Windows memory trending 🔗 link

I just noticed that all the memory trending graphs for Windows machines
are broken. They just show zero's all across the board. Is this a known
problem, or is something broken in my setup?

Thanks,
Larry Barber

On Tue, Jul 19, 2005 at 03:28:29PM -0400, user-7a6c75d6cc10@xymon.invalid wrote:

Do you have the bb-memory extension installed on your Windows boxes,
or just the BBNT client ?

If you dont know - is there a "memory" column for these boxes, or 
only the "cpu" column ? bb-memory generates a "memory" column, the
standard BBNT client does not.

I suspect it's the BBNT client that is feeding your memory graphs.
It has some problems with large-memory systems, but it could also be
a bug in Hobbit. Please send me the output from the "cpu" status on
one of these boxes.


Regards,
Henrik

We're just running the NT client, I have enough trouble getting the NT
admins to just install that, let alone anything else. It probably is the
memory size that's causing problems. Here's a sample of the return from
the cpu test:

Tue Jul 19 14:44:37 CDT 2005 [ANANKE] up: 83 days, 1 users, 40 procs,
load=1%, PhysicalMem: 3840MB(23%)


Memory Statistics
Total Physical memory:          4025925632 bytes (3.75GB)
Available Physical memory:      3102130176 bytes (2.89GB)
Total PageFile size:            4294967295 bytes (4.00GB)
Available PageFile size:        4294967295 bytes (4.00GB)
Total Virtual memory size:      2147352576 bytes (2.00GB)
Available Virtual memory size:  2119045120 bytes (1.97GB)

Most active processes
00.81%	tomcat5 (0x190 [400])
00.03%	System (0x4 [4])
00.02%	winlogon (0x2ac [684])
00.02%	Rtvscan (0x8a8 [2216])
00.00%	VxSvc (0x918 [2328])
00.00%	svchost (0x444 [1092])
00.00%	StarbaseServer (0x13d0 [5072])
00.00%	sqlservr (0x608 [1544])
00.00%	csrss (0x294 [660])


Thanks,
Larry Barber

On Tue, 2005-07-19 at 14:42 -0500, user-ce4a2c883f75@xymon.invalid wrote:

Now this is interesting, here's the 'cpu' output for a Windows box that
doesn't have as much memory as most of ours, that is displaying the same
behavior on the trending graphs:

Tue Jul 19 14:54:01 CDT 2005 [BARRSERVER] up: 6:14, 1 users, 32 procs,
load=1%, PhysicalMem: 512MB(32%)


Memory Statistics
Total Physical memory:           536334336 bytes (512.00MB)
Available Physical memory:       365035520 bytes (349.00MB)
Total PageFile size:            1307942912 bytes (1.22GB)
Available PageFile size:        1159303168 bytes (1.08GB)
Total Virtual memory size:      2147352576 bytes (2.00GB)
Available Virtual memory size:  2118647808 bytes (1.97GB)

Most active processes
00.06%	inetinfo (0x2e8 [744])
00.01%	bbnt (0x240 [576])

Thanks,
Larry Barber

On Tue, 2005-07-19 at 14:42 -0500, user-ce4a2c883f75@xymon.invalid wrote:

On Tue, Jul 19, 2005 at 03:52:08PM -0400, user-7a6c75d6cc10@xymon.invalid wrote:

Thanks, testing this clearly shows that it overflows the "unsigned long"
calculations being done by Hobbit. The attached patch should solve this,
hopefully also on those platforms that don't have a built-in strtoll()
routine.

This is for Hobbit 4.0.4 - the next version will have a (slightly
different) version of this included.


Regards,
Henrik

-------------- next part --------------
--- hobbit-4.0.4/hobbitd/larrd/do_la.c.orig	2005-05-08 21:36:09.000000000 +0200
+++ hobbit-4.0.4/hobbitd/larrd/do_la.c	2005-07-19 22:37:33.794305038 +0200
@@ -10,6 +10,26 @@
 
 static char la_rcsid[] = "$Id: do_la.c,v 1.15 2005/05/08 19:35:29 henrik Exp $";
 
+static long long str2ll(char *s, char **errptr)
+{
+	long long result = 0;
+	int negative = 0;
+	char *inp;
• +	inp = s + strspn(s, " \t");
+	if (*inp == '-') { negative = 1; inp++; }
+	while (isdigit((int)*inp)) { 
+		result = 10*result + (*inp - '0'); 
+		inp++;
+	}
• +	if (errptr && (*inp != '\0') && (!isspace((int)*inp))) *errptr = inp;
• +	if (negative) result = -result;
• +	return result;
+}
• int do_la_larrd(char *hostname, char *testname, char *msg, time_t tstamp)
 {
 	static char *la_params[]          = { "rrdcreate", rrdfn, "DS:la:GAUGE:600:0:U", rra1, rra2, rra3, rra4, NULL };
@@ -189,27 +209,27 @@
 	if (memhosts_init && (rbtFind(memhosts, hostname) == rbtEnd(memhosts))) {
 		/* Pick up memory statistics */
 		int found, realuse, swapuse;
-		unsigned long phystotal, physavail, pagetotal, pageavail;
+		long long phystotal, physavail, pagetotal, pageavail;
 
 		found = realuse = swapuse = 0;
 		phystotal = physavail = pagetotal = pageavail = 0;
 
 		p = strstr(msg, "Total Physical memory:");
-		if (p) { found++; phystotal = atol(strchr(p, ':') + 1); }
+		if (p) { found++; phystotal = str2ll(strchr(p, ':') + 1, NULL); }
 
 		if (found == 1) {
 			p = strstr(msg, "Available Physical memory:");
-			if (p) { found++; physavail = atol(strchr(p, ':') + 1); }
+			if (p) { found++; physavail = str2ll(strchr(p, ':') + 1, NULL); }
 		}
 
 		if (found == 2) {
 			p = strstr(msg, "Total PageFile size:"); 
-			if (p) { found++; pagetotal = atol(strchr(p, ':') + 1); }
+			if (p) { found++; pagetotal = str2ll(strchr(p, ':') + 1, NULL); }
 		}
 
 		if (found == 2) {
 			p = strstr(msg, "Available PageFile size:"); 
-			if (p) { found++; pageavail = atol(strchr(p, ':') + 1); }
+			if (p) { found++; pageavail = str2ll(strchr(p, ':') + 1, NULL); }
 		}
 
 		if (found == 4) {

On Tue, Jul 19, 2005 at 10:43:44PM +0200, Henrik Stoerner wrote:

For those not familiar with patching, to apply this you must:

  cd hobbit-4.0.4
  patch -p1 </tmp/hobbit-4.0.4-win32mem.patch
  make
  stop hobbit
  as root: "make install"
  start hobbit

Henrik