[Geocoding] Nominatim Install Problems
gergely.hanti.reg at gmail.com
Thu Feb 24 14:27:11 GMT 2011
I am experiencing the same problem with Debian, PostgreSQL 8.3.12, PostGIS
1.3.3 and 1.5.2, with a two months old osm2pgqsl version.
I have been testing it on two dedicated hardware configurations. The
stronger one is a System x3650 M2 with 24GB Ram.
I am trying to load the Europe extract. The oom-killer kills the postgre
process at 70% during the loaddata phase.
- With the weaker hardware config (with 4Gb Ram) it happened much earlier.
- Updating PostGIS didn't help. The process stopped nearly at the same
- I have tried setting Debian's overcommit mode to strict (as was suggested
but this only resulted that the postgresql stopped the process and not the
- I could process smaller extracts without any problems.
At first I thought that this was a pure pg issue. That is why I had asked
for help at the pg forum (
http://www.pg-forum.de/konfiguration/4971-out-memory-linux.html), but then I
have found this thread. I hope somebody can help me here to solve this
Here are some further informations:
For the stronger hardware config with 24Gb Ram:
I've set Debian's SHMMAX to 8GB = 8*1024*1024*1024 = 8589934592, and SHMALL
to SHMMAX/4096 = 2097152
< shared_buffers = 6GB
> shared_buffers = 32MB
< work_mem = 256MB
< maintenance_work_mem = 256MB
> #work_mem = 1MB
> #maintenance_work_mem = 16MB
< max_fsm_pages = 1592000
> max_fsm_pages = 204800
< fsync = off
> #fsync = on
< wal_buffers = 24MB
> #wal_buffers = 64kB
< checkpoint_segments = 20
> #checkpoint_segments = 3
< checkpoint_completion_target = 0.9
> #checkpoint_completion_target = 0.5
< random_page_cost = 1.5
> #random_page_cost = 4.0
< effective_cache_size = 20GB
> #effective_cache_size = 128MB
< autovacuum = off
> #autovacuum = on
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Geocoding