[rabbitmq-discuss] bug21673 crashed with very large transaction

mabrek mabrek at gmail.com
Mon Mar 29 16:05:03 BST 2010


Hello.
I've got a crash with new persister and very large transaction. How to
reproduce: open transaction and send large amount of messages without
committing it. Rabbitmq process will use all available memory and then
crash. I tried to send 2000000 messages, but rabbit died at 800000.

Interesting pieces from rabbitmq.log:
=INFO REPORT==== 29-Mar-2010::18:32:54 ===
vm_memory_high_watermark set. Memory used:1682971336 allowed:1648679321

=INFO REPORT==== 29-Mar-2010::18:32:54 ===
    alarm_handler: {set,{vm_memory_high_watermark,[]}}

=INFO REPORT==== 29-Mar-2010::18:33:07 ===
vm_memory_high_watermark clear. Memory used:1337501072 allowed:1648679321

=INFO REPORT==== 29-Mar-2010::18:33:07 ===
    alarm_handler: {clear,vm_memory_high_watermark}

=INFO REPORT==== 29-Mar-2010::18:33:10 ===
vm_memory_high_watermark set. Memory used:2280979328 allowed:1648679321

=INFO REPORT==== 29-Mar-2010::18:33:10 ===
    alarm_handler: {set,{vm_memory_high_watermark,[]}}

=INFO REPORT==== 29-Mar-2010::18:33:13 ===
vm_memory_high_watermark clear. Memory used:1542490912 allowed:1648679321

=INFO REPORT==== 29-Mar-2010::18:33:13 ===
    alarm_handler: {clear,vm_memory_high_watermark}

alarm was set and cleared several times.

startup_err contains:
Erlang has closed
Crash dump was written to: erl_crash.dump
eheap_alloc: Cannot allocate 1167696400 bytes of memory (of type "old_heap").

Regards,
Anton Lebedevich.




More information about the rabbitmq-discuss mailing list