|
From: <ha...@us...> - 2008-04-24 21:01:49
|
Revision: 1982
http://cogkit.svn.sourceforge.net/cogkit/?rev=1982&view=rev
Author: hategan
Date: 2008-04-24 14:01:35 -0700 (Thu, 24 Apr 2008)
Log Message:
-----------
only fail after 3 consecutive qstat failures
Modified Paths:
--------------
trunk/current/src/cog/modules/provider-localscheduler/src/org/globus/cog/abstraction/impl/scheduler/pbs/QueuePoller.java
Modified: trunk/current/src/cog/modules/provider-localscheduler/src/org/globus/cog/abstraction/impl/scheduler/pbs/QueuePoller.java
===================================================================
--- trunk/current/src/cog/modules/provider-localscheduler/src/org/globus/cog/abstraction/impl/scheduler/pbs/QueuePoller.java 2008-04-24 19:49:13 UTC (rev 1981)
+++ trunk/current/src/cog/modules/provider-localscheduler/src/org/globus/cog/abstraction/impl/scheduler/pbs/QueuePoller.java 2008-04-24 21:01:35 UTC (rev 1982)
@@ -25,12 +25,15 @@
public class QueuePoller extends Thread {
public static final Logger logger = Logger.getLogger(QueuePoller.class);
+
+ public static final int MAX_CONSECUTIVE_FAILURES = 3;
private LinkedList newjobs, donejobs;
private Set processed;
private Map jobs;
boolean any = false;
private int sleepTime;
+ private int failures;
public QueuePoller() {
setName("PBS-Local provider stream poller");
@@ -140,8 +143,14 @@
processStderr(pqstat.getErrorStream());
int ec = pqstat.waitFor();
if (ec != 0) {
- failAll("QStat failed (exit code " + ec + ")");
+ failures++;
+ if (failures >= MAX_CONSECUTIVE_FAILURES) {
+ failAll("QStat failed (exit code " + ec + ")");
+ }
}
+ else {
+ failures = 0;
+ }
if (logger.isDebugEnabled()) {
logger.debug("QStat done");
}
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|