osrm-backend: osrm-extract for planet on EC2 hs1.8xlarge: "External memory block allocation error"

Still trying to see if I can get OSRM working for the planet… This time I decided to get serious and started up an hs1.8xlarge instance on AWS with ~117 GB RAM, running Ubuntu 12.04.

After a few minutes, osrm-extract fails:

$ ./run_osrm-extract.sh                                                                                         
Wed Nov 13 19:31:33 UTC 2013
[info] Input file: planet-131106.osm.pbf
[info] Profile: profile.lua
[info] Threads: 8
[info] Using script profile.lua
[STXXL-MSG] STXXL v1.3.1 (release)
[STXXL-MSG] 1 disks are allocated, total space: 25000 MiB
[info] Using turn restrictions
[info] Found 3 exceptions to turn restriction
[info]    motorcar
[info]    motor_vehicle
[info]    vehicle
[info] Parsing in progress..
[STXXL-ERRMSG] External memory block allocation error: 2097152 bytes requested, 0 bytes free. Trying to extend the external memory space...
[STXXL-ERRMSG] External memory block allocation error: 2097152 bytes requested, 0 bytes free. Trying to extend the external memory space...
...

Here are the contents of my .stxxl file (/home/ubuntu/osrm_ebs_volume is a 100 GB EBS volume that I attached to my EC2 instance):

$ cat .stxxl
disk=/home/ubuntu/osrm_ebs_volume/osrm/stxxl,25000,syscall

I also tried using a .stxxl capacity of 40000 which resulted in the same error.

You can see here that I have enough disk space for the stxxl file:

$ pwd
/home/ubuntu/osrm_ebs_volume/osrm
$ ls -lh stxxl 
-rw-rw---- 1 ubuntu ubuntu 25G Nov 13 20:25 stxxl
$ df -h
Filesystem      Size  Used Avail Use% Mounted on
/dev/xvda1      7.9G  1.5G  6.1G  20% /
udev             58G   12K   58G   1% /dev
tmpfs            23G  256K   23G   1% /run
none            5.0M     0  5.0M   0% /run/lock
none             58G     0   58G   0% /run/shm
/dev/xvdb2       99G   47G   47G  51% /home/ubuntu/osrm_ebs_volume

Why would osrm-extract fail like this considering that I have enough memory and disk space?