[HPDD-discuss] Lustre 2.5.1. HSM

Greetings All, I have a few basic questions regarding the HSM framework in general and regarding the "copytool" in particular: 1) If multiple "agents" are registered with the "coordinator", are the HSM requests sent to all "agents" serially or in parallel? What is the overall state of a file if some agents report success and other agents report failure? 2) Using https://build.hpdd.intel.com/job/lustre-manual/lastSuccessfulBuild/artifa... as a reference, section 22.3 states "Only one copytool could be run by agent node." If I'm backing up to Tape and NFS, I cannot run a "tape copytool" and a "posix copytool" on the same agent node? Must I have 2 separate agent nodes, one for "tape copytool", and another for "posix copytool"? 3) Using https://build.hpdd.intel.com/job/lustre-manual/lastSuccessfulBuild/artifa... as a reference, section 22.3.1 states "A Lustre file system supports an unlimited number of copytool instances". Does it mean that I must have unlimited agent nodes, each of which is running a single instance of "copytool"? 4) When a copytool crashes, is there any mechanism within the HSM framework to restart a copytool? 5) Is there any load-balancing capability with the HSM framework? 6) Is there HA capability with-in the HSM framework? 7) Is there a way to cluster copytool instances to get better throughput? 8) Is there a later version of the HSM manual than https://build.hpdd.intel.com/job/lustre-manual/lastSuccessfulBuild/artifa... ? If so, where would I find it? Thank you for your prompt reply. Regards, -aamir