[OSRM-talk] planet prepare problems

Dennis Luxen luxen at kit.edu
Sat Dec 28 13:31:47 UTC 2013


Grant,

Am 28.12.2013 um 00:46 schrieb Grant Heffernan <grant at mapzen.com>:

> Hi all. Was wondering if anyone has successfully run osrm-prepare on the planet recently. I’ve successfully set up north-america and other extracts, but the planet load fails every time, and in a very odd way, which is making tracking the problem down an issue. More specifically, the host it’s being run on becomes completely unresponsive (I’m running on EC2) and I’m unable to recover the instance to track additional details down. I can terminate but not reboot the instance, which makes me suspect it’s getting starved of resources and the hypervisor is unable to get it to respond. I’m presently testing on the following setup:

Indeed, it looks like your system runs out of resources. Couple of questions:

- Are you using the latest code from develop branch? If not, do so.
- How much disk space is available under /tmp

[…]

> [info] constructing r-tree of 923822461 elements

It is sorting all the data at this stage. This is indeed quite resource intensive, but it shouldn’t die. What does df -h say?

—Dennis




More information about the OSRM-talk mailing list