From: Faiyaz Ahmed <faiyaza@cs.princeton.edu>
Date: Fri, 4 Apr 2008 19:44:34 +0000 (+0000)
Subject: File locking isn't exclusive in the same process across threads.  Switched to regular... 
X-Git-Tag: NodeManager-1.8-0~59
X-Git-Url: http://git.onelab.eu/?p=nodemanager.git;a=commitdiff_plain;h=be69c6fca3b19f87c483091cdfaa99d465d0c1c1

File locking isn't exclusive in the same process across threads.  Switched to regular semaphores.  The assumption is sliver_vs.configure is called before Vserver.start(), which sets disk_usage_initialized.  The disk check should be skipped during start, so the lock won't carry over.  ...I hope.
---

diff --git a/sliver_vs.py b/sliver_vs.py
index 13ff990..3060b9f 100644
--- a/sliver_vs.py
+++ b/sliver_vs.py
@@ -24,6 +24,9 @@ import vserver
 import accounts
 import logger
 import tools
+from threading import BoundedSemaphore
+
+globalsem = BoundedSemaphore()
 
 # special constant that tells vserver to keep its existing settings
 KEEP_LIMIT = vserver.VC_LIM_KEEP
@@ -42,7 +45,7 @@ class Sliver_VS(accounts.Account, vserver.VServer):
 
     SHELL = '/bin/vsh'
     TYPE = 'sliver.VServer'
-    _init_disk_info_sem = tools.NMLock("/var/run/nm-disk-info.lock")
+    _init_disk_info_sem = globalsem
 
     def __init__(self, rec):
         logger.verbose ('initing Sliver_VS with name=%s'%rec['name'])