System Slowwwww

[telemorgan]

Hey guys

Our V880 is running very slow. I know the basics but my knowledge ends there. Does anything from the output of this top command stand out to anyone out there.

# top
load averages: 1.05, 1.57, 1.62 enterprise 11:08:14
574 processes: 570 sleeping, 1 stopped, 3 on cpu
CPU states: 43.6% idle, 6.5% user, 3.1% kernel, 46.8% iowait, 0.0% swap
Memory: 8.0G real, 2.4G free, 4.4G swap in use, 11.6G swap free

PID USERNAME THR PR NCE SIZE RES STATE TIME FLTS CPU COMMAND
11401 oracle 258 60 0 978M 961M cpu02 11:55 664 3.10% oracle
14672 oracle 11 57 0 946M 928M sleep 0:03 0 2.02% oracle
11176 oracle 1 0 0 933M 917M sleep 1:18 0 1.83% oracle
12905 oracle 1 59 0 934M 916M cpu01 2:05 0 1.17% oracle
15653 oracle 11 43 0 945M 927M sleep 0:26 0 1.04% oracle
855 oracle 258 58 0 940M 917M sleep 45:30 0 0.69% oracle
8543 oracle 1 41 0 942M 924M sleep 0:50 0 0.45% oracle
15715 oracle 1 59 0 940M 923M sleep 0:01 0 0.43% oracle
15676 oracle 1 52 0 933M 916M sleep 0:03 0 0.40% oracle
858 oracle 11 59 0 934M 912M sleep 102:57 42 0.33% oracle
29440 oracle 1 33 0 933M 917M sleep 65:22 0 0.27% oracle
14007 oracle 1 44 0 933M 915M sleep 0:10 0 0.23% oracle
15717 oracle 1 50 0 933M 915M sleep 0:01 0 0.19% oracle
15757 oracle 1 48 0 2640K 1624K cpu03 0:00 0 0.16% top
15643 oracle 1 58 0 933M 916M sleep 0:04 0 0.13% oracle

 

[daFranze]

look at your messages, the iowait is very high, maybe you have faulty disks or other io problems!

Best Regards, Franz
--
Solaris System Manager from Munich, Germany
I used to work for Sun Microsystems Support (EMEA) for 5 years

 

[achech]

If your disks are fine, it looks like doing lots of memory swap. I think you need more memory, or check your Oracle's configurations.

Cheers,


Achech

 

[marrow]

Run
vmstat 2 10 # Look for entries in rbw (ignore line 1) to see if short of memory or cpu. If rbw =0 all OK
or..
uptime

 

[coffeysm]

Did your DBAS do the semaphore tweaks and stuff that are required by Oracle.

 

[telemorgan]

Thanks for your help guys, We loaded a package called Statspack to monitor the database. This thing works great. We were able to find out that there were three custom (in-house) written programs that went haywire. 20M gets each.

Thanks for your help with this but once we sorted it out and took appropriate action things went back to normal.