how to upgrade either an existing instance of HPC to a remote database server, or setup a new head node with a remote database server? RRS feed

  • Question

  • Hi,

    We have spun up a new cluster with more nodes than we have in the past and appear to be running into the 10gb limit of sql server express.   I am curious on what the process would be in order to do 2 different things.  We are currently on 2012R2Update3:

    1) move the existing local sql server express database to a remote instance of sql server enterprise

    2) how to create a new head node, and have it create the databases on a remote database server.

    I found this guide (https://docs.microsoft.com/en-us/powershell/high-performance-computing/deploy-an-hpc-pack-cluster-with-remote-databases-step-by-step-guide?view=hpc16-ps) but had a few questions.

    1) is it possible to move an existing local installation of the databases to a remote instance of sql server enterprise?   Is it as simple as backing up the database, restoring it on the new server, and then changing the config somewhere to tell it to look at the new location instead of the local instance?   

    2) what user does HPC try and access the database instance as?  I see a bunch of NT Authority connections on the local database when i sp_who2, and dont know how that would translate in a remote instance.

    3) If I'm setting a new head node, what level of access does the user we are installing this as need?  Does it need SA on the database server (likely not possible given these are shared database servers managed by our DB services team).   Is it possible to create the empty databases, and then pointing the installer to those with an account that is db_owner on each of the databases and have it create the schema?   

    4) is there anything else we should know before embarking on the remote database server path?

    Monday, December 2, 2019 9:51 PM

All replies

  • Hi,

    Also, related to this, do you have any suggestions on how I can manually clean up the database to free up some space given that we've hit the 10gb limit in the meantime?  We mostly don't care about any of the log entries that exist on this system currently.   Are there tables I can safely delete from/truncate to free up space?  I couldnt tell from the schema what exactly was actually used to store configuration changes, and what was just logs that could be safely deleted as long as we didnt need log entries.

    Our 2 biggest tables are LogEntries, and SettingValues.    Is dbo.LogEntries safe to truncate?  what about dbo.Changes?  Are there any values we can delete out of SettingValues?  I just want to get the database in a manageable state while we sort out how we're going to upgrade to SQL Server Standard re: licensing.   

    Tuesday, December 3, 2019 9:26 PM
  • Hi Jason,

    1. It is dangerous to truncate database manually! If you want to free up some spaces, you can reduce the number of days that archived operation log is retained in the HPCManagement database (by default 180 days), you can use the following PowerShell command to check the current number and set a new number (for example 30 days).

    Add-PsSnapin Microsoft.Hpc


    Set-HpcClusterProperty -OperationRetention 30

    2. Yes, you can move your local HPC databases to remote SQL server, following is a link for your reference: https://social.msdn.microsoft.com/Forums/sqlserver/en-US/ee8273a4-6452-4c8c-953b-8f65d41b8c0a/move-all-databases-to-new-server?forum=sqlservermigration. We are not SQL experts, so if you need assistance about SQL migration, you may need to contact SQL server support.

    3. DB owner permission is good enough, you shall grant the permission to the user who will run the HPC setup, and the head node machine account (NT Authority\System translated into the machine account when connecting to a remote instance).  

    Friday, December 6, 2019 9:52 AM