CLUSTER.LOG (AlwaysOn Availability Groups)
As a failover cluster resource, there are external interactions between SQL Server, the Windows Failover Cluster Service (WSFC) cluster, and the SQL Server resource DLL (hadrres.dll), that cannot be monitored within SQL Server. The WSFC log, CLUSTER.LOG, can diagnose issues in the WSFC cluster or in the SQL Server resource DLL.
The following diagram demonstrates the relationship between applications like SQL Server and Windows Cluster Manager that initiate availability group resource creation, destruction or state changes
Generate Cluster Log
You can generate the cluster logs in two ways:
cluster /log /gcommand at the command prompt. This command generates the cluster logs to the \windows\cluster\reports directory on each WSFC node. The advantage of this method is that you can specify the level of detail in the generated logs by using the
/leveloption. The disadvantage is that you cannot specify the destination directory for the generated cluster logs. For more information, see How to create the cluster.log in Windows Server 2008 Failover Clustering.
Use the Get-ClusterLog PowerShell cmdlet. The advantage of this method is that you can generate the cluster log from all nodes to one destination directory on the node that you run the cmdlet. The disadvantage is that you cannot specify the level of detail in the generated logs.
The following PowerShell commands generate the cluster logs from all cluster nodes from the last 15 minutes and place them in the current directory. You should run the commands in a PowerShell window with Administrative privileges.
Import-Modeul FailoverClusters Get-ClusterLog –TimeSpan 15 –Destination .
AlwaysOn Log Verbosity
You can increase the verbosity of the logs in CLUSTER.LOG for an availability group. To do this, follow the steps below:
From the Start menu, open the Failover Cluster Manger.
Expand the your cluster and the Services and applications node, then click the availability group name.
In the detail pane, right-click the availability group resource and click Properties.
Click the Properties tab.
Modify the VerboseLogging property. By default, VerboseLogging is set to
0which reports information, warnings and errors. VerboseLogging can be set from
Right-click the availability group resource again and click Take this resource offline.
Right-click the availability group resource again and click Bring this resource online.
Availability Group Resource Events
The table below shows the different kinds of events you can see in CLUSTER.LOG that pertain to the availability group resource. For more information on the Resource Hosting Subsystem (RHS) and Resource Control Monitor (RCM) in WSFC, see Resource Hosting Subsystem (RHS) In Windows Server 2008 Failover Clusters.
|Identifier||Source||Example from CLUSTER.LOG|
|Messages prefixed with RES and '[hadrag]'||hadrres.dll (AlwaysOn Resource DLL)||00002cc4.00001264::2011/08/05-13:47:42.543 INFO [RES] SQL Server Availability Group <ag>: [hadrag] Offline request.
00002cc4.00003384::2011/08/05-13:47:42.558 ERR [RES] SQL Server Availability Group <ag>: [hadrag] Lease Thread terminated
00002cc4.00003384::2011/08/05-13:47:42.605 INFO [RES] SQL Server Availability Group <ag>: [hadrag] Free SQL statement
00002cc4.00003384::2011/08/05-13:47:42.902 INFO [RES] SQL Server Availability Group <ag>: [hadrag] Disconnect from SQL Server
|Messages prefixed with [RHS]||RHS.EXE (Resource Hosting Subsystem, host process of hadrres.dll)||00000c40.00000a34::2011/08/10-18:42:29.498 INFO [RHS] Resource ag has come offline. RHS is about to report resource status to RCM.|
|Messages prefixed with [RCM]||Resource Control Monitor (Cluster Service)||000011d0.00000f80::2011/08/05-13:47:42.480 INFO [RCM] rcm::RcmGroup::Move: Bringing group 'ag' offline first...
000011d0.00000f80::2011/08/05-13:47:42.496 INFO [RCM] TransitionToState(ag) Online-->OfflineCallIssued.
|RcmApi/ClusAPI||An API call, which mostly means SQL Server is requesting the action||000011d0.00000f80::2011/08/05-13:47:42.465 INFO [RCM] rcm::RcmApi::MoveGroup: (ag, 2)|
Debug AlwaysOn Resource DLL in Isolation
It is a debugging best practice to configure your cluster to run the AlwaysOn resource DLL (hadrres.dll) in isolation from other resource DLLs. By default, the WSFC cluster runs all resource DLLs in a single instance of rhs.exe. This causes all resources within the cluster to share the same rhs.exe instance. When you attempt to debug hadrres.dll with a debugger, pausing at a break point may cause other resources that share the rhs.exe instance to be paused as well. Also, when you run multiple availability groups in the same cluster, the same configuration can cause all availability groups to be paused when you pause at ta break point to debug one availability group.
To isolate an availability group from the other cluster resource DLLs, including other availability groups, do the following to run hadrres.dll inside a separate rhs.exe process:
Open Registry Editor and navigate to the following key: HKEY_LOCAL_MACHINE\Cluster\Resources. This key contains the keys for all the resources, each with a different GUID.
Find the resource key that contains a Name value that matches your availability group name.
Change the SeparateMonitor value to 1.
Restart the clustered service for your availability group in the WSFC cluster.