michael goulish created DISPATCH-64:
---------------------------------------

             Summary: slow or sporadic memory leak
                 Key: DISPATCH-64
                 URL: https://issues.apache.org/jira/browse/DISPATCH-64
             Project: Qpid Dispatch
          Issue Type: Bug
            Reporter: michael goulish


In long-term soak tests, I am seeing router mem grow by 1 megabyte every 4 or 5 
minutes.


Test setup
=======================================
  1. single router on one box
  2. 10 senders, 10 receivers on separate box.
  3. each client handles 100 unique addresses.
  4. while test is running, I run 'top' in a loop to see router memory usage 
(resident set size).  I also run "qdstat -m" in a loop, to see router's report 
on usage of various data structures.
  5. clients all have single connection for duration of test.
  6. clients start once at beginning of test and do not stop until end.  No new 
clients are started after the beginning.
  7. no clients failed during the test.
  8. no new addresses were added after test startup.


Observations
=========================================
1. During a 64 minute period which started at least 15 minutes after the 
beginning of the test,  memory usage (resident set size) as measured by 'top' 
grew from 96 to 109 megabytes.

2. Some of the data types reported by 'qdstat -m'  increased.  Here is the 
list:  (using numbers from the 'total' column of qdstat report. )

    qd_connection_t        832 -->   896
    qd_hash_handle_t      1408 -->  1600
    qd_hash_item_t        1408 -->  1600
    qd_link_t             1536 -->  1664
    qd_log_entry_t        1152 -->  1216
    qd_message_content_t 10256 --> 10272
    qd_parsed_field_t      448 -->  1024
    qd_router_link_ref_t  1408 -->  1600
    qd_router_link_t      1536 -->  1664


3. The data structures that increased did *not* increase smoothly.  For 
example, qd_hash_handle_t and qd_hash_item_t remained constant for 6 minutes 
before increasing.





--
This message was sent by Atlassian JIRA
(v6.2#6252)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to