I recently got lockup using kernel versions 2.1.60 and 2.1.61. (console
dead, no telnet from remote machine possible, ping from remote machine does
respond (!). These lockups appear only sometimes during serving big
documents over http protocol (over ethernet to WinDos 95).
This problems does happen both with NCSA httpd (1.3) and Apache httpd
(1.2.1). Next to this I'm also using squid cache.
Most of the times I was not able to see what happens during the lockup, as I
running X on the console. When a VC console is up; I notice that the system
does not respond to keyboard input; but switching VC's and using alt-sysreq
is possible.
Before problem:
Nov 4 23:47:29 obelix kernel: SysRq: Show Memory
Nov 4 23:47:29 obelix kernel: Mem-info:
Nov 4 23:47:29 obelix kernel: Free pages: 524kB
Nov 4 23:47:29 obelix kernel: ( 1*4kB 1*8kB 6*16kB 7*32kB 3*64kB 0*128kB = 524kB)
Nov 4 23:47:29 obelix kernel: Swap cache: add 150/150, delete 634051/150, find 183/0
Nov 4 23:47:29 obelix kernel: Free swap: 130672kB
Nov 4 23:47:29 obelix kernel: 8192 pages of RAM
Nov 4 23:47:29 obelix kernel: 335 free pages
Nov 4 23:47:29 obelix kernel: 458 reserved pages
Nov 4 23:47:29 obelix kernel: 5739 pages shared
Nov 4 23:47:29 obelix kernel: Buffer memory: 2204kB
Nov 4 23:47:29 obelix kernel: Buffer heads: 2124
Nov 4 23:47:29 obelix kernel: Buffer blocks: 2108
Nov 4 23:47:29 obelix kernel: CLEAN: 274 buffers, 17 used (last=17), 0 locked, 0 protected, 0 dirty
Nov 4 23:47:29 obelix kernel: LOCKED: 1375 buffers, 42 used (last=394), 0 locked, 0 protected, 0 dirty
Nov 4 23:47:29 obelix kernel: DIRTY: 396 buffers, 5 used (last=390), 0 locked, 0 protected, 396 dirty
Nov 4 23:47:29 obelix kernel: Networking buffers in use : 4
Nov 4 23:47:29 obelix kernel: Total network buffer allocations : 11467
Nov 4 23:47:29 obelix kernel: Total failed network buffer allocs : 0
Nov 4 23:47:29 obelix kernel: IP fragment buffer size : 0
After recovered from the lockup; using Alt-Sysreq+SAK/Alt-Sysreq+R on a VC (I was happy
not to have X in the foreground at that time)
Active at the point of lockup task was httpd... (not logged in syslog)
Nov 4 23:56:44 obelix kernel: SysRq: Show Memory
Nov 4 23:56:44 obelix kernel: Mem-info:
Nov 4 23:56:44 obelix kernel: Free pages: 468kB
Nov 4 23:56:44 obelix kernel: ( 91*4kB 13*8kB 0*16kB 0*32kB 0*64kB 0*128kB = 468kB)
Nov 4 23:56:44 obelix kernel: Swap cache: add 150/150, delete 674719/150, find 183/0
Nov 4 23:56:44 obelix kernel: Free swap: 130672kB
Nov 4 23:56:44 obelix kernel: 8192 pages of RAM
Nov 4 23:56:44 obelix kernel: 343 free pages
Nov 4 23:56:44 obelix kernel: 458 reserved pages
Nov 4 23:56:44 obelix kernel: 5926 pages shared
Nov 4 23:56:44 obelix kernel: Buffer memory: 200kB
Nov 4 23:56:44 obelix kernel: Buffer heads: 308
Nov 4 23:56:44 obelix kernel: Buffer blocks: 200
Nov 4 23:56:44 obelix kernel: CLEAN: 43 buffers, 18 used (last=43), 0 locked, 0 protected, 0 dirty
Nov 4 23:56:44 obelix kernel: LOCKED: 127 buffers, 44 used (last=126), 0 locked, 0 protected, 0 dirty
Nov 4 23:56:44 obelix kernel: DIRTY: 3 buffers, 3 used (last=3), 0 locked, 0 protected, 3 dirty
Nov 4 23:56:44 obelix kernel: Networking buffers in use : 165
Nov 4 23:56:44 obelix kernel: Total network buffer allocations : 41264
Nov 4 23:56:44 obelix kernel: Total failed network buffer allocs : 21094055
Nov 4 23:56:44 obelix kernel: IP fragment buffer size : 0
Notice the EXTREME high figure of 'Total failed network buffer allocs'!!!
While there is no big change in free or consumed memory.
Linux 2.1.57 didn't show up this problem to me.
Machine:
Intel P60, 32MB ram
ah2940uw scsi disk
3C509 ethernet
Are other people also experiencing this behaviour? Any clues where to look
for it, or what to fix?
Regards
Frank.
========================---------------->
#define NAME "Frank van de Pol"
#define ADDRESS "mgr. Nelislaan 10"
#define CITY "4741 AB Hoeven"
#define COUNTRY "The Netherlands"
#define EMAIL "F.K.W.van.de.Pol@inter.NL.net
Linux - Why use Windows, since there is a door?