PDA

View Full Version : "Stale" units eventually fail



Larry Loen
01-14-2002, 08:42 AM
Anyone else seeing this problem? I'm sneakernetting G@H and the typical pattern is this:

1. The problem is so far confined to Linux.
2. I sneakernet a unit as usual.
3. I run maybe a half dozen work units as per normal. They are credited.
4. Then, at random, the client dies. Restarting it gets this mesage:



73
74
75
76
Initializing protein design algorithm
Designing protein sequence 24 of 30
Unrecognized amino acid designation!!!!
17 amino acid number:0


Anyone else seen this? I've had this several times and I already zipped up an example and sent it to Stefan. He has been courteous and prompt in past e-mail exchanges, but I've received no feedback on this from him.

The obvious work-around is to kill the unit and re-download a new one.

Anyone know anything about this?

I'd estimate the outage rate as costing me something like 3 per cent or so from lost work units alone, since one loses, on average, 1/2 of a WU's crunch when redownloading over the old one.

Dyyryath
01-14-2002, 01:14 PM
I've seen it, too. It's happened primarily on my Linux systems, but I'm pretty sure I've had it happen once or twice on my gaming box (which runs XP).

It hasn't happened often enough for me to really start digging for a reason, but I'd be interested to hear what Stefan has to say about it.

dnar
01-14-2002, 08:41 PM
I have experienced this exact error before as well, running the Linux 0.99 client. I would estimate less than 3% failure rate however....