05-02-2013 08:12 AM
We have 8GB of memory
Since running oracle 11g on this server we are getting excessive swapping and had to add more device swap
the database is not responding well and even the command prompt is hanging on the OS.
Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 8388608 1959628 6428980 23% 0 - 1 /dev/vg00/lvol2
dev 16777216 1928812 14848404 11% 0 - 1 /dev/vgsan/swap3
reserve - 6248632 -6248632
memory 7948276 1894712 6053564 24%
# vmstat 1 1
procs memory page
r b w avm free re at pi po fr de sr in
sy cs us sy id
2 4 0 1613999 73551 11 0 464 588 29 0 5531 1536
73155 2223 13 5 82
05-02-2013 08:15 AM
How are you determining that only 44% of RAM is being used?
8 GB of RAM seems awfully small for 11iv3 and Oracle 11g. I would do a minimum of 16GB and would go to 32 GB if I could afford it. This would enable larger SGA sizes as well, which could improve DB performance.
05-02-2013 08:19 AM
we only have 8GB of memory but this database ran on 4GB until recently
The Oracle minimum is 1GB and the SGA is set to 3GB
I have uploaded some more data I had to collect for the call I have open with support
05-02-2013 12:27 PM - edited 05-02-2013 12:29 PM
>memory only 44% used
>we are getting excessive swapping and had to add more device swap
It seems hard to believe that 44% and the vmstat PO output. Did you measure these at the same time?
From your attachment, it seems you did. But you seem to have a real variable load.
As Patrick said, you need more RAM. You seem to have allocated 6 GB of device swap.
05-04-2013 09:19 PM
The system is using 95% of physical memory. vmstat 78000 free pages of 4k per page , equals to around 400Mbyte of free memory. On a 8Gbyte RAM system, that means 5% is left over. Which basically means the system is close to trashing.
The memory line in the swap info output has nothing to do with how much memory the system has left. Its a swaprelated information.
05-08-2013 06:55 PM
>> we only have 8GB of memory but this database ran on 4GB until recently
What the database did in the past is only relevant if the DBAs made *no* changes to schemas or SGA or SQL procedures. The fact that the system is running out of RAM (and therefore paging very badly) means that something has dramatically cheanged. It may not be Oracle, it could be JavaJunk that has run away. But the system is struggling to accomodate the application memory demands.
I would start by identifying memory usage by programs. Paste this command exactly into a shell prompt:
UNIX95=1 ps -e -o vsz,pid,ppid,args | sort -rn | head -20
The first column is KBytes for the program's address space, so the biggest programs wil sort to the top of the list. What you are looking for are unusual values for some programs.
Take a look at shared memory with:
Again, unusually large values that are not expected may point to a runaway program.
When the pagout rate hits 3 digits (more than 99), the OS will be crippled as it tries to rearrange memory and processes. A high paging rate will destroy performance. The only fix without changing the programs is to double, or triple the amount of RAM.
06-04-2013 04:34 AM
For 11.2 database, you only need 1.5 times RAM of swap (see Metalink Note ID 169706.1): "Between 4 GB and 8 GB then 2 times RAM. Between 8 GB and 32 GB then 1.5 times RAM. More than 32 GB then 32 GB RAM". So it seems you may have to much swap. I'm not sure about that, but I think it would be faster on your internal disks (/dev/vg00) than on some SAN...
I also think 8 Gb of RAM is a little short:
1Gb for HP-UX
3 Gb for Oracle DB
2-3 Gb? for pga_aggregate_target (Process Global Area)
Check your database parameters with the following query (as system or sysdba):
where (name like '%sga%' or name like '%pga%')
and isdeprecated <> 'TRUE'