hi, i recently migrated our news box (which runs approx. 1MB/s nntp
traffic inbound and >2MB/s outbound) to several systems and got the
spool onto a vinum volume that spans 6 scsi busses (adaptec 3940UW) each
4 disks (IBM DNES 18GB). theres a private switched interconnect between
the reader box and the "fileserver" now and nfs seem to crash on me
somehow. nnrpd start to die after a few megabytes transfer and if i do
an fstat -vp on one of the nnrpd's i get
root@NEWS:black[~]31# fstat -vp 3820
USER CMD PID FD MOUNT INUM MODE SZ|DV R/W
news nnrpd 3820 root / 2 drwxr-xr-x 512 r
news nnrpd 3820 wd /sp 158720 drwxrwxr-x 923136 r
news nnrpd 3820 text / 376 -r-xr-xr-x 90424 r
news nnrpd 3820 0* internet stream tcp
news nnrpd 3820 1* internet stream tcp
news nnrpd 3820 2* internet stream tcp
news nnrpd 3820 3* local dgram d4c7f400 <-> d4c7efc0
news nnrpd 3820 4* internet dgram udp d44b1600
news nnrpd 3820 5 /sp 158720 drwxrwxr-x 923136 r
news nnrpd 3820 12* local dgram d4c7fe40
example lsof output of one of the dead nnrpd's:
COMMAND PID USER FD TYPE DEVICE SIZE/OFF NODE NAME
nnrpd 3820 news rtd VDIR 4,131072 512 2 /
nnrpd 3820 news txt VREG 4,131072 90424 376
/home/news/bin/nnrpd
nnrpd 3820 news txt VREG 4,131078 63832 182626
/usr/libexec/ld-elf.so.1
nnrpd 3820 news txt VREG 4,131078 7579 39707
/usr/lib/libscrypt.so.2
nnrpd 3820 news txt VREG 4,131078 31855 39805
/usr/lib/libutil.so.2
nnrpd 3820 news txt VREG 4,131078 517727 39742 /usr/lib/libc.so.3
nnrpd 3820 news 0u inet 0t0 TCP can't read inpcb
at 0x00000000
nnrpd 3820 news 1u inet 0t0 TCP can't read inpcb
at 0x00000000
nnrpd 3820 news 2u inet 0t0 TCP can't read inpcb
at 0x00000000
nnrpd 3820 news 3u unix 0xd42353c0 0t0 ->0xd422edc0
nnrpd 3820 news 4u inet 0xd44b1600 0t0 UDP
localhost:1276->localhost:1119
nnrpd 3820 news 5r VDIR 130,1 923136 158720
/sp/art/alt/binaries/sounds/mp3
nnrpd 3820 news 12u unix 0xd422e320 0t0 ->(none)
system config is as follows:
both boxes running freebsd 3.3-RELEASE
fileserver backend
tyan tomcat ii (i440BX) bios 1.18
p-iii 450 (options I686_CPU, NO_F00F_HACK)
256m sdram (will be 768m sdram/ecc)
ahc[0..5] 3x aha-3940uw (options AHC_ALLOW_MEMIO)
da0 ibm dnes 9GB uw
da[1..24] 24x ibm dnes 18GB uw
fxp0 intel etherexpress pro 100+/server (options NMBCLUSTERS=8192)
media 100BaseTX mediaopt full-duplex
filesystems on vinum w/ softupdates
/db stripe, 4disks, 68GB
/sp stripe, 20disks, 341GB
nfs_server_flags="-u -t -n 16"
newsreader frontend
asus p2b-s (i440BX) bios 1011
p-iii 450 (options I686_CPU, NO_F00F_HACK)
512mb sdram (will be 1gb sdram/ecc)
ahc0 adaptec u2w onboard
da0 ibm dnes 9GB u2w
fxp[0,1] intel etherexpress pro100+/server (options NMBCLUSTERS=8192)
media 100BaseTX mediaopt full-duplex
fxp0 -> internet
fxp1 -> private storage network 10.0.0.0/24
nfs_client_flags="-n 8"
fstab:
bucket:/sp /sp nfs \
rw,bg,noatime,-r=1024,-w=1024,intr,-a=3,-d,-t=10 0 0
bucket:/db /db nfs \
rw,bg,noatime,-r=1024,-w=1024,intr,-a=3,-d,-t=10 0 0
the main question: why do those guys hang there? the nnrpd i use is from
the inn-1.7.2-insync+actived distribution/patchkit
any ideas? as far as i understand it chokes because of nonexistent pcb's
that are set 0, but i'm not this deep into nfs anyway... any help
appreciated...
/k
--
> "There is a God, but He drinks" -- Blore
http://www.nacamar.de - http://www.nacamar.net - http://www.webmonster.de
http://www.apache.de - http://www.quakeforum.de - finger [EMAIL PROTECTED]
KR433/KR11-RIPE - PGP-KFP = F9 A0 DF 91 74 07 6A 1C 5F 0B E0 6B 4D CD 8C 44
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
