Subject: Re: NFS questions ..
To: Stephen M. Jones <smj@cirr.com>
From: Roland Dowdeswell <elric@imrryr.org>
List: port-alpha
Date: 01/08/2003 14:26:36
On 1042012366 seconds since the Beginning of the UNIX epoch
"Stephen M. Jones" wrote:
>

>2) How do you decide how many severs to spawn?  number of exported filesystems
>   or number of clients?

Generally, number of clients that are expected to concurrently
access I think.  But, each client may have multiple outstanding
accesses.  So, what I do is start, say, 16 of them and then use
the system normally for a while.  After that I'll do a:

$ ps aux | grep nfsd
root      248  0.0  0.1  112   768 ?? IL   Sun11AM  0:00.05 nfsd: server 
root      247  0.0  0.1  112   768 ?? IL   Sun11AM  0:00.05 nfsd: server 
root      246  0.0  0.1  112   768 ?? SL   Sun11AM  0:00.04 nfsd: server 
root      245  0.0  0.1  112   768 ?? SL   Sun11AM  0:00.04 nfsd: server 
root      244  0.0  0.1  112   768 ?? SL   Sun11AM  0:00.04 nfsd: server 
root      243  0.0  0.1  112   768 ?? SL   Sun11AM  0:00.04 nfsd: server 
root      242  0.0  0.1  112   768 ?? SL   Sun11AM  0:00.03 nfsd: server 
root      241  0.0  0.1  112   768 ?? SL   Sun11AM  0:00.03 nfsd: server 
root      240  0.0  0.1  112   768 ?? SL   Sun11AM  0:00.03 nfsd: server 
root      239  0.0  0.1  112   768 ?? SL   Sun11AM  0:00.02 nfsd: server 
root      238  0.0  0.1  112   768 ?? SL   Sun11AM  0:00.02 nfsd: server 
root      237  0.0  0.1  112   768 ?? SL   Sun11AM  0:00.02 nfsd: server 
root      236  0.0  0.1  112   768 ?? SL   Sun11AM  0:00.36 nfsd: server 
root      235  0.0  0.1  112   768 ?? SL   Sun11AM  0:00.58 nfsd: server 
root      234  0.0  0.1  112   768 ?? SL   Sun11AM  0:01.24 nfsd: server 
root      233  0.5  0.1  112   768 ?? SL   Sun11AM  9:43.84 nfsd: server 
root      230  0.0  0.0  136     0 ?? Ss   Sun11AM  0:00.14 nfsd: master 

and look at the times that each one has logged up.  As you'll
notice, most of them are rather low which is good.  If all of them
are high, then you need more.  :-)

>The problem I'm seeing is sometimes an NFS server will not be responding
>and then come back .. sometimes the 1.6 client hangs after 3 or so days
>of uptime while the 1.5.x client keeps going .. sometimes I see nfsd send
>error 55 on the server.. many times I see
>nfs server sdf1:/sys: not responding
>nfs server sdf1:/sys: is alive again
>uvn_attach: blocked at 0x0xfffffc0020b41ba8 flags 0x4
>uvn_attach: blocked at 0x0xfffffc000b926de0 flags 0x4

I think that there are some 1.5.x era problems that are in the bug
db.  The ones that I've seen are kern/13633 and kern/11618.  It
affects some of the 1.5 branch.

Also, error 55 is ENOBUFS which suggests that there are too many
outstanding requests or some such.

--
    Roland Dowdeswell                      http://www.Imrryr.ORG/~elric/