nfsstat(1M) nfsstat(1M)
NAME
nfsstat - output statistical information about NFS and Remote Pro-
cedure Calls (RPC)
SYNOPSIS
/usr/bin/nfsstat [-cnrszT] [-M [node ...]]
DESCRIPTION
This command outputs statistical information relating to:
- The number of RPCs sent and received and the errors that occurred.
A distinction is made between RPCs, depending on whether the system
is a client or server.
- Number and type of NFS calls and errors that occurred. A distinc-
tion is also made here between statistics for the computer in its
capacity as client or server.
This information can be accessed for RPCs of NFS version 2 and 3.
In addition, the information referred to above can be updated (reini-
tialized) with this command.
The statistics output are valid as of the last updating.
OPTIONS
No option specified:
outputs all statistical information. There is no updating.
-c only outputs information about NFS or RPC relevant to the com-
puter in its capacity as client.
-n only outputs information about NFS.
-r only outputs information about RPC.
-s only outputs information about NFS or RPC relevant to the com-
puter in its capacity as server.
-z initiates the output and subsequent updating (zeroing) of the
statistical information.
The information which is to be updated can be specified in more
detail with one of the above arguments. If no argument is speci-
fied, all the information is updated.
The information on time usage is not updated with this option.
-T Information on time usage is output for every NFS procedure.
-M The -M option is only effective within a MESH network. It outputs
statistical information on the node or nodes named in the option.
Page 1 Reliant UNIX 5.44 Printed 11/98
nfsstat(1M) nfsstat(1M)
If a node is not specified, statistical information on all nodes
is output.
EXAMPLE
You would like all statistical information about NFS and RPC to be
output which is relevant to your computer in its capacity as a server
or a client. Enter the following:
nfsstat
The system responds by displaying e.g. the following:
Server rpc:
calls badcalls nullrecv badlen xdrcall
850889 53 0 0 0
Server nfs version 2:
calls badcalls
593711 0
create fsstat getattr link lookup
6 0% 64536 10% 399705 67% 0 0% 110638 18%
mkdir null read readdir readlink
0 0% 46 0% 10032 1% 8480 1% 228 0%
remove rename rmdir root setattr
4 0% 0 0% 0 0% 0 0% 36 0%
symlink wrcache write
0 0% 0 0% 0 0%
Server nfs version 3:
calls badcalls
257150 0
access commit create fsinfo fsstat
0 0% 0 0% 6 0% 64824 25% 64848 25%
getattr link lookup mkdir mknod
10369 4% 0 0% 105936 41% 0 0% 0 0%
null pathconf read readdir readdirplus
54 0% 0 0% 584 0% 10495 4% 0 0%
readlink remove rename rmdir setattr
3 0% 16 0% 0 0% 0 0% 15 0%
symlink write
0 0% 0 0%
Client rpc:
calls badcalls retrans badxid timeout
139673 57 150 5 207
wait newcred badverf timer toobig
0 0 0 180 0
nomem cantsend bufulock
0 0 22
Page 2 Reliant UNIX 5.44 Printed 11/98
nfsstat(1M) nfsstat(1M)
Client nfs version 2:
calls badcalls nclget nclsleep
139657 41 139657 0
create fsstat getattr link lookup
51 0% 35095 25% 4126 2% 0 0% 92511 66%
mkdir null read readdir readlink
0 0% 0 0% 4486 3% 3185 2% 75 0%
remove rename rmdir root setattr
25 0% 0 0% 0 0% 0 0% 28 0%
symlink wrcache write
0 0% 0 0% 75 0%
Client nfs version 3:
calls badcalls nclget nclsleep
0 0 0 0
access commit create fsinfo fsstat
0 0% 0 0% 0 0% 0 0% 0 0%
getattr link lookup mkdir mknod
0 0% 0 0% 0 0% 0 0% 0 0%
null pathconf read readdir readdirplus
0 0% 0 0% 0 0% 0 0% 0 0%
readlink remove rename rmdir setattr
0 0% 0 0% 0 0% 0 0% 0 2800%
symlink write
0 0% 0 0%
The command outputs the following statistics:
Server rpc
Information on RPC calls received by the computer in its capacity as
server is given under Server rpc, and has the following meaning:
calls Number of RPCs received.
badcalls Number of bad RPCs received.
nullrecv Number of attempts to receive that failed because of
internal memory problems.
badlen Number of RPCs received that are too short.
xdrcall Number of RPCs, for which an acknowledgement could not
be produced.
Client rpc
Client rpc lists all information about RPC calls made by the computer
in its capacity as a client.
calls is the number of RPC calls made.
Page 3 Reliant UNIX 5.44 Printed 11/98
nfsstat(1M) nfsstat(1M)
badcalls is the number of RPC calls still unanswered after execu-
tion of the retrans option [see mount(1M), retrans
option].
retrans is the number of RPC packets within a call which had to
be sent more than once because no acknowledgment or an
incorrect acknowledgment was received.
If the retrans option lists less than 5% of the calls,
that is normal.
badxid is the number of acknowledgments for RPC packets which
arrived after termination of the RPC call.
timeout is the number of RPC packets sent within a call to which
there was no response within a specified time [see
mount(1M), timeo options].
wait is the number of calls where it was necessary to wait
for internal file structures.
newcred is the number of authentication parameters which were
requested again from the server.
badverf is the number of RPCs, for which the authentication
parameters returned by the server were invalid.
NFS statistics
Server nfs or Client nfs lists the parameters which contain informa-
tion about NFS:
calls is the number of NFS requests which were sent.
badcalls is the number of requests still unanswered after execu-
tion of the retrans option [see mount(1M)]
nclget is the number of times internal data structures were
requested (for client only).
nclsleep is the number of times it was necessary to wait for
nclget (for client only).
The following parameters affect internal RPC calls sent to or received
by a system. An absolute figure is always specified and a percentage
value. The meaning is as follows (sorted alphabetically):
NFS V2 and NFS V3:
create create a file
fsstat file system statistics
Page 4 Reliant UNIX 5.44 Printed 11/98
nfsstat(1M) nfsstat(1M)
getattr request attributes for a file entry
link simple link (hard link)
lookup search for a file entry
mkdir create directory
null no action (for test purposes)
read read in a file
readdir read in a directory
readlink read from a symbolic link
remove delete a file
rename rename a file
rmdir delete directory
root is not supported in this version (V2 only)
setattr set attributes for a file entry
symlink symbolic link
wrcache is not supported in this version (V2 only)
write write in a file
NFS V3 only:
access check file permissions
commit commit write request
fsinfo fetch static file system information
fsstat fetch dynamic file system information
mknod create device file
pathconf fetch information on the path configuration of a file
(maximum path length, maximum number of links etc.)
readdirplus extended read of a directory in accordance with NFS V3
Page 5 Reliant UNIX 5.44 Printed 11/98
nfsstat(1M) nfsstat(1M)
Information on time usage
If you specify the -T option, information on time usage is issued for
every NFS procedure. This information is divided, on the one hand,
according to client and server, and, on the other hand, according to
NFS V2 and V3.
When a procedure is called, the client calculates the time from when
the call is issued until the response is received from the server.
From the server's point of view, the time is calculated from the
receipt of the client request until the response is sent.
The following information is issued for every NFS procedure:
- The number of calls
- The number of calls not answered within the specified time
(timeouts)
- The average time per call in milliseconds (msec avg)
- The total time for all calls (in seconds)
- The distribution of times to specific intervals. The intervals are
arranged according to specific procedures, i.e. they are not the
same for all procedures.
The upper limits for the intervals are specified in milliseconds as
well as the absolute number and the percentage share of those pro-
cedure calls, whose time usage falls within the interval. Times
that exceed the highest interval, are assigned to it.
The information on time usage is not updated with the -z option.
Statistical information within a Reliant Cluster system
When you specify the -M option, statistical information is output
within a MESH network.
Examples:
The output from nfsstat [options]-M within a Reliant Client system
corresponds to the sum of the outputs from nfsstat [options] on all
nodes that are currently running within a MESH network.
The output from nfsstat [options]-M node-1 on a node within a Reliant
Cluster system corresponds to the output from nfsstat [options] on
node-1, provided that node-1 is located in the same MESH network and
is currently running.
Page 6 Reliant UNIX 5.44 Printed 11/98
nfsstat(1M) nfsstat(1M)
The output from nfsstat [options]-M node-1 node-6 on a node within a
Reliant Cluster system corresponds to the sum of the outputs from
nfsstat [options] on node-1 and node-6, provided that both are located
in the same MESH network and are currently running.
Page 7 Reliant UNIX 5.44 Printed 11/98