Hi,
  We're running online JFS on HP.  We had a recommendation given to us to replace the settings in fstab for the online redo logs and archive logs.  The settings were:
 
/dev/vgPRD03/lvPRDsaparch     /oracle/PRD/saparch    vxfs  delaylog   0       
/dev/vgPRD02/lvPRDoriglogB    /oracle/PRD/origlogB   vxfs  delaylog   0       
/dev/vgPRD01/lvPRDoriglogA    /oracle/PRD/origlogA   vxfs  delaylog   0       
/dev/vgPRD01/lvPRDmirrlogB    /oracle/PRD/mirrlogB   vxfs  delaylog   0       
/dev/vgPRD02/lvPRDmirrlogA    /oracle/PRD/mirrlogA   vxfs  delaylog   0
 
The recommended change would be:
 
/dev/vgPRD03/lvPRDsaparch     /oracle/PRD/saparch    vxfs  nodatainlog,convosync=direct   0       
/dev/vgPRD02/lvPRDoriglogB    /oracle/PRD/origlogB   vxfs  nodatainlog,convosync=direct   0       
/dev/vgPRD01/lvPRDoriglogA    /oracle/PRD/origlogA   vxfs  nodatainlog,convosync=direct   0       
/dev/vgPRD01/lvPRDmirrlogB    /oracle/PRD/mirrlogB   vxfs  nodatainlog,convosync=direct   0       
/dev/vgPRD02/lvPRDmirrlogA    /oracle/PRD/mirrlogA   vxfs  nodatainlog,convosync=direct   0       
 
What a different branch of the consultants gave us was:
 
/oracle/DV1/origlogA vxfs delaylog,convosync=direct  0 2
/oracle/DV1/origlogB vxfs delaylog,convosync=direct  0 2
/oracle/DV1/mirrlogA vxfs delaylog,convosync=direct  0 2
/oracle/DV1/mirrlogB vxfs delaylog,convosync=direct  0 2
 
They based this on a quick test using the following command:
 
/bin/time -p prealloc myfile 1000000000
 
which indicated a slight negative impact to I/O with the nodatainlog.  I've read the man pages on prealloc, but it isn't clear that this is a valid simulation of the online redo logs.  Does anyone have experience with this?  Was this a valid test?
 
Best regards,
Russ Brooks
 

Reply via email to