Re: [Aoetools-discuss] AOE v83 - unknown device

2014-03-31 Thread James R. Leu
Thank you for your response. Have I over extended the dynamic minor number allocation system by having AOE devices on one server many shelves apart? IE 0.1 and 19.10 Since I track my AOE assignments carefully, would I be better served by turning off dynamic minor number allocation and maintain

Re: [Aoetools-discuss] AOE v83 - unknown device

2014-03-31 Thread Ed Cashin
It occurs to me that old (circa 2005) kernels (and even old user-space tools) cannot handle large system minor numbers. So please consider that. No, using an AoE (major, minor) address space that is sparsely populated is OK. Large gaps in the address space don't mean that you run into a limit

Re: [Aoetools-discuss] AOE v83 - unknown device

2014-03-31 Thread Ed Cashin
On Mar 31, 2014, at 7:52 PM, James R. Leu j...@inoc.com wrote: We're on CentOS 6 kernels - 2.6.32 Which user space tools are you referring too? This was before CentOS 6, so I am having trouble remembering. I had to send a patch to LVM2, and there was even a problem with something basic like

Re: [Aoetools-discuss] AOE v83 - unknown device

2014-03-28 Thread Ed Cashin
Just some quick thoughts in a hurry... Talking about minor numbers is naturally confusing with the aoe driver, because there are at least two kinds: * The AoE address is comprised of a {major, minor} pair, often called shelf.slot. AoE minor number * The Linux system has special block