|
|
Mario Cárdenas
|
Hi.
We are using a rx-7640 with 32 GB of memory, HP-UX 11.31 running Oracle 10g and 600 users. On "top" is possible to see the vhand proccess high on the list, usually betwen the firsts 5. Also, top reflects few free memory, less than 300 MB. "sar -d" also reflects the swap disk is having a lot of access. "vmstat" also reflects high "page po" values (betwen 50 and 6.000).
I am almost sure we need more memory for the system, but I would like to have more "proves". Please advice.
Thank you for help!.
|
|
|
Note: If you are the author of this question and wish to assign points to any of the answers, please login first.For more information on assigning points ,click
here
|
|
|
Sort Answers By:
Date or Points
|
|
Patrick Wallek
|
|
Nov 4, 2009 20:15:09 GMT
5 pts
|
|
I'm not sure how much more proof you need. All of what you reference points to memory shortage.
Showing the output of 'swapinfo -tam' could also help. That will show how much has actually been paged out. |
|
|
Mario Cárdenas
|
|
Nov 4, 2009 20:21:06 GMT
N/A: Question Author
|
|
Hi Patrick.
Here it is.
mls01:/home/ust>>swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 32768 0 32768 0% 0 - 1 /dev/vg00/lvol2
dev 32764 2283 30481 7% 0 - 0 /dev/vg01/lv01
dev 32760 1867 30893 6% 0 - 0 /dev/vgn01/lvol1
dev 37600 1854 35746 5% 0 - 0 /dev/vgn03/lvol1
reserve - 23159 -23159
memory 31040 19904 11136 64%
total 166932 49067 117865 29% - 0 -
mls01:/home/ust>>date
Wed Nov 4 15:54:43 SAT 2009
mls01:/home/ust>> |
|
|
Patrick Wallek
|
|
Nov 4, 2009 20:31:50 GMT
5 pts
|
|
The 'swapinfo' output is showing that you are currently using about 6 GB of device swap which is the result of the page outs that you see.
I would add at least 16 GB additional RAM to this server. You will probably see an improvement in performance as well since you should not be paging out to disk as much (or at all). |
|
Tor-Arne Nostdal
|
|
Nov 4, 2009 20:46:08 GMT
4 pts
|
|
The alternative is to have your applications spend less ;)
=> send that request to your DB manager
Could [s]he have been too eager in consuming memory for caching. |
|
|
Sagar Sirdesai
|
|
Nov 5, 2009 04:45:05 GMT
4 pts
|
|
Hi
Check the vmstat command if the po value is more this suggests that server is short of physical memory.
As suggested by other members either increase the memory of reduce the number of processes on this box.
I know if you java process running then they are major candidate for memory consumption
Sagar |
|
Emil Velez
|
|
Nov 5, 2009 04:50:32 GMT
3 pts
|
|
As with redneck jokes
You know you have a memory shortage if you are using swapspace
You know you have a memory shortage if you are doing pageouts to your swap device
You know you have a memory shortage if you have disk activity on your swap device.
You know you have a memory shortage if vhand is a top process |
|
Raj D.
|
|
Nov 5, 2009 04:55:56 GMT
5 pts
|
|
Mario,
-How much memory usage do you see in glance during peak hours. Also you can check the measure ware graph with historical data to find out if you really need more memory.
- Use kmeminfo tool to see separate memory utilization between system and application .
- As per 600 oracle user , if one user consume ~5MB then Total memory required wold be 30GB , you got 32GB that is not enough in this case as there should be provision for system and kernel memory as well.
Hth,
Raj. |
|
|
Venkatesh BL
|
|
Nov 5, 2009 04:58:37 GMT
8 pts
|
|
what is the value of 'filecache_min' and 'filecache_max' tunables on the system?
Please note that the 'pageout' rate on 11.31 also refers to the reclaimation of file cache pages. So, it does not always mean that the system is under pressue. It could just be that the file cache is hogging more pages. |
|
|
Mario Cárdenas
|
|
Nov 5, 2009 13:16:24 GMT
N/A: Question Author
|
|
Hi every body, thank you for your help!.
I'll check the JAVA proccess memory usage, and the memory detail with kmeminfo. About the 'filecache_min' and 'filecache_max' , both are in "auto tuning". In fact, I don't like too much "the automatic" parameters, but this one is working that way. I accept suggestions.
Thank you!. |
|
|
Mario Cárdenas
|
|
Nov 5, 2009 15:24:51 GMT
N/A: Question Author
|
|
|
I this attachment could be interesting. |
|
|
Mario Cárdenas
|
|
Nov 9, 2009 15:22:54 GMT
N/A: Question Author
|
|
Hi.
The filecache_ min and filecache_max parameters was by default and I change them to 2% and 15%. This "free" some memory.
The system is "moving" better, but of course this is not the solution... I'll also aplay the "PHKL_40240 11.31 vm cumulative patch" this week.
Hope this help until we buy more memory. |
|
Steven E. Protter
|
|
Nov 9, 2009 20:25:51 GMT
Unassigned
|
|
Shalom,
'filecache_max' is the 11.31 eqpuivalent of dbc_max_pct, the old buffer cache.
Quickest way to get some needed memory back is to reduce this figure.
But the purpose of this memory is to buffer and speed io input/output to and from disk.
So be prepared for some slowdown if that was actually being used. But, the impact overall to a paging system should be positive.
SEP |
|
Printable version
;)