Return to BSD News archive
Xref: sserve comp.unix.wizards:31496 comp.unix.bsd:12821 comp.unix.programmer:12830 Newsgroups: comp.unix.wizards,comp.unix.bsd,comp.unix.programmer Path: sserve!newshost.anu.edu.au!munnari.oz.au!constellation!convex!convex!darwin.sura.net!europa.eng.gtefsd.com!uunet!destroyer!news1.oakland.edu!cs.uiuc.edu!gooshie!sparc17.cs.uiuc.edu!daigle From: daigle@sparc17.cs.uiuc.edu (Russell L. Daigle) Subject: I/O wizards needed for "disk sequencer error" Message-ID: <CFswMC.Mw@sparc0a.cs.uiuc.edu> Sender: news@sparc0a.cs.uiuc.edu Organization: University of Illinois at Urbana-Champaign Date: Mon, 1 Nov 1993 06:56:35 GMT Lines: 39 Greetings! As the subject says, I am getting "disk sequencer errors", and am mostly wandering what that means (and how I can resolve this). For more information (which may not necessarily be needed): I have a disk intensive workload being performed (>10 large makes). This however is only part of the root of the problem. The main problem is as follows: Connected between each processor and main memory (this is a multiprocessor system) is a hardware tracer that stores address traces (amongst other things) generated from the processors. When the hardware tracer buffers fill up, they halt all the processors while they dump the buffers to disk, and then resume the processors. (The halting/restarting is performed such that continuous traces may be obtained.) Here is where the problem comes in. Running the same workload without halting the processors does not cause the disk sequencer error.... only between halting and restarting processors. (BTW, the processors are typically halted for about 3 minutes for every 0.5 seconds of execution.) Hence my explanation is that upon halting the processors, some disk activity is outstanding. While the processors are halted, the data tries to return...but can't since the processors are halted (temporarily). Hence, (by modifying the OS which I have access to) I increased the disk timeout period from 0.5 seconds to 5 minutes. This actually solved one of the problems I was having which was a "timeout error". However, it did not solve the "disk sequencer error". Any suggestions would be greatly appreciated. Thanx, Russ. ---------------------------------------------------------------------- Russell L. Daigle University of Illinois @ Urbana-Champaign Center for Supercomputing Research & Development daigle@cs.uiuc.edu