Index: common.scm
==================================================================
--- common.scm
+++ common.scm
@@ -1969,30 +1969,36 @@
     (or (common:get-cached-info actual-host "num-cpus" age: (+ 2592000 (random 3600))) ;; hosts had better not be changing the number of cpus too often!
 	(let* ((proc   (lambda ()
 			 (let loop ((numcpu 0)
 				    (inl    (read-line)))
 			   (if (eof-object? inl)
-			       (begin
-				 (common:write-cached-info actual-host "num-cpus" numcpu)
-				 numcpu)
+			       (if (> numcpu 0)
+				   numcpu
+				   #f) ;; if zero return #f so caller knows that things are not working
 			       (loop (if (string-match "^processor\\s+:\\s+\\d+$" inl)
 					 (+ numcpu 1)
 					 numcpu)
 				     (read-line))))))
 	       (result (if remote-host
 			   (with-input-from-pipe 
 			       (conc "ssh " remote-host " cat /proc/cpuinfo")
 			     proc)
 			   (with-input-from-file "/proc/cpuinfo" proc))))
-	  (common:write-cached-info actual-host "num-cpus" result)
+	  (if (> result 0)(common:write-cached-info actual-host "num-cpus" result))
 	  result))))
 
 ;; wait for normalized cpu load to drop below maxload
 ;;
-(define (common:wait-for-normalized-load maxload msg remote-host)
+(define (common:wait-for-normalized-load maxload msg remote-host #!optional (rem-tries 5))
   (let ((num-cpus (common:get-num-cpus remote-host)))
-    (common:wait-for-cpuload maxload num-cpus 15 msg: msg remote-host: remote-host)))
+    (if num-cpus
+	(common:wait-for-cpuload maxload num-cpus 15 msg: msg remote-host: remote-host)
+	(begin
+	  (thread-sleep! (random 60)) ;; we failed to get num cpus. wait a bit and try again
+	  (if (> rem-tries 0)
+	      (common:wait-for-normalized-load maxload msg remote-host (- rem-tries 1))
+	      #f)))))
 
 ;; DO NOT CALL THIS DIRECTLY. It is called from common:wait-for-normalized-load
 ;;
 (define (common:wait-for-cpuload maxload-in numcpus-in waitdelay #!key (count 1000) (msg #f)(remote-host #f)(force-maxload #f))
   (let* ((loadavg (common:get-cpu-load remote-host))