the context(s).
-
- Let's say mail is a CDF. Then, you may find your server's name or the
- keyword "localroot" in there:
-
- /usr/mail+ or /usr/mail+
- \ \
- localroot server_name
-
- Basically, what would have to be done is to make a context for your
- clients. You may be able to do this (in this example) by:
-
- ln -s localroot remoteroot
-
- creating:
-
- /usr/mail+
- /\
- remoteroot localroot
-
- I can't guarantee any of this because I don't know where your CDFs are,
- or even if this is a CDF-related problem. But this may help.
-
- See also: context(5), getcontext(1), and cdf(4)
-
- --
- Regards,
- Mark DiPasquale
-
- _______ ________________________________________________________________
- /
- / All appropriate disclaimers apply...
- ____ ___ /
- / / / /
- __/ __/ _____/
- /
- ___________ __/ __________________________________________________________
-