Re: [2.6.10-rc1 and prev] System unuseable while writing to disk

From: Prakash K. Cheemplavam
Date: Thu Nov 11 2004 - 02:09:17 EST


Bill Davidsen wrote:
Prakash K. Cheemplavam wrote:

Hi,

I have a problem which doesn't seem to be connected to the i/o schedulers, because all I tested (cfq, deadline, noop) show the same:

While writing (when the kernel actually commits to hd) my system gets very unresponsive esp when another app I want to use wants to write (read?) from hd, as well. This is *not* a UDMA problem (at least no apparent...)! More specific:

I wrote this primitive code for writing sequentially:


May I suggest running "vmstat 1" while this is happening? Looking at the waitio time vs. transfer rates might reveal something. If it all looks

Here you go. I hope you can see anything. (I attached the file as well...):


procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
r b swpd free buff cache si so bi bo in cs us sy id wa
0 0 0 775044 14520 136460 0 0 935 86 1251 909 13 5 57 25
0 0 0 775036 14520 136460 0 0 0 0 1270 536 1 0 99 0
0 0 0 775036 14520 136460 0 0 0 0 1225 655 1 1 98 0
0 0 0 775036 14520 136460 0 0 0 0 1187 530 1 0 99 0
0 0 0 775036 14520 136460 0 0 0 0 1184 517 0 0 100 0
0 0 0 775036 14520 136460 0 0 0 0 1188 548 1 1 98 0
0 0 0 774876 14532 136460 0 0 12 0 1185 681 2 1 94 3
0 0 0 774876 14532 136460 0 0 0 0 1185 479 0 0 100 0
0 0 0 774876 14532 136460 0 0 0 0 1180 436 1 0 99 0
1 1 0 571740 14732 333736 0 0 32 42628 1325 633 3 43 52 2
2 2 0 217812 15072 676648 0 0 0 45184 1523 1771 14 78 0 8
1 2 0 220436 15072 676648 0 0 0 49664 1582 1365 9 6 0 85
0 2 0 85460 15984 808012 0 0 164 73984 1487 1143 18 32 0 49
0 1 0 85716 15984 808140 0 0 0 29056 1493 1381 4 6 0 90
0 1 0 86100 15984 808140 0 0 0 40320 1420 1367 3 2 0 95
0 1 0 86100 15984 808140 0 0 0 44416 1386 693 0 2 0 98
0 1 0 86100 15984 808140 0 0 0 50304 1518 1413 5 6 0 89
0 1 0 86164 15984 808140 0 0 0 28544 1452 827 1 2 0 97
0 1 0 86292 15984 808140 0 0 0 35840 1386 809 1 2 0 97
0 1 0 86356 15984 808140 0 0 0 37376 1450 1373 3 3 0 94
procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
r b swpd free buff cache si so bi bo in cs us sy id wa
1 1 0 86484 15984 808140 0 0 0 42880 1752 2494 7 6 0 87
0 1 0 86484 15984 808140 0 0 0 31872 1608 2098 7 4 0 89
0 1 0 86484 15984 808140 0 0 0 45952 1530 795 1 2 0 97
0 1 0 86484 15984 808140 0 0 0 44288 1515 814 0 3 0 97
0 1 0 90580 15984 808140 0 0 0 29052 1471 1365 5 5 0 90
1 0 0 90148 16108 809356 0 0 976 2088 1383 1319 22 11 0 67
0 0 0 90148 16108 809360 0 0 4 16 1444 833 10 5 84 1
0 0 0 90148 16108 809360 0 0 0 0 1256 473 0 0 100 0
0 0 0 90148 16108 809360 0 0 0 0 1379 611 3 1 96 0
0 0 0 90148 16108 809360 0 0 0 0 1526 2439 16 5 79 0
0 0 0 90148 16164 809360 0 0 0 120 1563 1824 16 4 80 0


Beginning and end the system is idle and then I start writing down the file. During that time thunderbird gets unusable.

I can't say that I see any such thing with ext[23], so it may be a reiser issue and someone else will have to help. Did you look at the logs to see that there are no useful warnings there?

Nothing in the logs, unfortunately.

Cheers,

Prakash

Attachment: vmstat.txt
Description: application/pgp-keys

Attachment: signature.asc
Description: OpenPGP digital signature