Archive for June, 2008
what IBM thinks about HP :)
DO you belive in coinicence? I think that it is part of unfair competition between two giants IBM and HP. so.. what IBM thinks about HP hard drives - that it is toaster 
take a look:
just like this:
original link to: nokaut.pl
DS3200 performance tests #1
today I made some tests at DS3200, I’m fully dissapointed of it’s power, but maybe I do sth wrong? Take a look:
# mount |grep reiser
/dev/sde1 on /tmp/test type reiserfs (rw,noatime,nodiratime)
/dev/sdf1 on /tmp/test2 type reiserfs (rw,noatime,nodiratime)
/dev/sdg1 on /tmp/test3 type reiserfs (rw,noatime,nodiratime)
everything connected from dual controller IBM DS3200 (SAS), with SATA 750GB disks.
/dev/sde1 is raid0 at 4 disks
/dev/sdf1 is raid10 at 4 disks
/dev/sdg1 is raid5 at 4 disks
and the tests:
# dd if=/dev/zero of=/tmp/test/pliczek count=5000 bs=1M 5000+0 records in
5000+0 records out
5242880000 bytes (5.2 GB) copied, 66.6999 s, 78.6 MB/s
# dd if=/dev/zero of=/tmp/test2/pliczek count=5000 bs=1M
5000+0 records in
5000+0 records out
5242880000 bytes (5.2 GB) copied, 191.962 s, 27.3 MB/s
# dd if=/dev/zero of=/tmp/test3/pliczek count=5000 bs=1M
5000+0 records in
5000+0 records out
5242880000 bytes (5.2 GB) copied, 130.41 s, 40.2 MB/s
raid5 looks better than raid10?????
now read tests:
# dd if=/tmp/test/pliczek of=/dev/null count=5000 bs=1M
5000+0 records in
5000+0 records out
5242880000 bytes (5.2 GB) copied, 21.179 s, 248 MB/s
# dd if=/tmp/test2/pliczek of=/dev/null count=5000 bs=1M
5000+0 records in
5000+0 records out
5242880000 bytes (5.2 GB) copied, 33.007 s, 159 MB/s
# dd if=/tmp/test3/pliczek of=/dev/null count=5000 bs=1M
5000+0 records in
5000+0 records out
5242880000 bytes (5.2 GB) copied, 22.7079 s, 231 MB/s
why raid10 is much slower than raid5?
raid 10 at DS3200,3300,3400
In every marketing documentation I can find, that IBM TotalStorage DS3xxx series support RAID10. But reality is not always such nice as the marketing shit. The official ds3200 documenation says that It can do raid 10. Because as you can see below it is not listed:
I tried to do two raid 1 arrays and make a raid 0 from them, but it is imposible… So i have to read and search a lot of IBM Redbooks, and finaly find that:
When RAID level 1 is implemented and the number of drives increases to more than two, RAID level 10 is automatically implemented.
Very good news and it really works:
x3650 pci device resource allocation failure
After plug in the two SAS controllers and RSAII into the IBM SystemX x3650, I would like to set them up. The startup combination was ctrl+c while initialization process of SAS HBA Controller (LSI), and it should make the controller BIOS to appear, but I got for 3 seconds alert:
00180106 10000058 pci device resource allocation failure
then I got:
Please wait, invoking SAS configuration Utility…
I waited about 50 minutes (I went for my lunchtime), and after that - I was sure, that something was wrong!
Google… hmmm forward me to developerworks and this linked to IBM MIGR-65754, answer is not clear:
If an “1801 PCI resource error” occurs, disable “PCI ROM Control Execution” for the second controller. This will free enough ROM space for both controllers to run properly.
Disabling this option is done within system BIOS setup.
After the system is powered on, press F1 when prompted.
Select the following menu choices:
* Advanced Setup
* PCI Settings
* PCI ROM Control ExecutionWithin the “PCI ROM Control Execution” Panel, select the slot number for the secondary controller and disable that slot.
Save the new BIOS settings and restart the server.
First reboot - no difference, second boot, no diference (note that this system reboot last about 10 minutes!), after leaving just one PCI ROM Slot active, everything worked well. But, after installing the OS i noticed that my network interfaces work but with no communication. The reason was that i turned off too much :). Ethernet card has it’s own ROM and it has to be activated.
So what is the reason of this state:
When two or more of the affected controllers are installed, the BIOS for each controller will be allocated system ROM space for execution with default system BIOS settings. It is only necessary for the first controller to load its BIOS into ROM space. This single instance of the respective controllers BIOS will manage both adapters of the same type. By disabling “PCI ROM Control Execution” in the system BIOS for the second controller, less ROM space is used and avoids the BIOS POST event “1801 PCI Resource Allocation Errors” when launching the WebBIOS Configuration Utility.
Various onboard devices and adapter cards can have special option BIOS enabled. During system boot, these device will occupy the system option ROM space. The option ROM space is limited to 128K and if too many devices have their BIOS enabled, then there will not be enough room for all of them. For example the iSCSI TOE adapter needs 46K of the option ROM space during boot (of the adapter) and 10K once the adapter is running. Other devices that may occupy the option ROM space in this configuration are ethernet adapters (including the two onboard NICs) with PXE boot enable, system video and the SCSI hard drives.
files with no valid users
Totay tip:
sometimes in some migrations, reinstall or sth, or even in backup places, you have some files owned by no valid user for current system. It is rather high security issue (in mulituser systems), so you can find every file that does not have a valid user which can be found in /etc/passwd, how?
find / -nouser > no_no_valid_user
glsa-check… securitity in gentoo
what is glsa-check - you can find here http://gentoo-wiki.com/Glsa-check
why to use it - because it shows every possible security hole made by outdated software. Because the output of glsa-check is rather chatty - my own grep method is like this:
glsa-check -d affected |grep -e “Affected package”