DOCUMENT ID:  1128-02

SYNOPSIS:     Gathering Network Integrity Data From NFS

OS RELEASE:   

PRODUCT:      Solaris x86

KEYWORDS:     network integrity data NFS


DESCRIPTION:

Using nfsstat to gather network integrity data from NFS.


SOLUTION:

Systems that are running NFS can get useful data about network
corruption by using nfsstat -c, which reports the system's client- side
NFS statistics (i.e., statistics about NFS requests that this system
originated).  Here's a typical report:

Server rpc:
calls     badcalls  nullrecv  badlen   xdrcall  dupchecks  dupreqs   
0         0         0         0        0        0          0         

Server nfs:
calls      badcalls  
0          0         
Version 2: (0 calls)
null      getattr   setattr   root     lookup   readlink   read      
0 0%      0 0%      0 0%      0 0%     0 0%     0 0%       0 0%  
wrcache   write     create    remove   rename   link       symlink
0 0%      0 0%      0 0%      0 0%     0 0%     0 0%       0 0%  
mkdir     rmdir     readdir   statfs    
0 0%      0 0%      0 0%      0 0%      

Client rpc:
calls     badcalls  retrans   badxids   timeouts  waits   newcreds
75094     1         14        4         14        0       0      
badverfs  timers    toobig    nomem     cantsend  bufulocks 
0         62        0         0         0         0         

Client nfs:
calls      badcalls   clgets     cltoomany 
73916      1          73916      0         
Version 2: (73916 calls)
null      getattr   setattr   root   lookup     readlink read
0 0%      27493 37% 246 0%    0 0%   12494 16%  4 0%     18041 24%
wrcache   write      create    remove   rename   link     symlink
0 0%      13696 18%  338 0%    313 0%   80 0%    3 0%     0 0%
mkdir      rmdir      readdir    statfs
11 0%      0 0%       505 0%     692 0%

The retrans field (under Client rpc) indicates the number of packets
that this host had to retransmit as an RPC client: that is, the number
of retransmissions it made while reading or writing an NFS file.  If
this field is greater than 5 percent of the total number of client NFS
calls, suspect trouble.  Compare the number of retransmissions to the
badxid field.  If badxid and retrans are roughly equal, one or more of
the network's NFS servers is having trouble keeping up with the client's
demands.  This isn't a data corruption problem, but a performance
problem on the part of the server.  If retrans is high but badxid is
relatively small or zero, the problem is in the network itself - the
network is either slow or suffering from data corruption.  In this
example, retransmissions are only about 0.3 percent of total number of
RPC calls, an acceptably low figure. 

nfsstat can do one thing that netstat lacks: it lets you zero the
counters at any time.  Therefore, you don't have to have a look at data
gathering over the last month - it's easy to collect data for day or so. 
To zero the counters, become root and give the command nfsstat -z.  It
may be useful to zero the nfsstat counters regularly.  To zero the
counters at 2 a.m.  daily, add one of the following entries to your
system's cron files:

0 2 * * *      nfsstat -z # add to root crontab file.


DATE APPROVED: 10/10/95