raw device usage tracking

Hello, To avoid dealing with offsets, I create separate lvols for every dbspace and one of my instances is reaching > 1500 chunks/lvols (22TB). Thinking it would simplify disk usage management, I built a couple of large lvols of 1.2TB ea. and allocated them to the storage pool. But over time, a few chunks were dropped using onspaces -d instead of the function 'drop chunk to storagepool'. Does anyone have recommendations or methods to inventory the unused parts of a large raw device in order to return them to the storage pool? So far I only found onstat -d / -D to list the allocated/used parts and find the unallocated 'holes' and compare with the storage pool available chunks. Benoît

Hi Benoît, I kind of think you've figured out the answer yourself already :-) The easiest way to do this is to be disciplined in the use of the storagepool and avoid dropping chunks with onspaces. If you are using storagepools you could rename the onspaces binary to onspaces-dont-use or similar. Given you've got gaps in your raw devices, you'll need to work out where they are using 'onstat -d'. I have written a script which takes the output from 'onstat -d' as input and works through the chunks, taking into account their individual page sizes and the system page size and spits out where the free space is. It also makes calls to the logical volume manager to get the volume sizes. It makes the assumption that if a volume is used by Informix it is exclusive to Informix. Unfortunately I can't share this script but hopefully this is still a useful reply. Ben.