On Wednesday 26 February 2003 08:49 am, thorsten Sideb0ard wrote: > Hi there, > > i recently subscribed to the list, and have been enjoying browsing the > threads, however i didn't think i would be posting as soon... > > I wonder if anyone can offer any advice... > > I have two two machines, built from scratch > comprising of Asus P4S8X motherboard, P4 2.4GHz, > with two IBM drives each, one Deskstar 120GXP 82.3GB UDMA100 for the > system drive and one Deskstar180GXP 185.2GB UDMA100 > as a data partition. > They have a clean install of Mandrake 9.0 > [EMAIL PROTECTED] root]# uname -an > Linux morpheus.cissme.com 2.4.19-16mdkcustom #4 SMP > > The drives themselves seem in good shape: > [EMAIL PROTECTED] root]# hdparm /dev/hda > > /dev/hda: > multcount = 16 (on) > IO_support = 1 (32-bit) > unmaskirq = 0 (off) > using_dma = 1 (on) > keepsettings = 0 (off) > readonly = 0 (off) > readahead = 8 (on) > geometry = 10011/255/63, sectors = 160836480, start = 0 > > > [EMAIL PROTECTED] root]# hdparm /dev/hdc > > /dev/hdc: > multcount = 16 (on) > IO_support = 1 (32-bit) > unmaskirq = 0 (off) > using_dma = 1 (on) > keepsettings = 0 (off) > readonly = 0 (off) > readahead = 8 (on) > geometry = 22526/255/63, sectors = 361882080, start = 0 > > and seem to get good throughput: > > [EMAIL PROTECTED] root]# hdparm -tT /dev/hda > > /dev/hda: > Timing buffer-cache reads: 128 MB in 0.28 seconds =457.14 MB/sec > Timing buffered disk reads: 64 MB in 1.40 seconds = 45.71 MB/sec > > [EMAIL PROTECTED] root]# hdparm -tT /dev/hdc > > /dev/hdc: > Timing buffer-cache reads: 128 MB in 0.28 seconds =457.14 MB/sec > Timing buffered disk reads: 64 MB in 1.18 seconds = 54.24 MB/sec > > > Running Bonnie++ also confirms these results and in addition tells me that > i get about 23MB a second on Rewrite speeds. > > All seems pretty good, until i start trying to move any large amounts > of data around on the drives while also having mysql and apache running. > > > One of the systems is still not being used, so i can use it as a test box: > When i copy 1.7GB of data from hda -> hdc, > it doesn't seem like much load: > > [EMAIL PROTECTED] root]# time cp *.gz /space/scratch.backup/ > 0.13user 17.59system 1:44.22elapsed 17%CPU (0avgtext+0avgdata > 0maxresident)k > 0inputs+0outputs (130major+23minor)pagefaults 0swaps > > Output of non-idle mode ps: > > 4:02pm up 8 days, 22:35, 6 users, load average: 0.97, 0.35, 0.12 > 86 processes: 83 sleeping, 3 running, 0 zombie, 0 stopped > CPU states: 0.5% user, 25.2% system, 0.0% nice, 74.1% idle > Mem: 1551472K av, 1521556K used, 29916K free, 0K shrd, 68920K > buff > Swap: 2097136K av, 154508K used, 1942628K free 1299612K > cached > > PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND > 380 root 18 0 512 512 440 R 15.3 0.0 0:17 cp > 421 thorsten 12 0 1044 1044 820 R 0.5 0.0 0:00 top > 383 thorsten 9 0 1816 1816 1628 R 0.0 0.1 0:00 sshd > > > And here is a sample of vmstat 1 while the copy is running: > procs memory swap io system > cpu r b w swpd free buff cache si so bi bo in cs us > sy id 1 1 1 154508 31488 69112 1297788 0 0 16000 36752 989 719 > 0 29 71 0 1 0 154508 31748 69064 1297580 0 0 16128 40648 1023 > 851 0 25 75 1 0 0 154508 31080 69060 1298236 0 0 16128 36692 > 928 695 1 26 73 1 0 0 154508 31044 69068 1298280 0 0 16768 > 36632 998 765 0 28 72 1 0 0 154508 31004 69036 1298588 0 0 > 21888 32592 973 922 0 31 69 0 1 1 154508 31084 68928 1298468 0 > 0 17664 40800 1004 774 2 29 69 1 0 0 154508 31348 68856 1298296 > 0 0 17792 32596 947 766 0 25 75 1 0 0 154508 31012 68892 > 1298652 0 0 20480 32592 958 884 1 26 73 0 1 1 154508 30488 > 68960 1299100 0 0 18048 36692 939 799 0 30 70 1 0 1 154508 > 29728 68968 1299748 0 0 16128 40932 1034 722 1 20 79 0 1 0 > 154508 30192 68940 1299324 0 0 16000 36492 971 834 1 24 75 0 > 1 0 154508 30060 68928 1299464 0 0 16128 38696 933 735 0 23 > 77 0 1 0 154508 29908 68920 1299612 0 0 16128 34544 943 656 0 > 28 72 > > > Howver on the box also running mysql and httpd at the same time: > > [EMAIL PROTECTED] root]# time cp *.gz /space/scratch.backup/ > 0.31user 15.43system 3:27.27elapsed 7%CPU (0avgtext+0avgdata > 0maxresident)k > 0inputs+0outputs (130major+23minor)pagefaults 0swaps > > > non-idle ps output: > > 4:26pm up 27 days, 3:03, 6 users, load average: 7.97, 4.42, 2.42 > 419 processes: 418 sleeping, 1 running, 0 zombie, 0 stopped > CPU states: 18.0% user, 58.9% system, 0.0% nice, 23.0% idle > Mem: 1551472K av, 1520456K used, 31016K free, 0K shrd, 17532K > buff > Swap: 2097136K av, 39756K used, 2057380K free 1129820K > cached > > PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND > 5120 thorsten 14 0 1332 1332 820 R 25.3 0.0 3:23 top > 5548 root 10 0 512 512 440 D 10.9 0.0 0:09 cp > 24561 www 9 0 29456 10M 6172 D 0.0 0.6 0:41 httpd > 5294 mysql 9 0 229M 226M 1184 D 0.0 14.9 0:00 mysqld > > > procs memory swap io system > cpu r b w swpd free buff cache si so bi bo in cs us > sy id 1 8 1 42636 31616 17280 1133684 0 0 10132 15848 729 563 > 11 17 73 0 10 1 42632 31200 17332 1134084 0 0 412 11816 443 > 185 1 5 94 1 6 1 42620 30992 17416 1134344 0 0 15200 10684 > 725 746 5 28 67 3 7 0 42620 32016 17488 1133476 0 0 6276 > 18580 713 422 15 84 1 4 3 1 42620 31060 17496 1134468 0 0 > 6180 8080 604 411 9 91 0 3 5 1 42620 31060 17528 1133944 0 > 0 11952 18676 793 513 14 86 0 2 8 1 42616 30500 17560 1134116 > 32 0 204 13876 497 248 4 17 80 2 8 1 42616 31264 17576 > 1132264 0 0 20576 15700 966 967 30 35 36 1 5 0 42616 31404 > 17648 1133588 0 0 3528 23448 1052 495 13 13 75 0 6 1 42616 > 31332 17676 1133400 0 0 16860 18296 851 740 4 18 78 5 2 0 > 42600 30752 17636 1134312 56 0 5528 14280 646 320 4 76 20 2 3 > 0 42600 31340 17800 1133148 0 0 8264 11276 674 516 16 84 0 > 4 6 0 42600 31352 17808 1132960 0 0 5236 0 593 545 30 70 > 0 2 2 0 42600 32028 17808 1131484 0 0 6772 0 857 806 42 > 57 1 3 3 0 42600 31360 17808 1130852 0 0 3440 0 638 629 > 40 60 0 > > And more extreme, if i move a large directory (4.6GB) > > quick snapshot before move: > 5:12pm up 27 days, 3:49, 6 users, load average: 1.25, 1.08, 1.11 > > > During move: > 5:16pm up 27 days, 3:53, 6 users, load average: 7.57, 3.93, 2.21 > 369 processes: 368 sleeping, 1 running, 0 zombie, 0 stopped > CPU states: 28.7% user, 57.8% system, 0.0% nice, 13.3% idle > Mem: 1551472K av, 1519628K used, 31844K free, 0K shrd, 31804K > buff > Swap: 2097136K av, 48888K used, 2048248K free 1120272K > cached > > PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND > 5120 thorsten 17 0 1332 1332 820 R 25.0 0.0 15:00 top > 6212 root 9 0 688 688 464 D 8.3 0.0 0:16 mv > 6302 mysql 9 0 233M 222M 1500 D 0.6 14.6 0:00 mysqld > 3871 mysql 9 0 233M 222M 1500 D 0.2 14.6 0:08 mysqld > 6257 mysql 9 0 233M 222M 1500 D 0.2 14.6 0:00 mysqld > 6194 mysql 9 0 233M 222M 1500 D 0.1 14.6 0:00 mysqld > 6290 mysql 9 0 233M 222M 1500 D 0.1 14.6 0:00 mysqld > > procs memory swap io system > cpu r b w swpd free buff cache si so bi bo in cs us > sy id 12 5 1 67424 31140 24336 1127380 0 0 12768 8104 915 837 > 36 63 1 3 17 1 67424 32564 24376 1125796 0 0 3956 16064 739 > 431 46 54 0 6 3 0 67552 32324 24348 1124728 0 0 9244 12160 > 753 606 41 58 1 3 11 1 68960 32856 24572 1127092 32 1528 9764 > 14744 970 775 42 58 0 1 16 1 68960 32296 24604 1127500 0 0 > 412 12176 482 178 16 84 0 0 14 0 68960 32808 24620 1127760 0 > 0 260 4980 420 312 10 8 83 2 8 1 69600 30464 24636 1129840 > 20 0 18088 15224 1021 1025 22 32 46 0 16 0 69600 31668 24704 > 1128148 0 0 140 14788 513 208 3 9 88 1 8 0 69600 31892 > 24832 1128680 0 0 412 5636 410 283 2 4 94 2 10 1 69600 > 32488 24996 1128272 0 0 11736 8212 1136 1260 11 39 50 > > > Huh. i'm looking over my figures and nothing looks too terrible to read, > however my real problem is the user front end, as all our web sites rely > on the backed mysql server, which really seems unusable, waiting several > minutes for a page to load. > > i may not be coherent enough to be of much help to myself. :) i've been > trying to track this down for days now, so any one has any advice as to > what to look for next, it would be much appreciated. > > thanks, > thorsten
Offhand I would expect you have hit a condition in a filesystem. You say "move" .... does that mean delete as well? If so and if you are using XFS, that is normal. XFS seems to do something really interesting with deletion, perhaps some form of defragging of open space. if you are using ext2 I would be shocked by this behavior, but the journaling filesystems do have additional overhead. Let's get a little better description of your setup and see if someone can reproduce the behavior. Civileme
Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com