From Bugzilla Helper: User-Agent: Mozilla/4.77 [en] (X11; U; Linux 2.4.2-2 i686) Description of problem: NFS writes hang now and then, resulting in a NFS server not responding message in /var/log/messages. This is most noticeable (and annoying) when using vi to edit a file in a NFS mounted directory. Vi freezes now and then, presumably when auto saving changes. Have also gotten NFS server not responding messages from other programs writing to NFS mounted directory, but this is less noticable than vi freezing. NFS service recovers and I have not experienced any data loss or corruption. How reproducible: Always Steps to Reproduce: 1. Use vi to edit a file in a NFS mounted directory 2. After awhile vi will freeze (no response to keystrokes) It seems to happen fairly often when you write changes with :w 3. After 10 sec or so vi resumes and processes your buffered keystrokes. Actual Results: Vi froze now and then. Got NFS server not responding errors in /var/log/messages. Have also gotten these messages with other programs that write files in a NFS mounted directory, but it is most aggravating with vi because it keeps freezing. I suspect this happens when vi auto saves periodically. Expected Results: Vi should not freeze. Other (not Linux 7.1) NFS clients on this network having no trouble with NFS server. Additional info: My system is a Dell OptiPlex GX100 with a 3Com PCI 3c905C Tornado Network Card (according to dmesg). I have experienced this problem with 2 different NFS servers: IBM RS/6000 running AIX 4.3 Network Appliance F630 fileserver running ONTAP 6.0.1R3 I have experienced this problem on another Linux 7.1 system on this same network. This box was built here from parts. not sure of the hardware details. Other NFS clients (Solaris, AIX, IRIX) on this network have no problems with these NFS servers. I am mounting rw,hard,intr I presume I am getting NFS vers 3. I am not specifying rsize or wsize. Never had problems with RedHat 6.2, but that used NFS vers 2. Have not tried vers=2 or smaller rsize, wsize.
I have the same problem. kernel 2.4.3-15enterprise (with knfsd/reiserfs/quota patches from www.namesys.com). The NFS exported filesystem is reiserfs and quotas are enabled. The exported device is accessed through an aic7xxx card using aic7xxx_mod.o. Client logs show 'NFS server ... not responding, still trying' message. IBM e(logo)Server xSeries 230 client and server. PCnet/FAST III 79C975 (PCI) using pcnet32 driver. Mount options rw,rsize=8192,wsize=8192,hard,intr. Is this just a throughput/latency tradeoff? Will upgrading to 2.4.7 help?
This is very similar to my problem #58747, also accessing a Netapp filer V6.1.1R2. However we are using the Intel 82557 Ethernet Pro 100 card. We have sniffer traces that show Redhat stops asking the filer for data while it is reporting that the filer is not responding. Then all of a suddent, it starts asking the filer for data again. Looks like something is getting lost in the kernel.
This seems to be fixed in later kernels.