OpenSolaris

Printable Version Enter a New Search
Bug ID 6588453
Synopsis Solaris 10 iostat/vmstat performance degradation, due to a large number of devices seen
State 10-Fix Delivered (Fix available in build)
Category:Subcategory utility:performance
Keywords performance | rtiq_regression | slow | snapshot | spbc_idr
Responsible Engineer William Roche
Reported Against
Duplicate Of
Introduced In solaris_10
Commit to Fix s10u6_01
Fixed In s10u6_01
Release Fixed solaris_10u6(s10u6_01) , solaris_nevada(snv_83) (Bug ID:2151461)
Related Bugs 6402814 , 6461051 , 6644731 , 6645066
Submit Date 2-August-2007
Last Update Date 23-May-2008
Description 
A very large configuration -- using a lot of disk devices (several thousands devices seen on the machine) can see the iostat and vmstat commands to take a lot of time on Solaris 10.

For example, a configuration with 7700 disks seen on the machine can take more than 16 hours to run the following command on a T2000 running Solaris 10 (before Update 4) - eating 100% of a CPU:
		iostat -xnCz -Td 60 61

With the new algorithms provided by S10 Update 4 (fixing for example 6402814 -- that should be included in the kernel update 120011-03/120012-03), the same command is running for a little bit more than 2 hours also eating 100% of a CPU (90% user time, 5% system).

--> This is still considered too long !
Work Around 
N/A
Comments 
N/A