root cause analysis

[rone]

After removing the suspect drive and suffering more crashes, all while i was trying to transfer files to and from drives, i determined that the culprit was the goddamned primary 80-wire UDMA cable.

I ended up moving some files instead of copying them from drive to drive, so in returning to my original drive, i was missing some stuff, most importantly some log files; syslogd (the program that writes system logs) is still stupid in that if it tries to log to a file and realizes that it's not there, it will fail with a whine instead of creating the file.  This is indisputably broken behavior and i don't know why we put up with it this way.

The worst of all this was that i blew away my mail server configuration.  Not a big deal, though, just a moment of panic.  There's also some brokenness in my news spool, but i think that'll sort itself out after tonight's expiration run.

Comments

[ltempt.livejournal.com]

haha! IDE will always be broken. One thing that can be said for FCAL is the cabling is neat and simple.

[wisn.livejournal.com]

I still think it's bizarre and inexplicable that an IDE cable will work one way in one direction, and another way in another direction, despite a fairly detailed and accurate-seeming description by somebody who, in another and unrelated conversation, ranted at length about why people who can distinguish between expensive high-end speaker cables and zipwire by sound alone are self-deluding.

[ptomblin-lj.livejournal.com]

I find it bizarre and inexplicable that somebody can't tell that digital signals travelling through an 80 wire cable might be just a wee tad more complicated and subject to one way effects than an analog level travelling through a single wire. In the 80 wire cable, different wires do different things - I wouldn't be at all surprised if some of the wires carry signals going one way and some carry signals going the other, and/or some carry commands and status and some carry data.

I'll tell you I was surprised as hell when, in spite of the fact that 40 of the wires aren't even connected to the connector and are only there for grounding and cross talk disappation, a break in one of the unconnected wires once caused me some bizarre and inexplicable data corruption, and one of the drive manufacturer's disk test programs was able to diagnose that as the cause of the problem.

[baljemmett.livejournal.com]

I seem to remember that a requirement for these 80-pin cables is that they support Cable Select addressing -- in which case the wiring looks different depending on which end of the cable you start at, too.