| Handling Dumps in SCO UnixWare |
| These pages give details on what to do if a teamserver running the SCO UnixWare Operating System PANICs, how to obtain diagnostic system dumps and how to package the information for despatch to support. |
Introduction
On rare occasions, your SCO UnixWare Server may "panic" and generate a system dump, after which it will require a reboot. Obviously, this error condition should be investigated. You will therefore need to extract the system dump and send it, together with related files, to your Technical Support representative.
System Panics
If the system reaches an inconsistent state from which it can no longer proceed, it will panic. This means that the system will immediately stop functioning, will save information for diagnostic purposes, and will then require a reboot.
If you are near the system console at the time of a panic, you will see a number of messages scrolling on the console screen, indicating that the system memory image is being saved. This memory image is the system dump; by default, it is initially written to the swap slice (slice 2) on your system disk.
Once the system dump is complete, the system will halt
and await a reboot; normally you will see the console message
displayed.
Rebooting After a Panic
If the system has paniced as described above, and if a s ystem dump has been written to the default area on disk, then you will see the following message on the console when the system reboots:
Do you want to save it ? (y/n)
Assuming you need to save the dump, answer y to this.
| WARNING |
| The default action of dumpcheck is to never time out, but if you edit /etc/default/dump to change the value of TIME to place a time limit on this message, after which the system will reboot anyway, you can lose your chance to save the dump. See KE0028350. |
| WARNING |
| There is a bug whereby, on UnixWare 2.1.x machines with VxVM (part of ODM) installed, the above message may sometimes scroll rapidly and repeatedly up the console screen, accepting no input and making a reboot impossible. If this happens, you have to place the disk in another machine for editing, OR use an emergency recovery disk. There is a fix for this problem. See KE0022779. |
The dumpsave mechanism will ask where you want to save the dump; the options are floppy disk (!), Cartridge Tape if you have one, or file (i.e. disk).
The preferred option is file - if you have enough space in your root filesystem. It is theoretically possible for a system dump to require an amount of disk space equal to your system's physical memory size; however, experience has shown that 100MB of disk space is normally sufficient for systems with up to 512MB of memory.
If you don't have that much space in your root filesystem, you can save to CMT instead, but saving to Exabyte or DAT is only supported as detailed below.
| NOTE |
| On UnixWare 7.0.0
and 7.0.1 systems, repair ptf7070a
can be applied to add the capability to dump to DAT or to DLT. See
KE0028143. The equivalent repair for UnixWare 7.1.0 is erg501138, see KE0028411. The configuration instructions are in KE0028383 |
If the dump is saved to disk, it will be placed in a file /dumpfile, unless that already exists, in which case you will be asked to choose a different filename, which MUST be in the root filesystem.
After saving the dump, allow the system to reboot.
This is part of a six page document, and is continued here.
