locked
Backups vs user access to farm RRS feed

  • Question

  • MOSS SP 2007 service pack 3, 32 bit OS and application. Content on SQL Server 2005 32 bit separate farm, but is a shared resource.

    We are encountering increasing number of occurances, particularly during 7pm-6am time period, when users are unable to get pages to display from our farm.

    Currently there are two kinds of backups that occur.  The database admins perform backups of the databases nightly.  We run batch jobs that perform stsadm -o backup of the 1600 site collections daily.

    The stsadm backups are not intended for use in recovery the farm. They are only intended to provide quick access to the contents of a site collection for recoverying a site/list/document that has been lost from the production farm.

    We schedule the stsadm bat jobs so that the mysites are backed up one at a time, with no other backups occuring. Then, after they complete, the production data is backed up in 2 parallel threads. In one, the largest site collection (approx. 100 GB) is backed up. This takes 6-8 hours.

    In the parallel thread, 12+ site collections, mostly small, largest 20 GB or less, are backed up one at a time.

    These commands are executing in a scheduled task on the farm server.

    What we are seeing is not a problem with CPU usage. Users just cannot get the farm to deliver pages.

    One of the alternatives that has been suggested is that we stop doing the stsadm commands and in the rare cases where we need to do a recovery, restore the database backup file to a different sql server, attach the database to a development sharepoint farm, and access the data in that manner.

    From what I have seen in http://technet.microsoft.com/en-US/library/cc512725(v=office.12).aspx , the process would be something that the database admins would be familar with doing.

    The only step that the microsoft article doesn't seem to mention is connecting the restored database to the farm. But I know how to add a content database to the farm.

    Are there any special issues that need to be taken into consideration?

    Wednesday, December 5, 2012 2:21 PM

Answers

  • Hi,

    No as long as you know how to add a ContentDb to the DEV server to grab the data you need its not an issue. However, whenever you are run multiplexed backups on SQL the performance degradation of the disk IO will cause you to lose the connection usually to the configurationDB. It has been my experience you should be able to under normal circumstances using a single threaded backup be able to complete 300-400gb in 6-7 hours. You can reduce the number of threads that are being used to backup up the DBs, perform differential backups during the day, with a full at night to reduce the amount of data being flushed from he DB Logs LDFs into the MDFs.. But usually the si8mplest approach works best. Personally imho I never run stsadm backups and depend on the recycle bin and SNAP Manager... I can restore terabytes in 15 minutes with a SAN that's been provisioned properly and using SNAP Manager...

    When Office SharePoint Server 2007 is installed or a new Web application is created, the following default Recycle Bin settings are configured:

    • The Recycle Bin is turned on.
    • Items are scheduled to be deleted from the Recycle Bin after 30 days.
    • The second-stage Recycle Bin is set to add 50 percent of live site quota for second-stage deleted items.

    If you are satisfied with these settings, no further action is required. If you want to change these settings, use the following procedure. Regardless as necessary attaching a ContentDB to a DEV WebApp is painless as long as the same ContentDB is not attached to the same farm...

    -Ivan


    Ivan Sanders My LinkedIn Profile, My Blog, @iasanders.

    Sunday, December 23, 2012 12:22 PM

All replies

  • Hi,

    No as long as you know how to add a ContentDb to the DEV server to grab the data you need its not an issue. However, whenever you are run multiplexed backups on SQL the performance degradation of the disk IO will cause you to lose the connection usually to the configurationDB. It has been my experience you should be able to under normal circumstances using a single threaded backup be able to complete 300-400gb in 6-7 hours. You can reduce the number of threads that are being used to backup up the DBs, perform differential backups during the day, with a full at night to reduce the amount of data being flushed from he DB Logs LDFs into the MDFs.. But usually the si8mplest approach works best. Personally imho I never run stsadm backups and depend on the recycle bin and SNAP Manager... I can restore terabytes in 15 minutes with a SAN that's been provisioned properly and using SNAP Manager...

    When Office SharePoint Server 2007 is installed or a new Web application is created, the following default Recycle Bin settings are configured:

    • The Recycle Bin is turned on.
    • Items are scheduled to be deleted from the Recycle Bin after 30 days.
    • The second-stage Recycle Bin is set to add 50 percent of live site quota for second-stage deleted items.

    If you are satisfied with these settings, no further action is required. If you want to change these settings, use the following procedure. Regardless as necessary attaching a ContentDB to a DEV WebApp is painless as long as the same ContentDB is not attached to the same farm...

    -Ivan


    Ivan Sanders My LinkedIn Profile, My Blog, @iasanders.

    Sunday, December 23, 2012 12:22 PM
  • Regardless as necessary attaching a ContentDB to a DEV WebApp is painless as long as the same ContentDB is not attached to the same farm...

    So, when you say "the same ContentDB", I presume you mean literally not to attach a backup of content_1 to the same machine where it resided. Is that a correct assumption?

    If my prod farm has content_1 and content_2, but the dev farm has a different set of databases, but also called content_1 and content_2, and I need to to access the sharepoint sites on the backup of the prod's content_1, is that still going to be "painless"?

    There are some site collections which are named the same. So I presume I would have to load the backup database into the dev instance - is there a way to give it a new content database name during that process?

    Then, I would have to detach (remove) the dev content databases and add the new database.

    Is there a tech note or tutorial etc that covers the various other steps that should be covered for this to work successfully?

    Thank you for your time.

    Monday, January 14, 2013 4:23 PM