On 2011-05-04 07:44, Marko Mäkelä wrote:
The processing time probably won't be reduced if the machine starts swapping. Like Thorsten pointed out, it is good to leave some "breathing room" for the computer.
Sure, but the machine has 8GB ram so it won't have to swap while doing multiple things at the same time and swapping did not take place (according to Munin). I'll run the script again with 5 or 6 GB heap, but I'm not convinced it will change the output or processing time significantly.
I am not sure if the --cache option still works. It was sort of superceded or made unnecessary by the Protobuf input format support.
Oh, if it's not working anymore then I assume it will ignore the option. Or is this too simple thinking?
I do not know. I never used the --cache option myself. One more possibility (a wild guess) is that the format of the cache directory changed and you had some old-format stuff there that was being misinterpreted.
The script does a full cleanup before starting the processing. There is no cache directory lingering around. But I try to remember if I saw a cache directory while processing and I don't think I did, so maybe the option is already ignored. Anyway, I've generated a pbf planet file from the bz2 XML last night, so can test if splitting pbf will be successful this evening.