[55369] in Hotline Meeting
Log 218422: more on intaglio NFS problem
daemon@ATHENA.MIT.EDU (Larry Stone)
Thu Jun 14 21:19:17 2001
Date: Thu, 14 Jun 2001 21:19:13 EDT
From: Larry Stone <lcs@MIT.EDU>
Reply-To: <lcs@MIT.EDU>
To: ds-sys@MIT.EDU, hotline@MIT.EDU
Cc: lcs@MIT.EDU
Message-ID: <CMM.0.90.4.992567953.lcs@defiant.mit.edu>
After Lou completed the reinstall and I redid the customizations
(Intaglio has an external disk drive, some special cron jobs, the
Hummingbird hclnfsd PC nfs authentication server, and is "mkserv nfs"
to be an NFS server), we observed the same problems that triggered the
original complaint.
Then I tested it with another Athena sun work station as the NFS client
and found no problems. In one load test I copied 63Mb (average file
size 173Kb, representative of image files) from intaglio to my workstation
and then back again. Observations:
1. Reading 63Mb took 48 sec, writing took 2:14, which is normal enough.
2. The client didn't log any retransmits, which means the server
(intaglio) was responding promptly.
3. Running "netstat -i 10" on intaglio showed a lot of ethernet
collisions and output errors. In the read test, when it was pumping
out a lot of large packets, the collision rate hit about 50%! This
implies to me there is something questionable about the network hardware,
either cabling, the Sun's interface, or the bridge/switch/repeater it's
connected to. (logs below)
Another thing I checked was a protocol trace of the PC's conversation with
intaglio. The PC *was* retrying NFS operations several times, apparently
without result. Perhaps it needs some NFS parameters tweaked -- could
its settings have gotten changed in the last few days? Normally Windows
PCs aren't that good at staying in sync :-) but maybe they all got
changed at once? Unfortunately, I know nothing about NFS clients on WIndows.
-- Larry
Logs of "netstat -i 10" (10sec intervals),
1. For NFS Write test, data moving defiant->intaglio (logged on intaglio):
input le0 output input (Total) output
packets errs packets errs colls packets errs packets errs colls
2741 0 1482 1 35 2741 0 1482 1 35
2715 0 1482 2 111 2715 0 1482 2 111
4901 0 2553 2 127 4901 0 2553 2 127
3287 0 1757 2 125 3287 0 1757 2 125
2817 0 1525 1 90 2817 0 1525 1 90
3218 0 1715 0 112 3218 0 1715 0 112
2426 0 1322 1 67 2426 0 1322 1 67
2657 0 1443 1 63 2657 0 1443 1 63
2425 0 1299 3 46 2425 0 1299 3 46
3074 0 1640 1 81 3074 0 1640 1 81
5261 0 2688 2 239 5261 0 2688 2 239
5261 0 2710 2 257 5261 0 2710 2 257
2513 0 1278 1 60 2513 0 1278 1 60
2. For NFS REad test, data going defiant<-intaglio:
input le0 output input (Total) output
packets errs packets errs colls packets errs packets errs colls
1175 0 1965 61 1020 1175 0 1965 61 1020
991 0 1666 54 847 991 0 1666 54 847
639 0 1030 34 370 639 0 1030 34 370
1189 0 1990 61 865 1189 0 1990 61 865
1161 0 1941 49 1048 1161 0 1941 49 1048
807 0 1335 40 465 807 0 1335 40 465
770 0 1262 36 539 770 0 1262 36 539
483 0 786 37 327 487 0 790 37 327
1014 0 1712 42 763 1014 0 1712 42 763
1248 0 2110 49 1002 1248 0 2110 49 1002
874 0 1407 47 603 874 0 1407 47 603
1214 0 2018 48 976 1214 0 2018 48 976
1408 0 2393 66 1036 1408 0 2393 66 1036
1358 0 2281 63 1026 1358 0 2281 63 1026
1610 0 2742 76 1349 1610 0 2742 76 1349
1220 0 2010 85 1089 1220 0 2010 85 1089
1196 0 2029 58 1052 1200 0 2033 58 1052
1283 0 2133 73 1021 1283 0 2133 73 1021
754 0 1198 37 582 754 0 1198 37 582
1324 1 2216 67 1107 1324 1 2216 67 1107
