Hardware and Software Setup
From SPSU Wiki
Contents |
Software
Linux Counter
2005-07-27: Registered this machine with http://counter.li.org as 296682 and installed the automagic machine-update script.
NTP (Network Time Protocol)
2005-07-28: NTP was not installed by default. Downloaded it from RHN with "up2date" and changed /etc/ntp.conf to point to einstein.spsu.edu. Einstein seems to be about 40 seconds off from USNO, although it claims to be a stratum 2 server. External time servers, such as the public servers operated by ntp.org, are firewalled off for reasons that have never been well explained. Anyway, ntpd is started, and set to restart at run levels 3 and 5 on reboot.
PowerChute
2005-07-27: Downloaded, but have not yet installed: ftp.apcc.com/restricted/software/pcbe/704/linux/pbeagent-7.0.4-114.i386.rpm
RAID Controller Software
2005-07-30: However, afacli required /dev/afa0, and RHEL 4 creates the contents of the /dev directory automatically. /dev/afa0 was installed by installing afacli, but disappeared on reboot. That is discussed here: https://www.redhat.com/archives/nahant-list/2005-June/msg00125.html For the moment, /dev/afa0 can be recreated by removing and re-installing afaapps-2.7-1. That drops a script, MAKEDEV.afa, into /dev; the script gets removed on reboot, too! There's now a copy of it in /usr/local/bin that's executed on startup by rc.local. This needs to be fixed right, and has been added to the To-Do list.
2005-07-27: The Red Hat Enterprise Linux Server software includes kernel support for Adaptec "aacraid" RAID controllers, but no auxiliary software. Downloaded http://ftp.us.dell.com/scsi-raid/afa-linux-app-A01.tar.gz and used rpm to install afaapps-2.7-1.i386.rpm, which gave me the "afacli" command line interface to the disk controller. That let me install the "raidwatch.sh" script for root, which I got from here: http://linux.dell.com/files/aacraid/aacraid_monitoring_script.txt. The script will watch for changes in the RAID configuration, presumably including failed drives, and send email. Good. Documentation is here: http://support.dell.com/support/edocs/storage/RAID/
Sendmail
2005-07-25: Sendmail needs to be running as a daemon for mail confirmations and other email functions to work. Started sendmail with "service sendmail start" then used "chkconfig" to start sendmail at run levels 3 and 5.
Hardware
Test Disks and Build RAID Container
Format all four disks using the built-in utilities on the controller, accessed with Control-A during boot.
With RAID disabled, run the stand-alone Dell diagnostics (booted from floppy) with user prompting disabled; test SCSI subsystem and disks. This takes about twelve hours. (All OK.)
In setup (F2 at boot) enable RAID controller, accept change in configuration at reboot, then enter controller setup with Control-A. Initialize all four disks for RAID, create a RAID-5 container with disks 0-2 and "scrub" it. Add disk 3 as a failover drive.
Install the Dell Utility Partition
Dell server-class machines have a special partition from which to run diagnostics. In reality, it's an MS-DOS partition with a special identifier, 0xDE, and the Dell diagnostics programs. It is explained here: http://www.goodells.net/dellutility/
Here are the steps to create and populate a utility partition that will respond to F10 during bootstrapping:
- Boot from an MS-DOS disk that contains the usual MS-DOS utility programs and a copy of the Ranish Partition Manager, part.exe. For this project, a DOS 6.22 disk was prepared.
- Use FDISK to create a DOS partition of 60-100 MB at the front of
the disk (or at the front of the RAID container for PERC systems.) If you have doubts about the MBR, also run FDISK /MBR to rewrite it.
- The DOS partition will become drive C. FORMAT /S c: to format it. You may use any volume label. DELLDIAG is one possibility.
- Copy everything from the boot disk to c:.
- ERASE C:AUTOEXEC.BAT and then RENAME C:AUTOEXEC.NEW AUTOEXEC.BAT.
- Copy everything from 3 Dell Diagnostic Programs floppies to C:.
- From A:, run PART.EXE (instructions in PART.TXT) and change the
partition type to DE. Set the partition inactive. Ignore the warning that there are no active partitions.
- Reboot and press F10. You should see the menu from AUTOEXEC.BAT
that you just put on C:. Now you can run DELLDIAG. Good.
Be careful not to overwrite the MBR or F10 will no longer work. In particular, do not put a boot loader like GRUB in the MBR.
