1847069 Members
5083 Online
110261 Solutions
New Discussion

swapinfo -ta

 
Youlette Etienne_2
Regular Advisor

swapinfo -ta

Hello:

I am on a k580, 3 gb RAM, 11.0 system with 2.9 gb of swap space configured. I am currently considering increasing the swap space. When I run swapinfo -ta, I only get values for the total amount of kb's and PCT used, but I do not get these values for the individual device swap areas (below). Has anyone encountered this before?

Tue Aug 29 14:00:02 EDT 2000
Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 1048576 0 1048576 0% 0 - 1 /dev/vg00/lvol2
dev 409600 0 409600 0% 0 - 1 /dev/vg00/lvol10
dev 512000 0 512000 0% 0 - 1 /dev/vg02/lvol4
dev 512000 0 512000 0% 0 - 1 /dev/vg03/lvol10
dev 512000 0 512000 0% 0 - 1 /dev/vg04/lvol6
reserve - 2189696 -2189696
total 2994176 2189696 804480 73% - 0 -
If at first you don't succeed, change the rules!
13 REPLIES 13
Rick Garland
Honored Contributor

Re: swapinfo -ta

Try swapinfo -atm
James R. Ferguson
Acclaimed Contributor

Re: swapinfo -ta

Hi:

This looks normal to me for a system that has not used any of its device swap; rather is using memory only at this point.

...JRF...
Anthony deRito
Respected Contributor

Re: swapinfo -ta

There appears to be no swap used for any of your swap devices. This is evident by the Kb Used column for each swap device (0%). As you can see, the Kb FREE values are the same as the Kb AVAIL values.

Also, the PCT USED column for eache device shows how much swap is being used for pageout activities. This is not an indication of how much swap is currently being used by running processes.

The total line is also confusing you. The Kb USED column for the "Total" line indicates that there is swap "reserved" for running processes. This does not mean that the swap is being used at this time - it is only reserved "in case" the kernel needs it to perform pageout activities.

Of your total swap pool, you have 2189696 KB allocated in the reserve pool. If you subtract that number from the total amount of device swap you have enabled (2994176 KB), you have get the 804480 Kb value that you have remaining.

Tony
CHRIS_ANORUO
Honored Contributor

Re: swapinfo -ta

Hi Youlette,

If you enable pseudo swapping by setting kernel parameter SWAPMEM_ON=1, you will see the memory swap percentage and know where the large % value is coming from.
I will also advice you to set nbuf=0 and bufpages=0 for dynamic cache reads.
Kernel parameters to set are:
swapmem_on=1
bufpages=0
nbuf=0
dbc_pct_min=5
dbc_pct_max=25
The last value id 25% of 3gb
When pseudo swapping is enabled, you get 75% of 3gb added to total swap space.
Find time and retune your kernel with these values for better swapping.

Chris

When We Seek To Discover The Best In Others, We Somehow Bring Out The Best In Ourselves.
Anthony deRito
Respected Contributor

Re: swapinfo -ta

As metioned before, the 73% number is the percentage of total available swap which has been allocated for the "reserve" pool. This is not an indication of how much virtual memory is currently being used.

This is done automatically by the Memory Management Subsystem. You have no control over this. The kernel wants to have the virtual memory available in case it needs to pageout the entire process(es). So what it does is "reserve" the space ahead of time.

Tony
Youlette Etienne_2
Regular Advisor

Re: swapinfo -ta

Thanks for your quick response. We have Oracle 8.0.5 installed. Some processes produced the following error:

ORA-04030: out of process memory when trying to allocate 10264 bytes (session he

The dba first suggested to increase maxdsiz,maxsiz and maxtsiz. I increased maxdsiz from 1.6 to 2.0 gb. However, the error reappeared a few days later. I also noticed that swap space usage increased from a high of 80% to 92%, which indicated that swap space may have to be increased. I have a cron job that runs swapinfo -ta, vmstat, and a few sar reports (until I can get sos installed-we do not have Glance) to run every 20 minutes. Yesterday, the swap space climbed to 100% use, but there was no swapping occurring. Shortly after, Oracle errored out with a memory core dump error and we were not able to open any telnet sessions, but I was able to login at the console. I ultimately had to reboot the system since restarting inetd still did not allow telnet access. This was the second occurrence with the following oracle error:

memory core dump errors.

and the following message from syslog:
rpcbind: terminate: rpcbind terminating on signal. Restart with "rpcbind -w"

Today, we will be changing maxdsiz back to 1.6gb and try increasing maxsiz from 8 mb to 80 mb. Have you encountered this scenario?

swapinfo -atm
# swapinfo -atm
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 1024 0 1024 0% 0 - 1 /dev/vg00/lvol2
dev 400 0 400 0% 0 - 1 /dev/vg00/lvol10
dev 500 0 500 0% 0 - 1 /dev/vg02/lvol4
dev 500 0 500 0% 0 - 1 /dev/vg03/lvol10
dev 500 0 500 0% 0 - 1 /dev/vg04/lvol6
reserve - 1973 -1973
total 2924 1973 951 67% - 0 -
If at first you don't succeed, change the rules!
Cheryl Griffin
Honored Contributor

Re: swapinfo -ta

Check for patch PHKL_16639 or it's supersede PHKL_18543 (s700_s800 11.0)
# swlist -l fileset -a state |grep 16639
# swlist -l fileset -a state |grep 18543
"Downtime is a Crime."
Anthony deRito
Respected Contributor

Re: swapinfo -ta

It is important to keep in mind the difference between "reserved" swap and "used" swap. The high percentages of memory you have seen were a result of the Memory Management Subsystem not being able to "reserve" enough swap when your processes are created.

In order to have the space available, create more device swap using the swapon command. You can do this while the system is up.

The high percentages you were seeing were not a reflection of page-out activities and memory pressure.


Tony
Youlette Etienne_2
Regular Advisor

Re: swapinfo -ta

Tony,

Thanks for clarifying this. My next step is to increase the swap space and I will let you all know what the outcome is.

Thanks to all once again for your help

If at first you don't succeed, change the rules!
John Palmer
Honored Contributor

Re: swapinfo -ta

Youlette,

As mentioned earlier by Chris, set the kernel parameter swapmem_on to 1. This will reduce your actual device swap requirement by 75% of your physical memory (2.25Gb) so you won't have to waste any more disk space for swap files that you are not using. You might even be able to reduce your device swap after monitoring swap usage for a while

This parameter modifies the swap algorithm used by the system and will have no adverse effects.

Regards,
John
James R. Ferguson
Acclaimed Contributor

Re: swapinfo -ta

Hi:

This seems to be a carry-on to yesterday's post:

http://my1.itrc.hp.com/cm/QuestionAnswer/1,1150,0x6f96c3d7fb78d4118fef0090279cd0f9,00.html

If so, have you inflated maxdsiz ???

...JRF...
Tom Danzig
Honored Contributor

Re: swapinfo -ta

BTW, if you have no more devices you can use for swap, you can either enable pseudo swap as posted earlier or create file system swap instead of device swap.
Youlette Etienne_2
Regular Advisor

Re: swapinfo -ta

I have read additional information on pseudo swapping. It is a more feasible and less strenuous option. Thanks.

If at first you don't succeed, change the rules!