Hi Andrej,
The invalid path patch will be in slurm version 2.2.5.
I believe the attached patch will work better than your original patch for
power save mode.
Your patch would result in every powered down node being rebooted whenever the
slurmctld daemon restarts. This patch will leave the nodes powered down until
needed.
This change should also be in slurm version 2.2.5.
Your other patches will be reviewed in the coming days.
Thank you for your contributions!
________________________________________
From: [email protected] [[email protected]] On Behalf
Of Andrej N. Gritsenko [[email protected]]
Sent: Thursday, April 07, 2011 3:18 AM
To: [email protected]
Subject: [slurm-dev] power save mode fix
Hello there!
There is a problem with power save, it's experienced in SLURM 2.2.3.
When node went into NODE_STATE_POWER_SAVE and slurmctld restarted then
node will stay in NODE_STATE_UNKNOWN | NODE_STATE_POWER_SAVE forever in
slurmctld after it restarted. Solution is simple - node in that state
should be awakened so it change state from NODE_STATE_UNKNOWN to state
NODE_STATE_IDLE (if there is no error of course). Patch attached.
Andrej.
Index: src/slurmctld/node_mgr.c
===================================================================
--- src/slurmctld/node_mgr.c (revision 23037)
+++ src/slurmctld/node_mgr.c (working copy)
@@ -413,7 +413,16 @@
node_ptr->node_state |=
NODE_STATE_FAIL;
if (node_state & NODE_STATE_POWER_SAVE) {
- if (power_save_mode) {
+ if (power_save_mode &&
+ IS_NODE_UNKNOWN(node_ptr)) {
+ orig_flags = node_ptr->
+ node_state &
+ NODE_STATE_FLAGS;
+ node_ptr->node_state =
+ NODE_STATE_IDLE |
+ orig_flags |
+ NODE_STATE_POWER_SAVE;
+ } else if (power_save_mode) {
node_ptr->node_state |=
NODE_STATE_POWER_SAVE;
} else if (hs)
