[erlang-questions] Windows OS crash dump at 2GB (64bit OTP)

[Eric Cotter]

Hello,

I am using the Windows 64 bit distribution of the latest Elrang vm.
I am using RabbitMQ's Windows installation as well.

When I start load testing RabbitMQ I'm sending a modest volume of messages
onto RabbitMQ 10,000 messages avg size <  1K.  When I get about 8K onboard
i'm running just about 2GB of Ram and the Erlang VM crashes.

I'm using the C# NuGet packgage via NServiceBus to make all of my calls
onto RabbitMQ transport.

I am not using Erlang at this time for anything other then RabbitMQ and
plan on moving it eventually to CentOS.   For now I need to get RabbitMQ
and Erlang to be able to stay up when  more then 2GB  of memory is
allocated.

I have included the crashdump on the machine in QA that is crashing.  I'm
using AppDymamics to stress the system.

8 cores 3.3ghz
25 GB ram
18.7 Terabytes drive
Windows 64bit server.


---------------------------
=erl_crash_dump:0.3
Tue Sep 02 16:11:36 2014
Slogan: eheap_alloc: Cannot allocate 762886488 bytes of memory (of type
"heap").
System version: Erlang/OTP 17 [erts-6.0] [64-bit] [smp:4:4]
[async-threads:30]
Compiled: Mon Apr  7 19:35:35 2014
Taints:
Atoms: 23356
=memory
total: 1972553664
processes: 1446544768
processes_used: 1446486544
system: 526008896
atom: 703377
atom_used: 685539
binary: 476176112
code: 20066158
ets: 24685176
=hash_table:atom_tab
size: 19289
used: 13572
objs: 23356
depth: 8
=index_table:atom_tab
size: 23552
limit: 1048576
entries: 23356
=hash_table:module_code
size: 797
used: 476
objs: 565
depth: 3
=index_table:module_code
size: 1024
limit: 65536
entries: 565
=hash_table:export_list
size: 6421
used: 4963
objs: 9312
depth: 7
=index_table:export_list
size: 10240
limit: 524288
entries: 9312
=hash_table:export_list
size: 6421
used: 4955
objs: 9285
depth: 7
=hash_table:process_reg
size: 97
used: 73
objs: 120
depth: 5
=hash_table:fun_table
size: 3203
used: 2257
objs: 4143
depth: 6
=hash_table:node_table
size: 5
used: 1
objs: 1
depth: 1
=hash_table:dist_table
size: 5
used: 1
objs: 1
depth: 1
=allocated_areas
sys_misc: 37549
static: 524928
atom_space: 360536 342698
atom_table: 342841
module_table: 205008
export_table: 400212
export_list: 1638912
register_table: 868
fun_table: 25714
module_refs: 27072
loaded_code: 17769240
dist_table: 547
node_table: 179
bits_bufs_size: 0
bif_timer: 99064
link_lh: 0
process_table: 12582912
port_table: 98304
ets_misc: 49192
=allocator:sys_alloc
option e: true
option m: libc
=allocator:temp_alloc[0]
versions: 2.1 3.0
option e: true
option t: 5
option ramv: false
option sbct: 524288
option rsbcmt: 80
option rmbcmt: 100
option mmbcs: 131072
option lmbcs: 10485760
option smbcs: 1048576
option mbcgs: 10
option acul: 0
option mbsd: 3
option as: gf
mbcs blocks: 0 0 4
mbcs blocks size: 0 0 9968
mbcs carriers: 1 1 1
mbcs sys_alloc carriers: 1
mbcs carriers size: 131200 131200 131200
mbcs sys_alloc carriers size: 131200
sbcs blocks: 0 0 0
sbcs blocks size: 0 0 0
sbcs carriers: 0 0 0
sbcs sys_alloc carriers: 0
sbcs carriers size: 0 0 0
sbcs sys_alloc carriers size: 0
temp_alloc calls: 8663
temp_free calls: 8663
temp_realloc calls: 1
sys_alloc calls: 1
sys_free calls: 0
sys_realloc calls: 0
=allocator:temp_alloc[1]
--------------------------------

The actual crashdump  is 25mb so I am not including it entirely.
I can see the "eheap_alloc: Cannot allocate 762886488 bytes of memory (of
type "heap")."   But that is not even a gig.

Any guidance would be greatly appreciated.

Thank you very much,

Eric Cotter
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://erlang.org/pipermail/erlang-questions/attachments/20140905/afc6593a/attachment.htm>