Add smart_

git-svn-id: svn+ssh://asteria.noreply.org/svn/weaselutils/trunk@125 bc3d92e2-beff-0310-a7cd-cc87d7ac0ede

1 files changed, 344 insertions, 0 deletions

diff --git a/munin/smart_ b/munin/smart_
new file mode 100755
index 0000000..9d1d062
--- /dev/null
+++ b/munin/smart_
@@ -0,0 +1,344 @@
+#!/usr/bin/env python
+
+# taken from the debian munin-node package, version 1.2.4-1
+###########################################################
+
+# -*- encoding: iso-8859-1 -*-
+#
+# Wildcard-plugin to monitor S.M.A.R.T attribute values through smartctl,
+# which is part of smartmontools package:
+#         http://smartmontools.sourceforge.net/
+# 
+# To monitor a S.M.A.R.T device, link smart_<device> to this file. 
+# E.g.
+#    ln -s /usr/share/munin/plugins/smart_ /etc/munin/plugins/smart_hda
+# ...will monitor /dev/hda.
+#
+# Needs following minimal configuration in plugin-conf.d/munin-node:
+#   [smart_*]
+#   user root
+#   group disk
+#
+# Parameters
+# 	smartpath   - Specify path to smartctl program (Default: /usr/sbin/smartctl)
+# 	smartargs   - Override '-a' argument passed to smartctl with '-A -i'+smartargs
+# 
+# Parameters can be specified on a per-drive basis, eg:
+#   [smart_hda]
+#   user root
+#   group disk
+#   env.smartargs -H -c -l error -l selftest -l selective -d ata
+#   env.smartpath /usr/local/sbin/smartctl
+#
+#   [smart_twa0-1]
+#   user root
+#   group disk
+#   env.smartargs -H -l error -d 3ware,1
+#
+#   [smart_twa0-2]
+#   user root
+#   group disk
+#   env.smartargs -H -l error -d 3ware,2
+#
+# Author: Nicolas Stransky <Nico@neo-lan.net>
+# 
+# v1.0 22/08/2004 - First draft
+# v1.2 28/08/2004 - Clean up the code, add a verbose option
+# v1.3 14/11/2004 - Compatibility with python<2.2. See comments in the code
+# v1.4 17/11/2004 - Deal with non zero exit codes of smartctl
+#                 - config now prints the critical thresholds, as reported by smartctl
+# v1.5 18/11/2004 - Plot smartctl_exit_code bitmask
+# v1.6 21/11/2004 - Add autoconf and suggest capabilities
+#                 - smartctl path can be passed through "smartpath" environment variable
+#                 - Additional smartctl args can be passed through "smartargs" environment variable
+# v1.7 29/11/2004 - Add suggest capabilities for NetBSD, OpenBSD, FreeBSD and SunOS.
+#                 - Allow to override completely the smartctl arguments with "smartargs"
+# v1.8 16/02/2005 - Exit status field now only triggers warnings, not criticals.
+# v1.9 07/07/2005 - Allow to query several drives on the same 3ware card.
+#                 - Correct a bug when '-i' was not listed in smartargs
+#                 - Don't fail if no value was obtained for hard drive model
+# v1.10 19/08/2005 - smartctl_exit_code is now a numerical value
+#
+# Copyright (c) 2004,2005 Nicolas Stransky.
+#
+# Permission to use, copy, and modify this software with or without fee
+# is hereby granted, provided that this entire notice is included in
+# all source code copies of any software which is or includes a copy or
+# modification of this software.
+#
+# THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR
+# IMPLIED WARRANTY. IN PARTICULAR, NONE OF THE AUTHORS MAKES ANY
+# REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE
+# MERCHANTABILITY OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR
+# PURPOSE.
+#
+#
+# Magic markers
+#%# capabilities=autoconf suggest
+#%# family=auto
+
+import os, sys, string, pickle
+from math import log
+
+# Increase verbosity of the plugin by putting verbose=True (for 
+# debugging purposes)
+# For python < 2.2 compliance, put verbose=0 instead of "False"
+verbose=False
+
+# Modify to your needs:
+statefiledir='/var/lib/munin/plugin-state/'
+
+
+def verboselog(s):
+    sys.stderr.write(plugin_name+': '+s+'\n')
+
+if not verbose :
+    verboselog = lambda s: None
+
+def read_values(hard_drive) :
+    num_exit_status=0
+    try :
+        verboselog('Reading S.M.A.R.T values')
+        os.putenv('LC_ALL','C')
+        smart_output=os.popen(os.getenv('smartpath','/usr/sbin/smartctl')+' '+os.getenv('smartargs','-a')+' -A -i /dev/'+hard_drive)
+        read_values=0
+        # For python < 2.2 compliance, switch comments on next 2 lines.
+        for l in smart_output :
+        #for l in smart_output.readlines() :
+            if l[:-1]=='' :
+                read_values=0
+            elif l[:13]=='Device Model:' or l[:7]=='Device:' :
+                model_list=string.split(string.split(l,':')[1])
+                try: model_list.remove('Version')
+                except : None
+                model=string.join(model_list)
+            if read_values==1 :
+                smart_attribute=string.split(l)
+                smart_values[string.replace(smart_attribute[1],'-','_')]={"value":smart_attribute[3],"threshold":smart_attribute[5]}
+            elif l[:18]=="ID# ATTRIBUTE_NAME" :
+                # Start reading the Attributes block
+                read_values=1
+        exit_status=smart_output.close()
+        if exit_status!=None :
+            # smartctl exit code is a bitmask, check man page.
+            num_exit_status=int(exit_status/256)
+            if num_exit_status<=2 :
+                verboselog('smartctl cannot access S.M.A.R.T values on drive '+hard_drive+'. Command exited with code '+str(num_exit_status)+' ('+str(exit_status/256)+')')
+                sys.exit(1)
+            else :
+                verboselog('smartctl exited with code '+str(num_exit_status)+'. '+hard_drive+' may be FAILING RIGHT NOW !')
+    except :
+        verboselog('Cannot access S.M.A.R.T values ! Check user rights or propper smartmontools installation. Quitting...')
+        sys.exit(1)
+    if smart_values=={} :
+        verboselog('Can\'t find any S.M.A.R.T value to plot ! Quitting...')
+        sys.exit(1)
+    smart_values["smartctl_exit_status"]={"value":str(num_exit_status),"threshold":"1"}
+    try: smart_values["model"]=model
+    # For some reason we may have no value for "model"
+    except : smart_values["model"]="unknown"
+
+def open_state_file(mode) :
+    return open(statefiledir+'/smart-'+string.join(hard_drive,"-")+'.state',mode)
+
+def update_state_file() :
+    try:
+        verboselog('Saving statefile')
+        pickle.dump(smart_values,open_state_file("w"))
+    except :
+        verboselog('Error trying to save state file ! Check access rights')
+
+def print_plugin_values() :
+    verboselog('Printing S.M.A.R.T values')
+    for key in smart_values.keys() :
+        if key=="model" : continue
+        print(key+".value "+smart_values[key]["value"])
+
+def print_config(hard_drive,smart_values) :
+    if os.path.exists(statefiledir+'/smart-'+string.join(hard_drive,"-")+'.state'):
+        try :
+            verboselog('Try to recall previous S.M.A.R.T attributes for '+string.join(hard_drive,","))
+            smart_values_state=pickle.load(open_state_file("r"))
+        except :
+            verboselog('Error opening existing state file !')
+            sys.exit(1)
+    else :
+        verboselog('No state file, reading S.M.A.R.T values for the first time')
+        read_values(hard_drive[0])
+        pickle.dump(smart_values,open_state_file("w"))
+        smart_values_state=smart_values 
+
+    verboselog('Printing configuration')
+    print('graph_title S.M.A.R.T values for drive '+string.join(hard_drive,","))
+    print('graph_vlabel Attribute S.M.A.R.T value')
+    print('graph_args --base 1000 --lower-limit 0')
+    print('graph_category disk')
+    print('graph_info This graph shows the value of all S.M.A.R.T attributes of drive '+string.join(hard_drive,",")+' ('+smart_values_state['model']+'). smartctl_exit_status is the return value of smartctl. A non-zero return value indicates an error, a potential error, or a fault on the drive.')
+    attributes=smart_values_state.keys()
+    attributes.sort()
+    for key in attributes :
+        if key in ['smartctl_exit_status','model'] : continue
+        print(key+'.label '+key)
+        print(key+'.draw LINE2')
+        # Following line can be commented out if necessary. Only useful for warning reports about S.M.A.R.T values
+        print(key+'.critical '+smart_values_state[key]["threshold"]+':')
+    # Let's place smartctl_exit_status at the end
+    print('smartctl_exit_status.label smartctl_exit_status')
+    print('smartctl_exit_status.draw LINE2')
+    # Following line can be commented out if necessary. Only useful for warning reports about S.M.A.R.T values
+    print('smartctl_exit_status.warning '+smart_values_state['smartctl_exit_status']["threshold"])
+
+def get_hard_drive_name() :
+    try :
+        hard_drive=[plugin_name[string.rindex(plugin_name,'_')+1:]]
+        if os.uname()[0]=="SunOS" :
+            try :
+                # if hard_drive name starts with "rdsk" or "rmt", try to reconstruct the path
+                if hard_drive[0][0:4]=="rdsk":
+                    hard_drive[0]=os.path.join("rdsk",hard_drive[0][4:])
+                elif hard_drive[0][0:3]=="rmt":
+                    hard_drive[0]=os.path.join("rmt",hard_drive[0][3:])
+            except :
+                verboselog('Failed to find SunOS hard_drive')
+        # For 3ware cards, we have to set multiple plugins for the same hard drive name.
+        # Let's see if we find a '-' in the drive name.
+        if hard_drive[0].find('-')!=-1:
+            # Put the drive name and it's number in a list
+            hard_drive=[hard_drive[0][:string.rindex(hard_drive[0],'-')],hard_drive[0][string.rindex(hard_drive[0],'-')+1:]]
+        # Chech that the drive exists in /dev
+        if not os.path.exists('/dev/'+hard_drive[0]):
+            verboselog('/dev/'+hard_drive[0]+' not found !')
+            sys.exit(1)
+        return(hard_drive)
+    except :
+        verboselog('No S.M.A.R.T device name found in plugin\'s symlink !')
+        sys.exit(1)
+
+def find_smart_drives() :
+    # Try to autodetect Linux, *BSD, SunOS drives. Don't try to autodetect drives on a 3Ware card.
+    drives=[]
+    if os.uname()[0]=="Linux" :
+        if os.path.exists('/sys/block/'):
+            # Running 2.6, yeah
+            try :
+                for drive in os.listdir('/sys/block/') :
+                    if drive[:2] in ['md','fd','lo','ra','dm'] : continue # Ignore MD, Floppy, loop , RAM and LVM devices.
+                    try :
+                        read_values(drive)
+                        drives.append(drive)
+                    except :
+                        continue
+            except : 
+                verboselog('Failed to list devices in /sys/block')
+        else :
+            verboselog('Not running linux2.6, failing back to /proc/partitions')
+            try :
+                partitions=open('/proc/partitions','r')
+                L=partitions.readlines()
+                for l in L :
+                    words=string.split(l)
+                    if len(words)==0 or words[0][0] not in string.digits : continue
+                    if words[0] in ['1','9','58','254'] : continue # Ignore RAM, md, LVM and LVM2 devices
+                    if words[-1][-1] not in string.digits : 
+                        try :
+                            read_values(words[-1])
+                            drives.append(words[-1])
+                        except :
+                            continue
+                verboselog('Found drives in /proc/partitions ! '+str(drives))
+            except :
+                verboselog('Failed to list devices in /proc/partitions')
+    elif os.uname()[0]=="OpenBSD" :
+        try :
+            sysctl_kerndisks=os.popen('sysctl hw.disknames')
+            kerndisks=string.strip(sysctl_kerndisks.readline())
+            for drive in string.split(kerndisks[string.rindex(kerndisks,'=')+1:],',') :
+                if drive[:2] in ['md','cd','fd'] : continue # Ignore Memory Disks, CD-ROM drives and Floppy
+                try :
+                    read_values(drive+'c')
+                    drives.append(drive+'c')
+                except :
+                    continue
+        except :
+            verboselog('Failed to list OpenBSD disks')
+    elif os.uname()[0]=="FreeBSD" :
+        try :
+            sysctl_kerndisks=os.popen('sysctl kern.disks')
+            kerndisks=string.strip(sysctl_kerndisks.readline())
+            for drive in string.split(kerndisks)[1:] :
+                if drive[:2] in ['md','cd','fd'] : continue # Ignore Memory Disks, CD-ROM drives and Floppy
+                try :
+                    read_values(drive)
+                    drives.append(drive)
+                except :
+                    continue
+        except :
+            verboselog('Failed to list FreeBSD disks')
+    elif os.uname()[0]=="NetBSD" :
+        try :
+            sysctl_kerndisks=os.popen('sysctl hw.disknames')
+            kerndisks=string.strip(sysctl_kerndisks.readline())
+            for drive in string.split(kerndisks)[2:] :
+                if drive[:2] in ['md','cd','fd'] : continue # Ignore Memory Disks, CD-ROM drives and Floppy
+                try :
+                    read_values(drive+'c')
+                    drives.append(drive+'c')
+                except :
+                    continue
+        except :
+            verboselog('Failed to list NetBSD disks')
+    elif os.uname()[0]=="SunOS" :
+        try :
+            from glob import glob
+            for drivepath in glob('/dev/rdsk/*s2') :
+                try :
+                    drive=os.path.basename(drivepath)
+                    read_values('rdsk'+drive)
+                    drives.append('rdsk'+drive)
+                except :
+                    continue
+            for drivepath in glob('/dev/rmt/*') :
+                try :
+                    drive=os.path.basename(drivepath)
+                    read_values('rmt'+drive)
+                    drives.append('rmt'+drive)
+                except :
+                    continue
+        except :
+            verboselog('Failed to list SunOS disks')
+    return(drives)
+
+### Main part ###
+
+smart_values={}
+model=''
+plugin_name=list(os.path.split(sys.argv[0]))[1]
+verboselog('plugins\' UID: '+str(os.geteuid())+' / plugins\' GID: '+str(os.getegid()))
+
+# Parse arguments
+if len(sys.argv)>1 :
+    if sys.argv[1]=="config" :
+        hard_drive=get_hard_drive_name()
+        print_config(hard_drive,smart_values)
+        sys.exit(0)
+    elif sys.argv[1]=="autoconf" :
+        if os.path.exists(os.getenv('smartpath','/usr/sbin/smartctl')) : 
+            print('yes')
+            sys.exit(0)
+        else : 
+            print('no (smartmontools not found)')
+            sys.exit(1)
+    elif sys.argv[1]=="suggest" :
+        for drive in find_smart_drives() :
+            print(drive)
+        sys.exit(0)
+    elif sys.argv[1]!="" :
+        verboselog('unknown argument "'+sys.argv[1]+'"')
+        sys.exit(1)
+
+# No argument given, doing the real job:
+hard_drive=get_hard_drive_name()
+read_values(hard_drive[0])
+update_state_file()
+print_plugin_values()
+