Error: "nfsserv warning 9001001, nfs request queue is full"
(Last modified: 05Dec2002)
This document (10022812) is provided subject to the disclaimer at the end of this document.
fact
Novell NetWare 6.0
Novell NetWare 5.1
Novell NetWare 4.2
Novell NFS Services 2.3
Novell NFS Services 3.0
symptom
Error: "nfsserv warning 9001001, nfs request queue is full"
Error: "nfs request queue is full, current request is dropped"
Error: "v2 request queue is full, any packets received could be fragmented
Slow NFS Server response.
Slow server response to user requests.
Copying a large number of files takes too long.
cause
NFS Server parameters and/or NFS Client parameters need to be adjusted.
fix
Slow Server Performance
If NFS server performance is very slow when a large file is copied to the server, the problem could be that the LAN adapter or adapter driver is losing packets. To determine if an unacceptable packet loss rate is the cause of the problem, check the IP ipReasmFails statistic.
To check this statistic, load TCPCON at the server console. From the Available Actions main menu, click Statistics, then click IP Statistics to see ipReasmFails.
NOTE: The value for ipReasmFails increments is 15 seconds (the timeout period) after the arrival of the last fragment.
If it is observed that the value increases, two courses of action are available:
Decrease the client's buffer size. On UNIX systems, buffer size relates to the size of the NFS write operations the client uses to write data to the server. By default, this size is 8 KB. From most UNIX systems, this size can be decreased when a NetWare file system is mounted using the wsize mount command option.
For example, the following command entry reduces the client write area to 2048 bytes:
mount -o soft,wsize=2048 <server>: /volume1/directoryx/NetWare
Obtain a faster network board for the server. For example, if the Ethernet board is an 8-bit board, replace it with a 16-bit board.
Set the switch to half duplex.
For additional information on TCPCON, refer to the TCP/IP Transport Supervisor's Guide.
NFS Server Parameters:
1) Increase the following parameters in the Set Parameters section of the Administer NFS Server section in UNICON:
Number of NFS Threads
Number of Open File Cache
Number of Request Cache Entries
NOTE: Adjustment of these parameters will affect usage of the total amount of RAM memory. A review of the total amount of RAM memory in the system may be in order.
2) Either apply the latest support pack available for the server or contact a Novell Technical Support Engineer.
Very Slow Response for NFS Clients:
Some UNIX file systems (in particular, BSD and its derivatives) support quota usage restrictions. If quota usage is active and an NFS server is not operating, response can be slow because login is trying to get quota information from the NFS system. If the response is too slow, use the mount option noquotausage or dismount attached clients
before taking the NFS server down.
I/O Failures and RPC Time-outs:
An NFS client can experience errors such as remote procedure calls (RPCs) timing out so that read operations fail. Various conditions can cause this problem, such as heavy network traffic or the server being overloaded. Parameters on both the client side and the server side can be tuned to better suit the network environment. The following section provides information on tuning these parameters.
NFS Client Parameters:
Some NFS clients have adjustable parameters for dealing with slower NFS server responses. For UNIX clients, the mount command provides the following options:
NOTE: The default values for the mount options vary among vendors and software releases. Refer to the operating system documentation for platform-specific options.
Soft/hard---indicates how the directory is mounted. A soft mount specifies that an error is returned if the server does not respond. A hard mount specifies that the retry request be made continually until the server responds. When the network traffic is heavy or the server is very slow, a soft mount is a good option to try.
NOTE: The use of a hard mount forces continual retries and keeps the network traffic high.
Timeo---the number of tenths of a second before time-out occurs. Set timeo higher if RPC times out or I/O failure occurs as a result of a slow network or NFS server.
Retrans---the number of retransmissions. Set retrans higher if RPC times out or if I/O failure occurs as a result of a slow network or NFS server.
Retry---the number of times to retry the mount operation. The NFS server might not be able to respond to the mount request if it is too busy or if network traffic is heavy. If the NFS server is operating but does not respond to a mount request, specify a higher number for the retry parameter.
Server Parameters:
Two NetWare parameters affect NFS performance. These two parameters control NFS packet size and receive buffer capacity. Specify these parameters in the STARTUP.NCF file by using the SET command. Refer to the SET command in the Utilities Reference in the NetWare documentation for syntax and related information.
Maximum Physical Receive Packet Size--- the largest packet that can be transmitted on any of the file server's networks. For Ethernet, set this parameter to at least 1518. Setting it to a smaller size would cause unnecessarily fragmentation or even complete failure to communicate with other devices. For token ring, the recommended minimum is 4202.
Minimum Packet Receive Buffers--- the minimum number of packet-receive buffers that NetWare can allocate. It is best to anticipate NFS traffic and set this parameter when starting NetWare. Otherwise, when the server receives a large number of NFS requests, response time increases while NetWare allocates more buffers. The recommended number of minimum packet-receive buffers for a server running NFS services is 500.
Maximum Packet Receive Buffers--- the maximum packet receive buffers that NetWare can allocate. If this is too low, the server will have to drop packets that it is not ready to process. The maximum needed by a server can vary extremely. Even on a lightly-loaded server running NFS, set the maximum to at least 1000. It would be better to use 3000 - 5000, however..
document
| Document Title: | Error: "nfsserv warning 9001001, nfs request queue is full" |
| Document ID: | 10022812 |
| Solution ID: | 1.0.42223022.2435908 |
| Creation Date: | 07Dec1999 |
| Modified Date: | 05Dec2002 |
| Novell Product Class: | Connectivity Products |
disclaimer
The Origin of this information may be internal or external to Novell. Novell makes all reasonable efforts to verify this information. However, the information provided in this document is for your information only. Novell makes no explicit or implied claims to the validity of this information.
Any trademarks referenced in this document are the property of their respective owners. Consult your product manuals for complete trademark information.