Please let me know your inputs on the following,
1) Under all path failure condition or the device is removed from array (using management station), IO continues if "queue_if_no_path" feature is enabled. As all the requests are cached all the memory gets used out for this. Can there be any mechanism where the all path failure is detected and marked, devices can be suspended or removed after some time (user configurable). By doing this the OS stalling can be avoided when the LUN is in dead state for prolonged period of time.
2) I am not able to give partial product ID string (substring) in /etc/multipath.conf as DM expects complete product string. This requirement is needed to cater different product id strings for same product. The XP array has the ids based on the emulation type of each LUN. With the current settings available in /etc/multipath.conf, the only option is to have separate “device” structs mentioning about the policies/features for each LUN.