1. customer wants to have the same behaviour on Linux as on NetWare when NCS fences a node
--> this should be "halt the server" instead of "reboot"
2. customer wants to make sure that when NCS fences a node - no core is written (even if KDUMP is configured)
--> otherwise NCS will fill up the HDD with cores
3. customer wants to "power off" the server instead of "halt" or "reboot" in case of fencing.
--> today NCS does a "echo c >/proc/sysrq-trigger"
but customers what to have "echo o > /proc/sysrq-trigger"
4. all the different options must be configurable via iManager in eDirectory.
We already have the functionality of "node isolation scripts" in eDirectory (on the node object) - however - changing it is not documented, tested or supported.
We tried it - and the "echo o >/proc/sysrq-trigger" creates a strange behavior
(node1 kills node2 at the same point in time when it kills itself)
by: Bhanuprakash B. | over a year ago | Cluster Services (NCS)
Comments