Re: But why is the RAM gone?! [message #27046 is a reply to message #27031] |
Thu, 07 February 2008 12:52 |
HubertD
Messages: 22 Registered: August 2006
|
Junior Member |
|
|
All right, I'm going to use the anticipatory scheduler, just to be on the safe side.
But I don't think it could be a "normal" IO bottleneck as the server has low io-load and, after all, I'm graphing it
Here is a 24h io graph (data generated by "iostat"):
http://www.denkmair.de/ramgone/iostats/
the peaks show periodical rsync-backups of the live data on sda and sdb, the large reading phase on sdc at night is the daily offsite-backup.
Other than those, there is almost no ioload on the server, so waiting io requests could easily be handled.
Concerning the raid driver...
# cat /proc/scsi/ips/0
IBM ServeRAID General Information:
Controller Type : ServeRAID 6M
Memory region : 0xfb000000 (4096 bytes)
Shared memory address : 0xf8802000
IRQ number : 16
BIOS Version : 7.10.18
Firmware Version : 7.10.18
Boot Block Version : 7.10.18
Driver Version : 7.12.05
Driver Build : 761
Max Physical Devices : 30
Max Active Commands : 64
Current Queued Commands : 0
Current Active Commands : 4
Current Queued PT Commands : 0
Current Active PT Commands : 0
This is the stock IBM ServeRAID driver from the kernel source, never used something else...
And with kernel 2.6.16, the machine has had an uptime of >400 days.
Also, the mentioned error messages have been there since day one and did not have any impact on stability.
They are gone now because I deactivated my raid status checks.
But, what's attracting my attention right now:
The displayed driver version differs from the firmware version.
No idea whether that is normal or could be a problem, I'm trying to find out...
|
|
|