i386/148204: UDP NFS causes overload

[Walter C. Pelissero]

>Number:         148204
>Category:       i386
>Synopsis:       UDP NFS causes overload
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    freebsd-i386
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Mon Jun 28 08:00:10 UTC 2010
>Closed-Date:
>Last-Modified:
>Originator:     Walter C. Pelissero
>Release:        FreeBSD 8.1-PRERELEASE i386
>Organization:
>Environment:
System: FreeBSD zaphod.home.lan 8.1-PRERELEASE FreeBSD 8.1-PRERELEASE #0: Mon Jun 21 22:28:44 CEST 2010 root at zaphod.home.lan:/usr/home/obj/usr/src/sys/TIGER-MP i386


	
>Description:
	Having UDP mounts, like as for this amd.map:

	* opts:=rw,grpid,resvport,vers=3,proto=udp,nosuid,nodev,intr,rsize=16384,wsize=16384

	causes NFS to grind to a halt when doing substantial i/o on
	the NFS mount.

	Top shows a number of nfsiod spinning out of control, causing
	a massive increase of system load (>20).  Killing the process
	responsible for the i/o doesn't help.

last pid:  2637;  load averages: 20.11, 14.63,  7.53    up 0+00:37:45  16:00:08
197 processes: 23 running, 156 sleeping, 18 waiting
CPU:  4.1% user,  0.0% nice, 95.9% system,  0.0% interrupt,  0.0% idle
Mem: 613M Active, 53M Inact, 101M Wired, 8228K Cache, 79M Buf, 1227M Free
Swap: 2014M Total, 2014M Free

  PID USERNAME      THR PRI NICE   SIZE    RES STATE   C   TIME   WCPU COMMAND
 2405 root            1  76    -     0K     8K RUN     1   0:34 10.06% nfsiod 10
 2403 root            1  76    -     0K     8K RUN     1   0:35  9.96% nfsiod 8
 2397 root            1  76    -     0K     8K RUN     1   0:35  9.96% nfsiod 2
 2406 root            1  76    -     0K     8K RUN     1   0:35  9.77% nfsiod 11
 2412 root            1  76    -     0K     8K RUN     1   0:36  9.57% nfsiod 17
 2409 root            1  76    -     0K     8K RUN     1   0:36  9.47% nfsiod 14
 2404 root            1  76    -     0K     8K RUN     1   0:35  9.47% nfsiod 9
 2396 root            1  76    -     0K     8K RUN     1   0:33  9.47% nfsiod 1
 2395 root            1  76    -     0K     8K RUN     1   0:35  9.38% nfsiod 0
 2398 root            1  76    -     0K     8K RUN     1   0:37  9.28% nfsiod 3
 2399 root            1  76    -     0K     8K RUN     1   0:34  9.28% nfsiod 4
 2402 root            1  76    -     0K     8K RUN     1   0:34  9.18% nfsiod 7
 2411 root            1  76    -     0K     8K RUN     1   0:35  9.08% nfsiod 16
 2410 root            1  76    -     0K     8K RUN     1   0:37  8.98% nfsiod 15
 2407 root            1  76    -     0K     8K RUN     1   0:36  8.98% nfsiod 12
 2413 root            1  76    -     0K     8K RUN     1   0:37  8.79% nfsiod 18
 2408 root            1  76    -     0K     8K RUN     1   0:34  8.79% nfsiod 13
 2401 root            1  76    -     0K     8K RUN     1   0:36  8.69% nfsiod 6
 2400 root            1  76    -     0K     8K CPU1    0   0:33  8.69% nfsiod 5
 2414 root            1  76    -     0K     8K RUN     1   0:34  8.40% nfsiod 19
[...]

>How-To-Repeat:
	cd /to/nfs/mount
	cp /dev/random .
>Fix:

	I worked around the problem using TCP instead of UDP.


>Release-Note:
>Audit-Trail:
>Unformatted: