We are using routing, ip firewalling and masqurading. Most network stuff
in the kernel is enabled, inclusive SYN/RST cookies.
Sometimes the machine becomes extremly slow with great packet losses. It
can take several hours just to login and reboot it.
Here is something I found that might help. There is over 50MB allocated in
network buffers. Is that normal?
Aug 6 13:30:56 brandstrup kernel: Mem-info:
Aug 6 13:30:56 brandstrup kernel: Free pages: 1956kB
Aug 6 13:30:56 brandstrup kernel: ( 293*4kB 46*8kB 6*16kB 4*32kB 1*64kB
1*128k B = 1956kB)
Aug 6 13:30:56 brandstrup kernel: Swap cache: add 0/0, delete 2901041/0,
find 19/0
Aug 6 13:30:56 brandstrup kernel: Free swap: 66468kB
Aug 6 13:30:56 brandstrup kernel: 8192 pages of RAM
Aug 6 13:30:56 brandstrup kernel: 497 free pages
Aug 6 13:30:56 brandstrup kernel: 423 reserved pages
Aug 6 13:30:56 brandstrup kernel: 1324 pages shared
Aug 6 13:30:56 brandstrup kernel: Buffer memory: 8160kB
Aug 6 13:30:56 brandstrup kernel: Buffer heads: 8164
Aug 6 13:30:56 brandstrup kernel: Buffer blocks: 8160
Aug 6 13:30:56 brandstrup kernel: CLEAN: 6861 buffers, 1 used
(last=1), 0 locked, 0 protected, 0 dirty
Aug 6 13:30:56 brandstrup kernel: LOCKED: 1292 buffers, 24 used
(last=1259), 0 locked, 0 protected, 0 dirty
Aug 6 13:30:56 brandstrup kernel: Networking buffers in use : 1
Aug 6 13:30:56 brandstrup kernel: Network buffers locked by drivers : 0
Aug 6 13:30:56 brandstrup kernel: Total network buffer allocations :
50559043
Aug 6 13:30:56 brandstrup kernel: Total failed network buffer allocs : 0
Aug 6 13:30:56 brandstrup kernel: Total free while locked events : 0
Aug 6 13:30:56 brandstrup kernel: IP fragment buffer size : 0
Baldur