Not particularly worried

The repair shop needed a few more days to finish work on the faster Thinkpad, and so I have kept the Pentium online in its role as torrent slave. In that time I haven’t had much chance to solve the problems it has been having, but I do have a few observations.

First, it appears to be a memory issue that’s caving in the system. I don’t know if I can be 100 percent sure of that, but the system lockups appear to be interspersed with kernel panics nowadays, which are slightly more informative. (I never thought I would be glad to see a kernel panic. 🙄 )

When those kernel panics occur, there’s usually a cascade of killed processes — dhcpcd, agetty and so forth — which might also explain why I sometimes lose a network connection, but nothing else. If the kernel is killing out dhcpcd in hopes of scavenging some memory, then it would make sense that sometimes, bewilderingly, the network disappears while everything else is still running.

Unfortunately, whatever is happening to dhcpcd is also happening to some other part of the system, because killing and restarting dhcpcd is just giving error messages about network hardware, etc. So that route is not quite working.

But the problem is definitely happening when I add too many torrents, or torrents with exceptionally large ISO sizes (in other words, the entire Fedora 10 DVD 😳 ) or add more than five or six torrents at a time. The system also seems to burst when rtorrent checks hashes, which can be rather frustrating since the first thing it needs to do when it starts up after a crash … is re-hash everything. 😐

As a consequence, I’m learning to treat this machine with kid gloves. I seed both the Lowarch torrent and the Ubuntu GTK1.2 Remix torrent from that machine, but much more than that seems to be causing undue stress. Final diagnosis: Sixteen megabytes is probably a bit low to run a torrent slave, under normal circumstances.

It does beg the question of what’s happening with swap space — there’s a generous 256Mb partition available to it, and vm.swappiness is set to 100, and so I was under the impression that things like this would be more or less “solved” by swapping out.

Of course, my understanding of Linux’s behavior under extreme duress (I consider 16Mb to be extreme duress) is rather primitive. It might be that all that network interaction is still too fast for a machine running at 100Mhz, with I/O channels that were cutting-edge in 1993, and the result is a backlog that can’t be solved that way.

Or at least, that’s how I rationalize it. 🙄

In any case, I’m not any closer to solving the problem, but I am learning a bit more about low-end systems. Granted, 16Mb is a bit unusual, but it illustrates how the kernel protects what’s working … when things aren’t quite working.

Moral of the story is to look for a little more memory really. Either that or rely on the machine solely as a seeder of one or two torrents at a time. If I had a little more memory it probably wouldn’t be an issue, but with the machine slated to be reassigned within the next week or so, I find myself in the same place I was when I originally reported these issues … not particularly worried, and not inclined to fix them. 😐