-----Original Message-----
From: Jacob Robert Wilkins [mailto:jrw@nplus1.net]
Sent: Wednesday, February 11, 2004 1:45 PM
To: Gabe Newell
Cc: [email]spark@1sks.com[/email]
Subject: bladeforums diagnostics...
I turned searhing back on. Running vmstat with 2 second intervals on the
box, I did a search for the word "lock"
"bi" is the interesting field here. It represents the number of blocks
read from the disk.
[root@bladedb root]# vmstat 2
procs memory swap io system cpu
r b w swpd free buff cache si so bi bo in cs us sy id
1 0 0 1216 35620 216 797176 0 0 4 0 3 5 4 3 4
0 0 0 1216 35620 216 797176 0 0 0 0 187 96 1 0 99
0 0 0 1216 35584 216 797192 0 0 8 0 269 210 2 6 92
1 0 0 1216 23356 232 798596 0 0 702 18 433 294 4 5 91
0 0 0 1216 33472 232 798660 0 0 32 0 266 205 47 6 48
# boom, search starts here
0 1 0 1216 20028 232 804500 0 0 2920 0 495 512 9 7 84
0 2 0 1216 9084 232 811848 0 0 16888 0 572 516 5 5 90
1 1 0 1216 7356 232 817848 0 0 17034 0 458 356 3 7 90
0 1 0 1216 6136 232 821896 0 0 10788 64 496 358 3 5 92
1 1 0 1216 6440 232 805328 0 0 8398 0 502 299 47 9 44
0 1 0 1216 7292 232 817956 0 0 12114 64 569 504 3 10 86
1 0 0 1216 9044 232 815708 0 0 10836 0 613 536 4 7 89
0 1 0 1216 7196 232 821632 0 0 9702 0 631 542 4 2 94
1 1 0 1216 8948 232 815844 0 0 9622 0 720 711 3 9 87
0 1 0 1216 7304 232 821564 0 0 9864 0 652 595 5 4 91
1 1 0 1216 7360 232 821464 0 0 12218 64 590 530 3 8 89
0 1 0 1216 9240 232 819496 0 0 9018 0 599 558 3 4 93
0 1 0 1216 9880 232 814388 0 0 11224 0 519 431 3 4 92
1 1 0 1216 13404 232 815164 0 0 5624 64 666 776 10 10 79
0 1 0 1216 6364 232 819716 0 0 10420 0 438 1138 10 3 87
1 1 0 1216 9984 232 810552 0 0 14544 256 536 493 8 13 79
0 1 0 1216 7468 232 815812 0 0 10162 0 386 362 4 5 91
0 1 0 1216 6620 232 816348 0 0 13432 0 444 386 3 2 95
0 1 0 1216 6336 232 816604 0 0 13474 0 546 489 4 13 84
0 1 0 1216 7444 232 815820 0 0 15978 0 634 584 3 6 91
0 1 0 1216 7416 232 815364 0 0 14144 0 537 482 4 5 91
0 1 0 1216 9956 232 811228 0 0 17188 0 513 428 2 9 89
0 1 0 1216 7744 232 818208 0 0 12090 0 594 564 6 4 91
0 1 0 1216 6648 200 816064 0 0 10494 64 531 526 4 11 86
1 1 0 1216 7392 200 815332 0 0 15122 0 529 439 5 3 92
0 1 0 1216 6608 200 818672 0 0 10776 0 695 707 7 8 85
0 0 0 1216 8972 200 813784 0 0 7388 0 533 625 4 11 85
# search is over
0 0 0 1216 8972 200 813796 0 0 6 0 306 245 2 1 97
1 0 0 1216 9960 200 813176 0 0 374 64 374 327 5 8 87
0 1 0 1216 9824 200 813404 0 0 114 0 277 234 1 2 96
0 1 0 1216 10056 200 812468 0 0 172 0 360 302 4 4 92
0 0 0 1216 10056 200 812492 0 0 12 0 238 169 1 6 93
0 0 0 1216 10056 200 812500 0 0 4 0 236 169 2 1 97
0 0 0 1216 10004 200 812528 0 0 14 64 199 107 1 2 98
0 0 0 1216 9736 200 812604 0 0 38 0 320 280 4 7 90
0 0 0 1216 11492 200 811016 0 0 530 0 399 496 2 6 92
All while the system was running this query, CPU utilization was low.
Both processors were averaging about 80% idle. More CPU time was spent
in system than in user.
Memory on the box is fairly full, but not swapping.
[root@bladedb root]# free -m
total used free shared buffers cached
Mem: 1009 984 25 0 0 784
-/+ buffers/cache: 198 810
Swap: 2047 1 2046
It looks to me like we are I/O, and a little low in memory.
The processor is willing, but the disks can feed the data fast enough. Adding
memory would help with caching, but that'll only get you so far.
jrw