auf meiner Kiste treten mindestens nach jedem Reboot (nein, kein Hard-Reset ;-) Dateisysteminkonsistenzen (ext3) auf: Beispiel:
Code: Select all
xena:~# fsck.ext3 -fy /dev/lg0/var
e2fsck 1.40-WIP (14-Nov-2006)
Pass 1: Checking inodes, blocks, and sizes
Inode 540819 is in use, but has dtime set. Fix? yes
Inode 540832 is in use, but has dtime set. Fix? yes
Pass 2: Checking directory structure
Entry 'screen.list' in /lib/dpkg/info (458760) has deleted/unused inode 459703. Clear? yes
Entry 'screen.postrm' in /lib/dpkg/info (458760) has deleted/unused inode 459685 . Clear? yes
Entry 'screen.conffiles' in /lib/dpkg/info (458760) has deleted/unused inode 459 688. Clear? yes
Entry 'screen.md5sums' in /lib/dpkg/info (458760) has deleted/unused inode 45969 4. Clear? yes
Entry 'status' in /lib/dpkg (458757) has deleted/unused inode 459664. Clear? ye s
Entry 'templates.dat-old' in /cache/debconf (540673) has deleted/unused inode 37 19175. Clear? yes
Pass 3: Checking directory connectivity
'..' in /tmp/vi.recover (2326530) is /lost+found (11), should be /tmp (2326529).
Fix? yes
Pass 4: Checking reference counts
Inode 65537 ref count is 2, should be 1. Fix? yes
Inode 458923 ref count is 1, should be 2. Fix? yes
Unattached inode 459489
Connect to /lost+found? yes
Inode 459489 ref count is 2, should be 1. Fix? yes
Unattached inode 459495
Connect to /lost+found? yes
Inode 459495 ref count is 2, should be 1. Fix? yes
Unattached inode 459503
Connect to /lost+found? yes
Inode 459503 ref count is 2, should be 1. Fix? yes
Unattached inode 459523
Connect to /lost+found? yes
Inode 459523 ref count is 2, should be 1. Fix? yes
Unattached inode 459666
Connect to /lost+found? yes
Inode 459666 ref count is 2, should be 1. Fix? yes
Unattached inode 459686
Connect to /lost+found? yes
Inode 459686 ref count is 2, should be 1. Fix? yes
Unattached inode 459692
Connect to /lost+found? yes
Inode 459692 ref count is 2, should be 1. Fix? yes
Unattached zero-length inode 540819. Clear? yes
Unattached zero-length inode 540832. Clear? yes
Unattached inode 3719172
Connect to /lost+found? yes
Inode 3719172 ref count is 2, should be 1. Fix? yes
Pass 5: Checking group summary information
Block bitmap differences: +1337 +4960 +4983 +6692 +6783 +6787 +6867 +6869 +6988 +6992 +6996 +7008 -9043 -(9151--9153) -(9155--9162) -927778 -927811 -(931898--9 31955) -(938549--938551) -1102067 +(1150200--1150202) -(1150216--1150238) +(1178 522--1178529) -6934529 +(6936576--6936631) +6942720
Fix? yes
Free blocks count wrong for group #0 (24395, counted=24044).
Fix? yes
Free blocks count wrong for group #3 (32258, counted=32251).
Fix? yes
Free blocks count wrong for group #5 (32261, counted=32251).
Fix? yes
Free blocks count wrong for group #6 (32264, counted=32254).
Fix? yes
Free blocks count wrong for group #24 (32356, counted=32254).
Fix? yes
Free blocks count wrong for group #28 (26166, counted=26099).
Fix? yes
Free blocks count wrong for group #33 (1, counted=102).
Fix? yes
Free blocks count wrong for group #35 (1483, counted=1484).
Fix? yes
Free blocks count wrong for group #36 (836, counted=1155).
Fix? yes
Free blocks count wrong (5350118, counted=5349992).
Fix? yes
Inode bitmap differences: -459024 -459685 -459688 -459694 -459703 -459717 +5408 49 -540853 +3457038
Fix? yes
Free inodes count wrong for group #28 (15426, counted=15416).
Fix? yes
Free inodes count wrong (3920010, counted=3920000).
Fix? yes
/dev/lg0/var: ***** FILE SYSTEM WAS MODIFIED *****
/dev/lg0/var: 12160/3932160 files (1.4% non-contiguous), 2514328/7864320 blocksSporadisch treten auch Input/output-Errors auf (etwa bei einem find /). Es handelt sich hier um Logical Volumes auf einem Software-RAID1. Der Hoster (H******) behauptete auf Nachfrage, einen RAM- und Plattentest durchgeführt und keine Fehler festgestellt zu haben. Das Problem trat bis vor einer Woche nicht auf, Änderungen an der Kernel-Konfiguration oder sonstige "tiefergehenden" Systemeingriffe ich nicht durchgeführt; ein Kernel-Update (von 2.6.19.2-grsec auf 2.6.19.3-grsec) bin ich nach Auftreten des Problems auch einmal gefahren, geändert hat sich nichts.
/proc/mdstat sieht soweit ok aus.
Hat jemand einen Tip für mich, ob bzw. wo es am ehesten Sinn machen würde, mit einer weiteren Suche zu beginnen? Das RAM scheint in Ordnung zu sein, zumindest lief der stress test von mprime gestern einige Stunden lang ohne Fehler. Hat evtl. jemand einen Tip für einen "hard disk stress test"? ;-)


