sql-server – 监控SQL Server 2008

我知道除非操作系统需要,否则sql Server不会释放内存.然后,监视可用字节(可用内存)不是监视服务的最佳方法.还有哪些其他变量可以让我真正衡量sql Server的行为？可能是Pages / sec或Page Faults / sec？
我正在使用nagios来监控服务,有时会出现警报,因为执行了大查询.

解决方法

您可以使用Perfmon计数器,特别是总服务器内存,使用sql获得正确的内存使用量.使用Nagios,您可以使用 NSClient++和nrpe来轮询这些计数器.以下是一些sql计数器供您参考(具有愚蠢的警告和关键级别)：

define service {
        service_description     sql Buffer Cache Hit Ratio
        use                     generic-service
        hostgroup_name          database_servers
        check_command           check_nrpe_counter!BufferCacheHitRatio!\\sqlServer:Buffer Manager\\Buffer cache hit ratio!100!200
}

define service {
        service_description     sql Full Scans Per Sec
        use                     generic-service
        hostgroup_name          database_servers
        check_command           check_nrpe_counter!scans!\\sqlServer:Access Methods\\Full Scans/sec!100!200
}

define service {
        service_description     sql Latch Wait Time in MS
        use                     generic-service
        hostgroup_name          database_servers
        check_command           check_nrpe_counter!ms!\\sqlServer:Latches\\Total Latch Wait Time (ms)!100!200
}

define service {
        service_description     sql Batch Req Per Sec
        use                     generic-service
        hostgroup_name          database_servers
        check_command           check_nrpe_counter!req!\\sqlServer:sql Statistics\\Batch Requests/sec!100!200
}

define service {
        service_description     sql Re-Compilations Per Sec
        use                     generic-service
        hostgroup_name          database_servers
        check_command           check_nrpe_counter!req!\\sqlServer:sql Statistics\\sql Re-Compilations/sec!100!200
}

define service {
        service_description     sql Total Memory in KB
        use                     generic-service
        hostgroup_name          database_servers
        check_command           check_nrpe_counter!mem!\\sqlServer:Memory Manager\\Total Server Memory (KB)!100!200
}

define service {
        service_description     sql Memory Grants Pending
        use                     generic-service
        hostgroup_name          database_servers
        check_command           check_nrpe_counter!PenMemGrant!\\sqlServer:Memory Manager\\Memory Grants Pending!100!200
}

#define service {
#        service_description     sql Mirror Log Remaining For Undo
#        use                     generic-service
#        check_command           check_nrpe_counter!kb!\\sqlServer:Database Mirroring(StackOverflow)\\Log Remaining for Undo KB!100!200
#}

#define service {
#        service_description     sql Mirror Log Send Queue
#        use                     generic-service
#        check_command           check_nrpe_counter!kb!\\sqlServer:Database Mirroring(StackOverflow)\\Log Remaining for Undo KB!100!200
#}

#define service {
#        service_description     sql Mirror Redo Queue
#        use                     generic-service
#        check_command           check_nrpe_counter!kb!\\sqlServer:Database Mirroring(StackOverflow)\\Log Remaining for Undo KB!100!200
#}

define service {
        service_description     sql Number of Connected Users
        use                     generic-service
        hostgroup_name          database_servers
        check_command           check_nrpe_counter!UserCon!\\sqlServer:General Statistics\\User Connections!100!200
}

示例命令Def：

define command {
  command_name check_nrpe_counter
  command_line $USER1$/check_nrpe -H $HOSTADDRESS$-c CheckCounter -a "Counter:$ARG1$=$ARG2$" ShowAll MaxWarn=$ARG3$MaxCrit=$ARG4$
}

最后,您可能对Brent Ozar推荐的sql Perfmon计数器感兴趣：http://www.brentozar.com/archive/2006/12/dba-101-using-perfmon-for-sql-performance-tuning/