NVBU and NVSD are on the same machine.
For testing purposes, Exchange databases and logs are restored to the local disk on the NVBU server using the Exchange Server APM 'Copy files to target directory' feature.
The restore job fails with the below error:
Failed to connect to all known addresses for device 'NVBUSERVER'
Failed to restore log file 'C:\Exchangerestore\NVBU\\Mailbox Database 0136324319\E000002E855.log'
At the same time the System Event logs show event ID 4227 source TCPIP:
http://technet.microsoft.com/en-us/library/cc735929%28WS.10%29.aspx
The Exchange restore process trace shows:
0 NET ??? 2196 433 0 132935310226 Failed to bind to any allowed local port
0 PLGDAV ??? 2196 5 0 132935310226 Failed to connect to all known addresses for device 'NVBUSERVER'
Upon checking the restored items, the database files are intact but 1 or 2 of the log files we managed to restore are 0 bytes in size.
Restoring a copy of the same backup from VTL or tape completes successfully.
When performing resotre of transaction logs for Exchange, the current Exchange Server APM v4.5.x will make as many restore (GET) requests to the same SmartDisk segment, as there ar elogs to restore.
Meaning, a stream is created for each log to be restored. However the plugin requests rate to SmartDisk is so high that the OS connections are not "freed" up in time and are still in the process of being closed.
The request rate being much higher than the system's ability to provisioning binding ports, the system will forcibly close the connection which results in logs being restored with 0 bytes in size.
The 0 byte files restored are part of the restore process. During restores, a file is first created as 0 byte then written to.
However the restore failure happens just after the file is created.
In this case the plugin ran out of of available ports as it was unsuccessful in binding several ports.
Bug 21195 has been created to change the Exchange Server APM restore request behaviour to SmartDisk and this will be fixed in a future version of the plug-in.
In the meantime, it is important to understand that this rare situation only occurs when some of the following conditions are met:
- both NVBU and NVSD are on the same machine
- restore is done locally from NVSD to the same server's local disks.
- and the 'Copy files to targer directory' feature is used.
Restoring from SmartDisk back to the original Exchange Server for production recovery purpose will still work.
The restore transfer rate being extremely high as the connections are local, the key would be to introduce an intended bottleneck to slow down the restore requests.
Possible workarounds could be:
- restore the needed files to a remote location, introducing the network link as limiting factor in speed
- or duplicate the needed backup to another remote device then restore from that location
