Send Linux-ha-cvs mailing list submissions to
[email protected]
To subscribe or unsubscribe via the World Wide Web, visit
http://lists.community.tummy.com/mailman/listinfo/linux-ha-cvs
or, via email, send a message with subject or body 'help' to
[EMAIL PROTECTED]
You can reach the person managing the list at
[EMAIL PROTECTED]
When replying, please edit your Subject line so it is more specific
than "Re: Contents of Linux-ha-cvs digest..."
Today's Topics:
1. Linux-HA CVS: lrm by alan from ([email protected])
2. Linux-HA CVS: cts by alan from ([email protected])
3. Linux-HA CVS: lrm by alan from ([email protected])
----------------------------------------------------------------------
Message: 1
Date: Thu, 9 Feb 2006 21:06:34 -0700 (MST)
From: [email protected]
Subject: [Linux-ha-cvs] Linux-HA CVS: lrm by alan from
To: [EMAIL PROTECTED]
Message-ID: <[EMAIL PROTECTED]>
linux-ha CVS committal
Author : alan
Host :
Project : linux-ha
Module : lrm
Dir : linux-ha/lrm/lrmd
Modified Files:
lrmd.c
Log Message:
Put in a fix suggested by Sun Jiang Dong for an lrm problem where
we try and read from a closed file descriptor.
===================================================================
RCS file: /home/cvs/linux-ha/linux-ha/lrm/lrmd/lrmd.c,v
retrieving revision 1.211
retrieving revision 1.212
diff -u -3 -r1.211 -r1.212
--- lrmd.c 10 Feb 2006 03:20:01 -0000 1.211
+++ lrmd.c 10 Feb 2006 04:06:34 -0000 1.212
@@ -1,4 +1,4 @@
-/* $Id: lrmd.c,v 1.211 2006/02/10 03:20:01 alan Exp $ */
+/* $Id: lrmd.c,v 1.212 2006/02/10 04:06:34 alan Exp $ */
/*
* Local Resource Manager Daemon
*
@@ -3200,7 +3200,7 @@
}
op_type = ha_msg_value(op->msg, F_LRM_OP);
- if (op->first_line_ra_stdout[0] == EOS) {
+ if (op->first_line_ra_stdout[0] == EOS && op->ra_stdout_fd >= 0) {
handle_pipe_ra_stdout(-1, op);
}
rc = RAExec->map_ra_retvalue(exitcode, op_type,
op->first_line_ra_stdout);
@@ -3376,14 +3376,14 @@
return FALSE;
}
- if (fd == -1) {
+ if (fd < 0) {
fd = op->ra_stdout_fd;
}
if (fd < 0) {
lrmd_log(LOG_CRIT
- , "%s:%d: Attempt to read from closed file descriptor."
- , __FUNCTION__, __LINE__);
+ , "%s:%d: Attempt to read from closed file descriptor %d."
+ , __FUNCTION__, __LINE__, fd);
lrmd_op_dump(op, "op w/closed fd");
return FALSE;
}
@@ -3645,6 +3645,10 @@
}
/*
* $Log: lrmd.c,v $
+ * Revision 1.212 2006/02/10 04:06:34 alan
+ * Put in a fix suggested by Sun Jiang Dong for an lrm problem where
+ * we try and read from a closed file descriptor.
+ *
* Revision 1.211 2006/02/10 03:20:01 alan
* Put in checks for unallocated (freed) operation structures.
* Also print out more information in the cases where I/O fails
------------------------------
Message: 2
Date: Thu, 9 Feb 2006 21:07:42 -0700 (MST)
From: [email protected]
Subject: [Linux-ha-cvs] Linux-HA CVS: cts by alan from
To: [EMAIL PROTECTED]
Message-ID: <[EMAIL PROTECTED]>
linux-ha CVS committal
Author : alan
Host :
Project : linux-ha
Module : cts
Dir : linux-ha/cts
Modified Files:
CTS.py.in CTStests.py.in
Log Message:
Separated out the "wait for all nodes to come up" function into
a more usable class...
===================================================================
RCS file: /home/cvs/linux-ha/linux-ha/cts/CTS.py.in,v
retrieving revision 1.48
retrieving revision 1.49
diff -u -3 -r1.48 -r1.49
--- CTS.py.in 13 Sep 2005 16:32:15 -0000 1.48
+++ CTS.py.in 10 Feb 2006 04:07:41 -0000 1.49
@@ -886,32 +886,17 @@
j=j-1
-class InitClusterManager(ScenarioComponent):
- (
-'''InitClusterManager is the most basic of ScenarioComponents.
-This ScenarioComponent simply starts the cluster manager on all the nodes.
-It is fairly robust as it waits for all nodes to come up before starting
-as they might have been rebooted or crashed for some reason beforehand.
-''')
-
- def IsApplicable(self):
- '''InitClusterManager is so generic it is always Applicable'''
- return 1
-
-
- def _IsNodeBooted(self, node):
+class NodeStatus:
+ def IsNodeBooted(self, node):
'''Return TRUE if the given node is booted (responds to pings'''
return os.system("@PING@ -nq -c1 @PING_TIMEOUT_OPT@ %s >/dev/null
2>&1" % node) == 0
-
- def _WaitForNodeToComeUp(self, node, Timeout=300):
-
- '''Return TRUE when given node comes up, or FALSE if timeout'''
-
+ def WaitForNodeToComeUp(self, node, Timeout=300):
+ '''Return TRUE when given node comes up, or None/FALSE if timeout'''
timeout=Timeout
anytimeouts=0
while timeout > 0:
- if self._IsNodeBooted(node):
+ if self.IsNodeBooted(node):
if anytimeouts:
# Fudge to wait for the system to finish coming up
time.sleep(30)
@@ -926,22 +911,35 @@
timeout = timeout - 1
self.Env.log("%s did not come up within %d tries" % (node, Timeout))
- return None
- def _WaitForAllNodesToComeUp(self, nodes, timeout=300):
+ def WaitForAllNodesToComeUp(self, nodes, timeout=300):
'''Return TRUE when all nodes come up, or FALSE if timeout'''
for node in nodes:
- if not self._WaitForNodeToComeUp(node, timeout):
+ if not self.WaitForNodeToComeUp(node, timeout):
return None
return 1
+class InitClusterManager(ScenarioComponent):
+ (
+'''InitClusterManager is the most basic of ScenarioComponents.
+This ScenarioComponent simply starts the cluster manager on all the nodes.
+It is fairly robust as it waits for all nodes to come up before starting
+as they might have been rebooted or crashed for some reason beforehand.
+''')
+ def __init__(self, Env):
+ self.ns=NodeStatus()
+
+ def IsApplicable(self):
+ '''InitClusterManager is so generic it is always Applicable'''
+ return 1
+
def SetUp(self, CM):
'''Basic Cluster Manager startup. Start everything'''
- if not self._WaitForAllNodesToComeUp(CM.Env["nodes"]):
+ if not self.ns.WaitForAllNodesToComeUp(CM.Env["nodes"]):
return None
CM.prepare()
@@ -963,7 +961,7 @@
def TearDown(self, CM):
'''Set up the given ScenarioComponent'''
- self._WaitForAllNodesToComeUp(CM.Env["nodes"])
+ self.ns.WaitForAllNodesToComeUp(CM.Env["nodes"])
# Stop the cluster manager everywhere
===================================================================
RCS file: /home/cvs/linux-ha/linux-ha/cts/CTStests.py.in,v
retrieving revision 1.135
retrieving revision 1.136
diff -u -3 -r1.135 -r1.136
--- CTStests.py.in 7 Jan 2006 21:22:20 -0000 1.135
+++ CTStests.py.in 10 Feb 2006 04:07:41 -0000 1.136
@@ -2309,9 +2309,10 @@
if self.CM.ShouldBeStatus[node] == self.CM["down"]:
self.CM.StartaCMnoBlock(node)
if watch.lookforall():
- if self.CM.cluster_stable():
- return self.success()
- return self.failure("Cluster did not stablize")
+ for attempt in (1, 2, 3, 4, 5):
+ if self.CM.cluster_stable():
+ return self.success()
+ return self.failure("Cluster did not stabilize")
did_fail=0
unstable = []
------------------------------
Message: 3
Date: Thu, 9 Feb 2006 22:00:36 -0700 (MST)
From: [email protected]
Subject: [Linux-ha-cvs] Linux-HA CVS: lrm by alan from
To: [EMAIL PROTECTED]
Message-ID: <[EMAIL PROTECTED]>
linux-ha CVS committal
Author : alan
Host :
Project : linux-ha
Module : lrm
Dir : linux-ha/lrm/lrmd
Modified Files:
lrmd.c
Log Message:
Fixed two very small lrmd errors.
Improved checking for bad lrmd file descriptors - hopefully catching things
sooner.
===================================================================
RCS file: /home/cvs/linux-ha/linux-ha/lrm/lrmd/lrmd.c,v
retrieving revision 1.212
retrieving revision 1.213
diff -u -3 -r1.212 -r1.213
--- lrmd.c 10 Feb 2006 04:06:34 -0000 1.212
+++ lrmd.c 10 Feb 2006 05:00:35 -0000 1.213
@@ -1,4 +1,4 @@
-/* $Id: lrmd.c,v 1.212 2006/02/10 04:06:34 alan Exp $ */
+/* $Id: lrmd.c,v 1.213 2006/02/10 05:00:35 alan Exp $ */
/*
* Local Resource Manager Daemon
*
@@ -1151,7 +1151,7 @@
for (j=0; j < 3; ++j) {
close(j);
- (void)open("/dev/null", j == 0 ? O_RDONLY : O_RDONLY);
+ (void)open("/dev/null", j == 0 ? O_RDONLY : O_WRONLY);
}
CL_IGNORE_SIG(SIGINT);
CL_IGNORE_SIG(SIGHUP);
@@ -3094,15 +3094,15 @@
cl_perror("%s::%d: dup2"
, __FUNCTION__, __LINE__);
}
+ close(stdout_fd[1]);
}
if (STDERR_FILENO != stderr_fd[1]) {
if (dup2(stderr_fd[1],
STDERR_FILENO)!=STDERR_FILENO) {
cl_perror("%s::%d: dup2"
, __FUNCTION__, __LINE__);
}
+ close(stderr_fd[1]);
}
- close(stdout_fd[1]);
- close(stderr_fd[1]);
RAExec = g_hash_table_lookup(RAExecFuncs,rsc->class);
if (NULL == RAExec) {
lrmd_log(LOG_ERR,"perform_ra_op: can not find
RAExec");
@@ -3380,9 +3380,9 @@
fd = op->ra_stdout_fd;
}
- if (fd < 0) {
+ if (fd <= 0) {
lrmd_log(LOG_CRIT
- , "%s:%d: Attempt to read from closed file descriptor %d."
+ , "%s:%d: Attempt to read from closed/invalid file
descriptor %d."
, __FUNCTION__, __LINE__, fd);
lrmd_op_dump(op, "op w/closed fd");
return FALSE;
@@ -3393,7 +3393,7 @@
G_main_del_fd(op->ra_stdout_gsource);
op->ra_stdout_gsource = NULL;
}
- if (op->ra_stdout_fd != -1) {
+ if (op->ra_stdout_fd >= 0) {
close(op->ra_stdout_fd);
op->ra_stdout_fd = -1;
}
@@ -3446,6 +3446,14 @@
return FALSE;
}
+ if (fd <= 0) {
+ lrmd_log(LOG_CRIT
+ , "%s:%d: Attempt to read from closed/invalid file
descriptor %d."
+ , __FUNCTION__, __LINE__, fd);
+ lrmd_op_dump(op, "op w/closed fd");
+ return FALSE;
+ }
+
if (0 != read_pipe(fd, &data, op)) {
if ( NULL != op->ra_stderr_gsource) {
G_main_del_fd(op->ra_stderr_gsource);
@@ -3645,6 +3653,10 @@
}
/*
* $Log: lrmd.c,v $
+ * Revision 1.213 2006/02/10 05:00:35 alan
+ * Fixed two very small lrmd errors.
+ * Improved checking for bad lrmd file descriptors - hopefully catching things
sooner.
+ *
* Revision 1.212 2006/02/10 04:06:34 alan
* Put in a fix suggested by Sun Jiang Dong for an lrm problem where
* we try and read from a closed file descriptor.
------------------------------
_______________________________________________
Linux-ha-cvs mailing list
[email protected]
http://lists.community.tummy.com/mailman/listinfo/linux-ha-cvs
End of Linux-ha-cvs Digest, Vol 27, Issue 43
********************************************
