New server configuration Diary

The new server configuration diary seems to have not yet developed the habit of writing a blog. As a result, I have thought about several articles a few weeks ago and I am not in the mood to summarize them. I want to stick to it... in the past few weeks, the Community (http://oss.lzu.edu.cn) bought a new server to replace the old one. The storage on the server is very powerful, with tens of thousands to 450 GB x 2 in the west, 1 Tb x 6 in the West black disk, and two array cards (rocketRAID 2720, don't laugh, the budget is not enough, A good RAID card is too expensive. It can only be used in this way). Then, we can make two slave G into RAID 1, which is used to store key data of the system. Six pieces of 1 t are used as RAID 5 to store other data. Before configuring the server, you will know that the system is difficult to install. Because the kernel does not support the driver of the RAID card, you need to manually enter it. But what about it? Stick it to your head! First, create a system boot disk (here is Centos 6.4 x86_64), start the system, and click Install. As a result, the two raid files are not identified when the partition is reached. Well, I guess. First download the official driver and installation manual: Drivers. No way. I looked at the specific installation script and started the manual installation. The following are detailed steps: 1. Prepare two U disks, where disk A is the system boot image (centos-mininal is recommended) and disk B is the storage drive package. 2. Insert disk A and disk B to the server, start the server, press F8 to enter the startup Item Selection, and select disk A to enter the character installation interface. 3. After selecting the system language and keyboard type, the installer enters the graphical interface. In this case, do not click the next step. Press Ctrl + Alt + F2 to enter the terminal and mount disk B. Assume that disk B is mounted under/drv. 4. unzip the driver package to the current directory: tar-xvf ***. tgz 5. go to the decompressed directory and run the following command: gzip-dc modules. cgz | cpio-idum 6. then, you will get a rr272x_1x.ko driver file, and then run the following command to load the driver: modprobe sd_mod insmod rr272x_1x.ko 7. after the driver is loaded successfully, you can switch back to the graphical installation interface. if the system is correctly installed, it will be OK. After the system is installed, do not restart the system. Press Ctrl + Alt + F2 to enter the terminal again. 8. copy all your previously decompressed driver packages to/mnt/sysimage/tmp and run the following command: chroot/mnt/sysimage 9. run the following command after chroot: mv/boot/initramfs-'uname-R '. img/boot/initramfs-'uname-R '. img. save cp rr272x_1x.ko/lib/modules/'uname-R'/kernel/driver/scsi/depmod-a dracut -- add-drivers rr272x_1x.ko-f/boot/initramfs-'uname-r '. img 10. then you can mount the U disk before umount, and then restart it so that you can normally identify the array card. Note: This section only describes how to insert the array card driver when installing the system. To upgrade the kernel, follow the steps above to re-mount the array card driver to the initramfs file. Finally, note that if lvm is used during partitioning, the boot partition must be placed out of lvm because the existing grub does not recognize lvm, then the system will not be able to get up. Google, it seems that the new version of grub can recognize lvm, but I do not dare to use it on the server. It is better to be honest. Now that the system has been set up, let's give it a performance test and a stress test! To begin. But I don't want to manually run lm-sensors every time to get the temperature. To know that the stress test will last for a day or two, write a script as follows: [plain] while true do sleep 30 time = 'date + % t' tmp = 'sensors | awk-F ":" '/temp/{print substr ($2, index ($2, "+") + 1, 7)} ''for value in $ tmp do time = $ time ", "$ value done echo $ time done then installs a screen to prevent me from exiting the remote session and killing the test process. All right, everything is ready. Start running: First enable the temperature record Script: sh get_temp.sh & then start running ltp :. /runltp-c 100-x 2-D 100,1000, 0-m 100,1000, 12000,0-d/var-p-l result. log-g result.html can then wait for the result. Here we will not describe the usage of ltp. For more information, see http://ltp.sourceforge.net/documentation/how-to/ltp.php. To test the performance of reading and writing small file blocks on a hard disk, use dd: dd if =/dev/zero of =/var/test bs = 10 k count = 10240, you can forcibly skip the hard disk cache and directly write it to the hard disk to check its performance. You can add conv = fsync after dd. Of course, if you look at the io wait rate, sar is indispensable. I will not elaborate on it here!