| 13 |
|
############################################################ |
| 14 |
|
|
| 15 |
|
# for computing machines |
| 16 |
< |
group.computing=Host.raptor.ukc.ac.uk;Host.slate.ukc.ac.uk;Host.granite.ukc.ac.uk; |
| 17 |
< |
config.rocks=rocks.conf |
| 16 |
> |
group.computing=Host.129.12.4.*; |
| 17 |
> |
config.computing=computing.conf |
| 18 |
|
|
| 19 |
|
# for library machines |
| 20 |
< |
group.library=Host.killigrew.ukc.ac.uk; |
| 20 |
> |
group.library=Host.129.12.58.*; |
| 21 |
|
config.library=library.conf |
| 22 |
|
|
| 23 |
+ |
# for SSB machines |
| 24 |
+ |
group.ssb=Host.stue*.ukc.ac.uk; |
| 25 |
+ |
config.ssb=ssb.conf |
| 26 |
+ |
|
| 27 |
|
# mySQL configuration |
| 28 |
|
config.mySQL=mySQL.conf |
| 29 |
|
|
| 30 |
+ |
############################################################ |
| 31 |
+ |
#### Misc system-wide configuration |
| 32 |
+ |
############################################################ |
| 33 |
|
|
| 34 |
+ |
# the time (in seconds) that checks should be made for |
| 35 |
+ |
# changes in configuration |
| 36 |
+ |
ConfigurationProxy.updateTime=60 |
| 37 |
|
|
| 38 |
+ |
# The time interval at which the Queue's should be monitored |
| 39 |
+ |
Queue.MonitorInterval=15 |
| 40 |
+ |
|
| 41 |
+ |
|
| 42 |
+ |
|
| 43 |
|
############################################################ |
| 44 |
|
#### FilterManager configuration |
| 45 |
|
############################################################ |
| 94 |
|
############################################################ |
| 95 |
|
|
| 96 |
|
# The default filter for a new host |
| 97 |
< |
Host.filter=computingFilter |
| 97 |
> |
Host.filter=computingFilter; |
| 98 |
|
|
| 99 |
|
# The default update times for a new host |
| 100 |
|
Host.UDPUpdateTime=10 |
| 101 |
|
Host.TCPUpdateTime=60 |
| 102 |
+ |
Host.AveragerUpdateTime=5 |
| 103 |
|
|
| 104 |
|
# The services checks to run on a host |
| 105 |
|
Host.serviceChecksPackage=uk.ac.ukc.iscream.filter.plugins |
| 107 |
|
|
| 108 |
|
|
| 109 |
|
|
| 94 |
– |
# The time interval at which the Queue's should be monitored |
| 95 |
– |
Queue.MonitorInterval=15 |
| 96 |
– |
|
| 97 |
– |
|
| 98 |
– |
|
| 110 |
|
############################################################ |
| 111 |
|
#### Monitor Configuration |
| 112 |
|
############################################################ |
| 113 |
|
|
| 114 |
|
# the monitors to run |
| 115 |
|
Monitor.PluginsPackage=uk.ac.ukc.iscream.client.monitors |
| 116 |
< |
Monitor.Plugins=CPU; |
| 116 |
> |
Monitor.Plugins=CPU;Heartbeat |
| 117 |
|
|
| 118 |
< |
# CPU monitor configuration |
| 119 |
< |
Monitor.CPU.level.1=90 |
| 120 |
< |
Monitor.CPU.level.2=95 |
| 121 |
< |
Monitor.CPU.level.3=100 |
| 118 |
> |
## CPU monitor configuration |
| 119 |
> |
# threshold values |
| 120 |
> |
# |
| 121 |
> |
# note: if the UPPER threshold is broken, |
| 122 |
> |
# the alertTimeout values are halved. |
| 123 |
> |
# This is to escalate the alert quicker. |
| 124 |
> |
Monitor.CPU.threshold.LOWER=90 |
| 125 |
> |
Monitor.CPU.threshold.UPPER=99 |
| 126 |
|
|
| 127 |
+ |
# timings of the alerts (in seconds) |
| 128 |
+ |
Monitor.CPU.alertTimeout.NOTICE=60 |
| 129 |
+ |
Monitor.CPU.alertTimeout.WARNING=900 |
| 130 |
+ |
Monitor.CPU.alertTimeout.CAUTION=1800 |
| 131 |
+ |
Monitor.CPU.alertTimeout.CRITICAL=3600 |
| 132 |
|
|
| 133 |
+ |
## Heartbeat monitor configuration |
| 134 |
+ |
# threshold values |
| 135 |
+ |
# |
| 136 |
+ |
# note: if the UPPER threshold is broken, |
| 137 |
+ |
# the alertTimeout values are halved. |
| 138 |
+ |
# This is to escalate the alert quicker. |
| 139 |
+ |
Monitor.Heartbeat.checkPeriod=50 |
| 140 |
+ |
Monitor.Heartbeat.threshold.LOWER=120 |
| 141 |
+ |
Monitor.Heartbeat.threshold.LOWER=300 |
| 142 |
|
|
| 143 |
+ |
# timings of the alerts (in seconds) |
| 144 |
+ |
Monitor.Heartbeat.alertTimeout.NOTICE=60 |
| 145 |
+ |
Monitor.Heartbeat.alertTimeout.WARNING=900 |
| 146 |
+ |
Monitor.Heartbeat.alertTimeout.CAUTION=1800 |
| 147 |
+ |
Monitor.Heartbeat.alertTimeout.CRITICAL=3600 |
| 148 |
+ |
|
| 149 |
|
############################################################ |
| 150 |
|
#### Alerter Configuration |
| 151 |
|
############################################################ |
| 155 |
|
Alerter.Plugins=EMail;IRC; |
| 156 |
|
|
| 157 |
|
# EMail alerter configuration |
| 158 |
< |
Alerter.EMail.level = 2 |
| 159 |
< |
Alerter.EMail.destList = dev@i-scream.org.uk |
| 158 |
> |
Alerter.EMail.level = CAUTION |
| 159 |
> |
Alerter.EMail.defaultDestList = dev@i-scream.org.uk |
| 160 |
|
Alerter.EMail.sender = dev@i-scream.org.uk |
| 161 |
|
Alerter.EMail.smtpServer = mercury.ukc.ac.uk |
| 162 |
< |
Alerter.EMail.subject = i-scream alert: level %level% alert on %source for %attributeName% |
| 163 |
< |
Alerter.EMail.message = The i-scream distributed central monitoring system has\nraised a level %level% alert for the host %source%.\n\nThe value for %attributeName% of %value% has exceeded the\nthreshold value of %thresholdValue%. |
| 162 |
> |
Alerter.EMail.subject = i-scream alert: %level% alert on %source% for %attributeName% |
| 163 |
> |
Alerter.EMail.message = The i-scream distributed central monitoring system has\nraised a %level% alert for the host %source%.\n\nThe value for %attributeName% of %value% has exceeded the\n%threshold% threshold value of %thresholdValue%.\n\nThe next alert (should one occur) will be sent in %timeTillNextAlert%. |
| 164 |
|
|
| 165 |
|
# IRC alerter configuration |
| 166 |
< |
Alerter.IRC.level = 1 |
| 166 |
> |
Alerter.IRC.level = OK |
| 167 |
|
Alerter.IRC.IRCServer = killigrew.ukc.ac.uk |
| 168 |
|
Alerter.IRC.IRCPort = 6667 |
| 169 |
< |
Alerter.IRC.nick = iscreamBot |
| 169 |
> |
Alerter.IRC.nickList = iscreamBot;_iscreamBot;i-screamBot |
| 170 |
> |
Alerter.IRC.user = i-scream |
| 171 |
> |
Alerter.IRC.comment = i-scream alerting bot |
| 172 |
|
Alerter.IRC.channel = #i-scream |
| 173 |
< |
Alerter.IRC.message = level %level%: %attributeName% on %source% has reached %value% |
| 173 |
> |
Alerter.IRC.message = %level%: %attributeName% on %source% has passed %threshold%(%thresholdValue%) threshold with %value% - time till next alert (should one occur), %timeTillNextAlert% |
| 174 |
> |
Alerter.IRC.reconnectDelay = 30 |
| 175 |
> |
Alerter.IRC.startupNotice = i-scream alerting bot activated |
| 176 |
> |
Alerter.IRC.stopCommand = stop alerts |
| 177 |
> |
Alerter.IRC.startCommand = start alerts |
| 178 |
> |
Alerter.IRC.lastAlertCommand = last alert |
| 179 |
> |
Alerter.IRC.joinCommand = join |
| 180 |
> |
Alerter.IRC.nickChangeCommand = nick |
| 181 |
> |
Alerter.IRC.statCommand = statistics |
| 182 |
> |
Alerter.IRC.uptimeCommand = uptime |
| 183 |
> |
Alerter.IRC.timeSinceLastAlertCommand = time since last alert |
| 184 |
> |
Alerter.IRC.versionCommand = version |
| 185 |
> |
Alerter.IRC.helpCommand = help |
| 186 |
> |
Alerter.IRC.rejectMessage = sorry, I don't understand your request |
| 187 |
|
|
| 188 |
< |
############################################################ |
| 188 |
> |
############################################################ |
| 189 |
> |
|
