I start a slrn (newsreader) session on a very big newsgroup; 150000 headers
that are read from disk, then sorted. The process takes about 100 Mb memory,
I have 512 Mb, so there should be enough to spare.
This is without any special values in /proc, BTW, just the defaults.
here we start (the logfile is "vmstat 2")
procs memory swap io system cpu
r b w swpd free buff cache si so bi bo in cs us sy id
0 1 0 8180 103364 86792 257804 15 16 109 156 85 151 6 6 88
0 0 0 8180 103328 86792 257828 6 0 8 0 113 81 0 1 99
0 0 0 8152 102800 86792 257864 0 0 24 68 117 69 1 0 99
1 0 0 8152 92976 88652 264304 0 0 1136 0 391 635 7 8 85
1 0 0 8152 72700 93704 276452 0 0 2060 4260 701 1141 12 16 71
1 0 0 8152 53348 97780 287576 0 0 1810 3594 608 1057 12 13 75
1 0 0 8152 35516 101452 297956 0 0 1786 3594 616 1027 9 13 77
0 1 1 8160 11376 107952 310756 0 0 2184 6452 731 1236 13 17 70
2 0 1 8160 1588 109112 315664 0 0 1836 1798 597 964 9 32 59
1 0 0 8160 2224 106436 318588 0 0 1578 3750 624 923 10 16 74
3 0 0 8172 1956 109248 315820 0 0 986 3594 402 593 5 36 58
0 1 0 8172 1588 108156 315932 0 0 1878 1996 590 1033 10 34 56
0 1 0 8172 1592 104476 317924 0 0 1336 2580 514 785 7 17 76
1 0 0 8172 1668 107912 313736 0 0 1096 3598 462 650 7 40 53
1 0 0 8172 1780 108448 312740 0 0 1382 3680 520 796 7 39 53
0 1 1 8172 1588 107212 312552 0 0 1778 3604 597 987 9 43 49
1 0 2 8172 1748 104880 313196 0 0 1906 1234 613 1060 11 23 66
0 1 1 8172 2432 107072 308980 0 0 1018 4570 443 695 6 39 55
0 1 1 8172 1588 104880 310532 0 0 1828 1806 589 999 10 39 51
1 0 1 8172 2032 104616 309316 4 0 1492 2570 548 925 8 19 74
0 1 1 8172 1740 106308 306164 0 0 1082 3894 442 678 7 39 54
0 1 1 8172 1588 106056 304992 0 0 1796 2656 603 989 11 45 44
0 1 0 8172 1752 104732 305664 0 0 1310 3082 501 813 7 18 75
0 1 1 8172 2112 104288 304576 0 0 1042 3808 475 703 9 33 57
0 1 0 8140 1592 104400 304076 0 0 1174 514 488 694 5 16 79
0 1 2 8200 1740 105144 302364 0 0 1252 2688 454 831 5 49 46
1 0 0 8388 1596 104128 302276 0 0 1292 1008 478 876 7 51 42
1 0 0 8388 1592 105004 301368 0 0 394 4410 380 434 2 16 82
free space is gone, so let's hit the buffers and start to swap.
0 3 2 11648 2976 105408 302268 0 0 1238 1730 594 15566 3 52 45
0 1 0 33284 15572 87288 328872 0 8 846 1668 383 717 5 72 23
0 1 0 34160 1560 91440 336572 0 0 1502 3596 558 896 7 11 82
1 0 0 34116 1560 95112 329560 0 0 1572 3662 587 915 9 13 77
0 1 0 34604 1560 98780 323048 0 0 1582 3602 586 894 9 13 77
0 1 1 35112 1560 102448 315872 0 0 1832 3604 606 1002 8 17 75
1 0 2 36244 1808 103312 316984 0 0 1834 442 620 999 11 33 56
1 0 0 44836 1592 103132 327536 0 2 1364 4966 525 860 7 32 61
1 0 0 44816 1588 103516 326796 0 166 1350 4282 546 925 8 38 54
1 0 0 44772 2088 100884 327524 0 26 1868 1616 618 1095 12 46 42
1 0 1 44772 1588 101784 326180 0 2 1418 2060 501 994 9 38 53
1 0 0 44772 1768 99704 327772 0 2 700 4884 433 575 6 22 72
0 1 1 44772 1588 100308 326276 0 6 1922 2096 611 1049 11 44 45
1 0 1 44768 2656 102604 322084 0 10 1658 1786 567 1151 7 46 47
1 0 1 44768 1876 101896 322604 0 0 940 5120 466 643 6 12 82
1 0 0 44768 3000 99608 322388 0 0 1644 1480 533 1409 7 42 50
1 0 1 44800 2480 100648 320656 0 6 1622 2304 566 1086 8 54 39
1 0 1 46168 1700 99072 323320 0 0 958 4950 476 716 7 29 64
1 0 0 50824 1592 99512 326356 0 48 1708 2196 583 1033 7 54 39
0 1 1 55288 3044 98584 329060 0 2 1342 1728 495 854 9 59 32
0 1 1 58612 1588 98408 332860 10 6 912 4818 485 764 6 40 54
1 0 0 62768 2032 97168 336944 0 466 1116 1942 462 938 8 60 32
1 0 0 65364 1872 98640 336960 26 574 1510 2246 541 997 11 49 41
0 1 1 66552 1588 98256 337872 0 16 884 3844 478 708 6 36 58
0 1 1 67228 1588 99456 336400 0 1090 962 3112 440 778 5 56 40
1 0 1 67228 1996 99200 335796 0 0 762 2602 465 900 5 68 26
0 1 2 67236 1604 98408 335364 0 520 1210 3108 551 913 7 53 40
0 1 0 67232 1588 98652 334412 0 0 880 836 353 625 6 28 66
0 1 1 69624 1980 97792 336468 0 504 680 4184 442 733 5 35 60
1 0 0 72316 1564 98948 337568 0 488 994 1698 411 645 7 48 44
1 0 0 72932 1576 97868 337164 0 472 1828 2148 584 1003 9 43 47
0 1 2 75296 1908 96488 339768 0 0 1234 3924 505 845 8 53 39
1 0 1 77548 1588 97468 340488 0 458 724 3334 400 633 4 43 53
1 0 0 78932 2408 97648 340136 0 444 1298 2044 509 966 9 59 32
0 1 1 78956 1588 98172 339328 0 1254 778 4880 488 783 5 59 36
0 1 1 78932 1732 98600 337380 0 404 1338 2474 511 978 10 41 49
1 0 2 78940 2308 98488 336096 0 748 744 1776 342 546 3 44 52
0 1 1 80004 1588 99212 336680 0 3106 498 6316 469 520 4 37 59
0 1 1 81780 2740 101060 333984 0 260 1124 1570 449 899 6 34 60
1 0 1 87872 1588 100032 341440 0 258 770 1614 393 964 5 73 23
1 0 1 88476 1600 100052 341260 0 0 614 3398 443 613 4 30 66
1 0 0 88440 1588 103084 334824 0 6284 1864 6466 646 1086 13 46 41
0 1 0 88440 1588 102560 334276 0 386 1116 4828 455 694 8 35 57
0 1 1 88440 2036 103448 332268 0 0 1056 3634 476 719 6 37 57
0 1 0 88440 1588 105036 330036 0 0 1534 1542 512 901 9 36 55
0 1 1 88440 1588 103372 330084 0 184 1724 2288 567 981 9 44 47
0 1 1 88448 1640 101032 331504 0 0 860 3226 443 732 5 29 66
1 0 0 88444 1592 100640 329644 0 0 1192 1872 506 791 6 22 72
0 1 0 88444 1588 103748 325540 0 5692 1366 8482 549 851 8 62 29
0 1 2 90508 1588 103408 326220 0 4582 1176 7582 533 943 7 48 45
1 0 1 92488 1588 104004 326168 0 0 744 2194 427 677 4 28 68
0 1 1 101796 2544 105052 333184 0 7076 32 9146 267 692 1 93 6
1 0 0 101780 2200 107328 327812 24 0 2290 2094 754 1261 12 43 45
1 0 0 101912 1596 106396 327540 0 0 1152 2644 514 863 5 23 72
0 1 1 101912 1760 108412 324548 0 0 734 3596 375 433 3 37 60
1 0 1 102596 1588 107716 326700 3432 0 3620 106 211 254 36 9 55
deeper and deeper into swap we grow, yet the 300+ megabyte cache stays intact?
1 0 1 102936 1588 105724 329432 3986 0 3986 144 187 218 42 17 41
1 0 0 102936 2840 102536 331608 3912 0 3912 28 190 234 42 12 46
1 0 0 102936 1588 101452 334532 4058 0 4058 26 185 218 42 4 55
1 0 0 102936 1588 100152 336736 4096 0 4096 10 173 202 42 4 54
1 0 0 102936 1588 96716 340532 3684 0 3684 40 171 192 43 4 54
Now we're sorting headers, notice how we read back swap pages - we
must've swapped out my slrn-process to keep the cache intact.
1 0 0 102936 1588 96444 341836 3866 0 3866 8 163 177 43 5 52
1 0 0 102936 3116 93416 344060 4012 0 4012 2 178 203 44 4 52
1 0 1 102952 1596 91724 347792 3968 0 3968 38 172 187 42 32 26
1 0 0 102936 1600 91724 347672 20 0 20 0 104 51 49 2 48
1 0 0 102936 1748 91724 347624 0 0 0 38 111 68 50 2 48
The program is idle:
Name: slrn
State: R (running)
Pid: 11167
PPid: 519
TracerPid: 0
Uid: 500 500 500 500
Gid: 100 100 100 100
FDSize: 256
Groups: 100 13 14 16 17 33 101
VmSize: 103220 kB
VmLck: 0 kB
VmRSS: 101152 kB
VmData: 100892 kB
VmStk: 72 kB
VmExe: 296 kB
VmLib: 1752 kB
SigPnd: 0000000000000000
SigBlk: 0000000000000000
SigIgn: 8000000000201000
SigCgt: 0000000008184083
CapInh: 0000000000000000
CapPrm: 0000000000000000
CapEff: 0000000000000000
after closing slrn; i see
procs memory swap io system cpu
r b w swpd free buff cache si so bi bo in cs us sy id
0 0 0 8076 91408 91484 262912 22 22 131 198 90 159 6 7 87
Is there any tuning I can do myself in /proc, or is starting a 100 Mb
program in 512 Mb of memory without swapping beyond the possibilities of
the current VM? (Don't take this personnally, VM-guru's :-) )
Thanks,
Jurriaan
-- BOFH excuse #400:We are Microsoft. You are not experiencing problems. GNU/Linux 2.4.5-ac15 SMP/ReiserFS 2x1402 bogomips load av: 0.07 0.04 0.08 - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/