Configuration file resource allocation
Posted: Wed Jun 13, 2007 10:24 pm
I am assigning multiple resource disks in a configuration along the lines of:
According to the documentation and prior experience, these will share scratch activity - at least when they are sized identically.
In my case, I have a small Disk1 {which is local and fast} and a larger /Disk2 {which is on a very busy SAN}. I would like to use the local scratch area for all scratch activity, but it isn't large enough. Ideally I would like to use /Disk1 until it fills up, then continue and use /Disk2 through /Disk{n}. I would have been happy enough to have at least some of the scratch (and perhaps sort and buffer) activity going to the local disk.
Unfortunately, it seems that PX chooses the largest disks (or perhaps those with the most space available) first.
In the 2 disk configuration above I've seen that /Disk1 doesn't get used at all in most cases. It will get used once the /Disk2 fills up with data {I"m not sure but think the trigger used is available space left}.
I think that, particularly with an overloaded SAN or a SAN which uses RAID for all mounted drives, it would be greatly beneficial to use any available local space for scratch, buffer, sort and perhaps even dataset data and then "failing" over to the next resource when full.
Does anyone know more about this mechansim - particularly if the order of precedence for resources can be set either in the configuration file itself or with environment settings elsewhere?
(I haven't gone to support, as I thought that people here not only might know the answer already but the solution might help others.)
Code: Select all
resource scratchdisk "/Disk1/Scratch" {pools ""}
resource scratchdisk "/Disk2/Scratch" {pools ""}According to the documentation and prior experience, these will share scratch activity - at least when they are sized identically.
In my case, I have a small Disk1 {which is local and fast} and a larger /Disk2 {which is on a very busy SAN}. I would like to use the local scratch area for all scratch activity, but it isn't large enough. Ideally I would like to use /Disk1 until it fills up, then continue and use /Disk2 through /Disk{n}. I would have been happy enough to have at least some of the scratch (and perhaps sort and buffer) activity going to the local disk.
Unfortunately, it seems that PX chooses the largest disks (or perhaps those with the most space available) first.
In the 2 disk configuration above I've seen that /Disk1 doesn't get used at all in most cases. It will get used once the /Disk2 fills up with data {I"m not sure but think the trigger used is available space left}.
I think that, particularly with an overloaded SAN or a SAN which uses RAID for all mounted drives, it would be greatly beneficial to use any available local space for scratch, buffer, sort and perhaps even dataset data and then "failing" over to the next resource when full.
Does anyone know more about this mechansim - particularly if the order of precedence for resources can be set either in the configuration file itself or with environment settings elsewhere?
(I haven't gone to support, as I thought that people here not only might know the answer already but the solution might help others.)