| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| Swap Space and Dataflow Engines |
« View previous topic :: View next topic » |
| Author |
Message
|
| moepi |
 Posted: Fri Aug 08, 2003 2:05 am Post subject: Swap Space and Dataflow Engines |
 |
|
Novice
Joined: 11 Jul 2001
Posts: 21
Location: South Africa
|
Hi All,
We are running WMQI 2.1, DB2 7.2 CSD05, MQSeries 5.2 on Solaris 2.8.
We have a problem where the Dataflows engines do not really
release memory.
We get an error that says the system is running out of swap space. I've
been monitoring the system and this has lead me to belief its WQMI that's
causing the problem.
How can we make sure that WMQI(Dataflow engines) releases memory once its
done processing messages ?
Please find below the stats regarding memory usage.
Thanks,
Moepi
load averages: 0.02, 0.05, 0.05 10:15:24
196 processes: 195 sleeping, 1 on cpu
CPU states: 99.2% idle, 0.0% user, 0.8% kernel, 0.0% iowait, 0.0% swap
Memory: 1024M real, 17M free, 617M swap in use, 509M swap free
PID USERNAME THR PRI NICE SIZE RES STATE TIME CPU COMMAND
985 mqadmin 18 58 0 570M 336M sleep 17:29 0.00% DataFlow
951 mqadmin 15 59 0 452M 213M sleep 11:26 0.00% DataFlow
896 mqadmin 12 59 0 167M 47M sleep 0:11 0.00% DataFlow 957 mqadmin 18 59 0 156M 51M sleep 0:15 0.00% DataFlow 954 mqadmin 15 59 0 134M 38M sleep 0:09 0.00% DataFlow 960 mqadmin 12 59 0 131M 37M sleep 0:08 0.00% DataFlow 956 mqadmin 13 59 0 130M 37M sleep 0:07 0.00% DataFlow 976 mqadmin 11 59 0 130M 37M sleep 0:07 0.00% DataFlow 955 mqadmin 9 59 0 129M 36M sleep 0:07 0.00% DataFlow 1015 mqadmin 4 58 0 88M 54M sleep 0:00 0.00% db2sysc
1100 mqadmin 4 58 0 86M 49M sleep 0:01 0.00% db2sysc
878 mqadmin 4 58 0 86M 51M sleep 0:00 0.00% db2sysc
1140 mqadmin 4 34 0 85M 51M sleep 0:00 0.00% db2sysc
1146 mqadmin 4 54 0 85M 48M sleep 0:00 0.00% db2sysc
1018 mqadmin 4 58 0 85M 52M sleep 0:00 0.00% db2sysc
986 mqadmin 4 58 0 85M 51M sleep 0:00 0.00% db2sysc
952 mqadmin 4 58 0 85M 51M sleep 0:00 0.00% db2sysc
965 mqadmin 4 58 0 85M 51M sleep 0:00 0.00% db2sysc
971 mqadmin 4 58 0 85M 51M sleep 0:00 0.00% db2sysc
894 mqadmin 4 58 0 85M 51M sleep 0:00 0.00% db2sysc
962 mqadmin 4 46 0 85M 51M sleep 0:00 0.00% db2sysc
1934 mqadmin 4 54 0 85M 52M sleep 0:00 0.00% db2sysc
1928 mqadmin 4 54 0 85M 52M sleep 0:00 0.00% db2sysc
877 mqadmin 4 58 0 85M 51M sleep 0:00 0.00% db2sysc
1128 mqadmin 4 28 0 85M 51M sleep 0:00 0.00% db2sysc
1268 mqadmin 4 58 0 85M 51M sleep 0:03 0.00% db2sysc
1932 mqadmin 4 58 0 85M 51M sleep 0:00 0.00% db2sysc
1960 mqadmin 4 58 0 85M 51M sleep 0:00 0.00% db2sysc
2852 mqadmin 4 58 0 85M 51M sleep 0:00 0.00% db2sysc
1473 mqadmin 4 58 0 85M 51M sleep 0:00 0.00% db2sysc
swap -l
swapfile dev swaplo blocks free
/dev/dsk/c2t1d0s1 32,609 16 2100704 1041664
/swapfile - 16 204784 0 |
|
| Back to top |
|
 |
| DanielG |
| Posted: Fri Aug 08, 2003 4:32 am Post subject: |
|
|
Novice
Joined: 29 Mar 2003
Posts: 13
|
Moepi,
which CSD level is your WMQI running on? There have been several memory related problems that have been fixed in CSD04 / CSD05 (NNSY components).
If you're up to date, pls. describe your situation in a bit more detail: is memory consumption steadily growing over time or is it just rising when processing a single message, what sizes are your messages of, format/parser/message domain, complexity of message flow... |
|
| Back to top |
|
|
| Craig B |
| Posted: Wed Aug 13, 2003 6:54 am Post subject: |
|
|
Partisan
Joined: 18 Jun 2003
Posts: 316
Location: UK
|
Once the operating system allocates memory to your DataFlowEngine process and it has been committed to that process, then the operating system will never reclaim the storage from the process until the process is terminated. What this means is that you will never see the memory usage statistics of your DataFlowEngine process reduce while it is running.
The message flows within your DataFlowEngine will free storage at the end of the processing of your message flow, but this will be freed up on the memory heap, and not back to the operating system.
At some point during your processing, your DataFlowEngine processes storage usage should plateau at the point where its internal memory heap has enough storage to satisfy its message flows without requesting more from the operating system.
Some operating system heap managers may be able to reclaim unused storage from a process during its lifetime, but I believe the Solaris operating system does not do this.
_________________
Regards
Craig |
|
| Back to top |
|
|
|
|
|
| |
|
Page 1 of 1 |
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
|
|
