System crash without any warnings on log

[Sk Noorul Hassan]

Hi,

I have three VAX 4105A servers in Duty,Hot & Warm mode running SCADA application in an area.A process which runs in all the three servers monitors Duty check every 60 secs(one duty should be in network) and Data Disk check every 30 secs.So, if there are two duties due to network blip, this process will crash the old duty immediately. In second case, it will issue a read request to external array(BA356) data disks and if does not get response from disk within 30 secs, it will crash the system as there are no data update for 30 secs.

In recent past three different servers from different area have crashed due to disk read error. So, is there any problem with array controller that my process is unable to read the disk in this case and crashed the server.
Please suggest ...

[Uwe Zessin]

It sounds like the system is doing a voluntary 'crash' (whatever you mean by crash) - how does this 'crash' look like? If this is not an operating system bugcheck, then you will loose the contents of the errorlog buffers in main memory, which are normally saved as part of the crash dump.

I am afraid you cannot really work around the problem of defective hardware. I beleive the default timeout for a SCSI disk happens to be 30 seconds and hardcoded in the driver. The only 'fix' I see is to replace defective or unreliably (disk read errors) hardware.