This is the information that we think that the operations staff currently needs to run the machine.

Today Gnosis does I/O only to disks and the Tymnet base and on rare occasions to tape.

If Gnosis is running on a virtual machine it is important to know that the Gnosis kernel notes disk errors and performs transient error recovery since it does SIO's to the disks.

The kernel provides error information to a program via the EREP key {(erep)}.

Gnosis does record data on different drives redundantly and can thus survive a dead drive. When the pack is brought online again, redundancy is automatically reestablished by rewriting the pack. This has caused severe temporary performance impairment.

The two drives of Gnosis should be on different channels or at least different controllers when possible.

Gnosis does checkpoints every few minutes. When Gnosis is IPL'd, it restarts from the most recent checkpoint. This happens without operator intervention. Some users have a key called "CHECKPOINT". If so the command "KEYCALL CHECKPOINT 0" causes a checkpoint to be done. If this is done just before taking the system down then everyone's last few minutes of work will be saved.

Gnosis under VM

The user name "GNOSIS2" runs Gnosis on system 85. It is setup to be "VLOG'd" when CP is brought up. It uses 1 meg V=R on 85 in Dallas.

An enabled wait state PSW = '003CXC00 000000AA' is the normal state of the CPU when it has no work.

000A00000B01 arises in GRANGET concerning failure to find swapping area.

000A00000B03 arises in IDIRECT about trouble in building directory entries.

The disabled wait state IA = '00000C01' arises upon machine check during IPL.

The disabled wait state PSW '00000C02' arises when the IPL code finds a disk format code that it does not understand.

The disabled wait states PSW '00000C03', '00000C04', '00000C05' and '00000C06' arise when the IPL code can't find an IPL range on the pack. I.e., no kernel! Presumably some system programmer goofed. You may be able to IPL from another of the disk packs.

The disabled wait state PSW '00000C07' arises upon correctable data checks or other unusual I/O status codes in the IPL code. It might suffice to move the pack to another drive. Otherwise IPL'ing from another pack may be necessary.

A PSW ending in 00000D0A is caused when the kernel can not take a dump. General register 15 (an error code) should be recorded. This has occurred in the past when all disk controllers were powered down. It also occurs if critical packs are inaccessible to the machine. See (d0a) for error code explanations.

0000D07 - permanent error on the IPL device.