[rabbitmq-discuss] Possible memory leak in the management plugin

[Pavel]

Hello,

We are running RabbitMQ 3.1.5 on Erlang R15B.

Last week we had a couple incidents, where RabbitMQ would go from stable
~1Gb RAM to ~10Gb in ~10 mins and keep growing up to the high watermark. It
was quickly identified that mgmt_db is a culprit, so we disabled management
plugin and attempted to reproduce the issue in the lab. Although the lab
test wasn't exactly representing the production setup, we managed to achieve
similar behavior with rapid unbounded growth of mgmt_db using the following
scenario:

- declare 1000 exchanges with 1 queue (maxlength=5) bound to each one
- run a single threaded script that will
   - create 1 connection with 1000 channels
   - publish 1000 messages (1 per channel) to each of exchanges sequentially

Here is the output for the rabbit_mgmt_db process after running test above
for about a minute:
https://gist.github.com/maisenovich/10226474

I'm very new to Erlang, so couple questions about this output: 
Q1: There are three ets tables (as far as I understand) that have million+
items and keep growing rapidly (5754973, 5759070 and 5763167). What are
those?
Q2: Total memory used by ets storage (~120Mb) is far less than total memory
used by rabbit_mgmt_db process (~3.5Gb). Is there any way to inspect what
consumes the most of the memory?

It was noticed also, that in the configuration described above (1000
exchanges x 1 queue each) stats events for publishing channels become very
big. It also seems that increasing collect_statistics_interval from default
5 seconds to 60 seconds helps - memory still goes up after first iteration,
but then stays within ~2Gb (test was running for at least 1 hour with no
crash).

Here is the output of fprof analysis for ~5 seconds of the rabbit_mgmt_db
process during the test at a slower publish rate:
https://gist.github.com/maisenovich/10226881

Q3: Is there any better way to profile Erlang/Rabbit? Attempts to use fprof
under load that is causing the crash didn't work so far.

We are suspecting that the issue is caused by massive channel_stats events
which at the high enough rate causing Erlang GC to start falling behind. The
function invocation below (from linked profile) is particularly raising
questions:

{[{{gb_trees,update,3},                        2892,  138.949,   43.548},      
  {{gb_trees,update_1,3},                      19243,    0.000,   94.664}],     
 { {gb_trees,update_1,3},                      22135,  138.949,  138.212},    
%
 [{garbage_collect,                              60,    0.737,    0.737},      
  {{gb_trees,update_1,3},                      19243,    0.000,   94.664}]}.  

Q4: Any better theories on what might be causing the unbounded memory
growth? Any GC tuning you would recommend to attempt?

Thank you!



--
View this message in context: http://rabbitmq.1065348.n5.nabble.com/Possible-memory-leak-in-the-management-plugin-tp27414p34663.html
Sent from the RabbitMQ mailing list archive at Nabble.com.