From: Russ Allbery <rra@stanford.edu>
Date: Wed, 9 Nov 2005 20:43:04 +0000 (+0000)
Subject: rxdebug -noconn tells us how many waiting connections there are; there's
X-Git-Url: https://git.michaelhowe.org/gitweb/?a=commitdiff_plain;h=6489294af19b1114a499472aa8a8ced5cd4547dc;p=packages%2Fa%2Fafs-monitor.git

rxdebug -noconn tells us how many waiting connections there are; there's
no need to parse the whole output and study it.  Also, it's helpful to
look for waiting connections to the file server, rather than waiting
connections to the local cache manager (doh).  Lower the warning threshold
to two from four while we're at it.
---

diff --git a/check_rxdebug b/check_rxdebug
index f1f002e..3ee184e 100755
--- a/check_rxdebug
+++ b/check_rxdebug
@@ -1,18 +1,18 @@
 #!/usr/bin/perl -w
 $ID = q$Id$;
 #
-# check_rxdebug -- Check AFS servers for blocked connections in Nagios.
+# check_rxdebug -- Nagios AFS server check for waiting connections.
 #
 # Written by Quanah Gibson-Mount based on work by Neil Crellin
 # Updated by Russ Allbery <rra@stanford.edu>
-# Copyright 2003, 2004 Board of Trustees, Leland Stanford Jr. University
+# Copyright 2003, 2004, 2005 Board of Trustees, Leland Stanford Jr. University
 #
 # This program is free software; you may redistribute it and/or modify it
 # under the same terms as Perl itself.
 #
 # Expects a file server with the -H option and runs rxdebug against that file
-# server, looking for any connections that are waiting for a process.  Exits
-# with status 1 if there are more than four connections in that state (a
+# server, looking for any connections that are waiting for a thread.  Exits
+# with status 1 if there are more than two connections in that state (a
 # warning) and with status 2 if there are more than eight connections in that
 # state.  The thresholds can be overridden from the command line.
 
@@ -22,7 +22,7 @@ $ID = q$Id$;
 
 # The default count of blocked connections at which to warn or send a critical
 # alert.  These can be overridden with the -w and -c command-line options.
-$WARNINGS = 4;
+$WARNINGS = 2;
 $CRITICAL = 8;
 
 # The default timeout in seconds (implemented by alarm) for rxdebug.
@@ -86,19 +86,26 @@ alarm ($TIMEOUT);
 
 # Run rxdebug and parse the output, counting the number of waiting for process
 # connections that we have.
-unless (open (RXDEBUG, "$RXDEBUG $host 7001 |")) {
+unless (open (RXDEBUG, "$RXDEBUG $host -noconn |")) {
     warn "$0: cannot run rxdebug\n";
     exit 3;
 }
-my $blocked = 0;
+my $blocked;
 while (<RXDEBUG>) {
-    $blocked++ if /waiting_for_process/;
+    if (/^(\d+) calls waiting for a thread/) {
+        $blocked = $1;
+        last;
+    }
 }
 close RXDEBUG;
 if ($? != 0) {
     print "AFS CRITICAL: cannot contact server\n";
     exit 2;
 }
+unless (defined $blocked) {
+    print "AFS CRITICAL: cannot parse rxdebug output\n";
+    exit 2;
+}
 
 # Check the connection count against our limits and make sure that it's okay.
 if ($blocked >= $CRITICAL) {
@@ -132,7 +139,7 @@ there are client connections waiting for a free thread.  If there are more
 than a few of these, AFS performance tends to be very slow; this is a fairly
 reliable way to catch overloaded file servers.  By default, B<check_rxdebug>
 returns a critical error if there are more than eight connections waiting
-for a free thread and a warning if there are more than four.  These
+for a free thread and a warning if there are more than two.  These
 thresholds can be changed with the B<-c> and B<-w> options.
 
 B<check_rxdebug> will always print out a single line of output including the
@@ -170,7 +177,7 @@ Print out the version of B<check_rxdebug> and quit.
 =item B<-w> I<threshold>, B<--warning>=I<threshold>
 
 Change the warning blocked connection threshold to I<threshold>, which
-should be an integer.  The default is 4.
+should be an integer.  The default is 2.
 
 =back
 
@@ -212,7 +219,7 @@ more uniform coding style.
 
 =head1 COPYRIGHT AND LICENSE
 
-Copyright 2003, 2004 Board of Trustees, Leland Stanford Jr. University.
+Copyright 2003, 2004, 2005 Board of Trustees, Leland Stanford Jr. University.
 
 This program is free software; you may redistribute it and/or modify it
 under the same terms as Perl itself.