Open geoneutrino opened 2 years ago
I think you still need more memory.
unfortunately not, i added lots of additional swap but got the same error, with similar ram peak usage of ~ 200GB. htop in parallel shows no exzessive ram usage
/usr/local/bin/osrm-customize ./planet-latest.osrm [info] Loaded edge based graph: 3038031592 edges, 748175550 nodes [info] Loading partition data took 828.337 seconds [info] RAM: peak bytes used: 205587542016 [error] [exception] std::bad_alloc [error] Please provide more memory or consider using a larger swapfile
hmm. memory errors with lots of ram free - maybe heap related ? (or how to avoid/trace local issue ?)
Can you try building with -DCMAKE_BUILD_TYPE=RelWithDebInfo
and see if we hit any assertions?
Its 9x slower but same effect and message, no more hints/output I cross-checked with a smaller map (Germany) which works without problems. So installation seems ok and this somehow points to a problem with too few resources (but enough free) or some other effect with the planet data
compile only mentions some deprecations (system has boost libs 1.74) but this shouldnt affect the code
include/util/ieee754.hpp:91:18: warning: ISO C++ does not support '__int128' for 'p' [-Wpedantic]
This issue seems to be stale. It will be closed in 30 days if no further activity occurs.
I'm aware of Issue #6122 and complied 5.26 with cmake .. -DCMAKE_BUILD_TYPE=Release
osrm-extract ... works osrm-partition --max-cell-sizes=1024,16384,262144,4194304 ... works
but osrm-customize crashes with [info] RAM: peak bytes used: 205588832256 [error] [exception] std::bad_alloc
6122 mentions this should not happen with Release build
The server is a debian 11 with 256 GB Ram + 256 GB Swap, so no memory pressure
Doing a osrm-contract to switch to CH on the planet.osrm crashes with a Segmentation fault between 50 and 55 %
Any idea how to get more details on this ? Or is there a current problem/howto build the current planet net with bike ?
Thanks