Thanks for attending the Webcast. We had a great crowd with lots of great questions.
Here’s the Deck: SharePoint Backup and Restore Strategies.
The slides are there and we’ll have the recording posted there as soon as it’s available (later today.) Those who have already registered for your webcast should be able to just click “Register Now” to download the file. (I know it sounds confusing.)
Thanks Brent Ozar (the SQL Guru accidental SharePoint Admin) and Ghazwan Khairi (demo)
Key resources:
Backup and Restore Strategies Q & A
Question: Will a backup snapshot of each server on the farm on a SAN be a good way to recover from disaster?
Answer: Not really. The snapshots won't necessarily be consistent, and it only gives you the ability to recover an entire server - not pick individual documents out of the database, like we're showing onscreen now.
Question: What is RPO again?
Answer: Recovery Point Objective… at what point in time can you recover to. Do you need to restore to up to the minute of failure, or is 24 hours from your last full backup ok… (Note the data loss… in that scenario. What’s your tolerance for that?)
Question: What databases would you put in Full recovery - just the content databases?
Answer: It depends on the requirements. For point in time, you would do it on all of them. The key is the configuration database needs to be the newest. Content databases can detach and reattach (SharePoint terms) to the config db to work out their site configuration details.
Question: Are we going to cover in this session a scenario where a SP DB is log-shipped to a DR server and there is a need to DR resumption?
Answer: Yes, we are. <We covered that. See the slides and recording.>
Question: Where is the recycle bin data stored? In the SQL databases?
Answer: Yes, you got it. It is simply a flipped bit.
Question: Is the index you are talking about the search index? If so we are using Search Server 2008 so do we want to make sure it is backed up well?
Answer: Correct, absolutely. You can use the SharePoint backup to backup that index, but also backup all those databases as well. The index configuration details are stored in a few different places, so backup the box as well. All the other rules apply here as well.
Question: Are we going to see a product demo? I was under the impression thats what this session was for.
Answer: Yes, absolutely - that's the last 10 minutes of the session.
Question: What is DPM?
Private Answer: Microsoft Data Protection Manager more at http://www.microsoft.com/dpm
Question: Why do you need to back up indexes? Could you just rebuild them?
Answer: If Search/Index isn’t critical that is the easiest answer. Especially if you have only the default content sources and don’t do anything with it. Those that do care and feeding and the search is really part of their data, it’s an important focus.
Question: SQL Cluster does not work in a VMWare Environment !!!
Answer: Brent: Correct, it does not. Future versions of VMware will support that. Joel: SQL is one of those things that is usually better on physical hardware due to high disk IO performance requirements, even if you’re not doing clustering. In the past my old teams have done Virtual Server SQL clustering and it’s great for a test or validation and doesn’t even require special disks :)
Question: Is using replication a good method of sychronizing databases in Sharepoint?
Answer: SQL Replication is not supported, but some vendors like Infonic and Syntergy have SharePoint “replication” like solutions. Some vendors do this at the byte level. Talk to your product support contacts about any solutions you delve into that might question this.
Question: We have implemented log shipping in that particular scenario for a DR SQL server on standy. it ships every 15 minutes and worked great when we failed over in a test scenario
Answer: OK, great. Do you have a question? Just making sure we don't miss it.
Question: no, sorry was just making a comment about log shipping. but one question i do have is why doesn't SQL clustering work in VMWare environment?
Answer: Brent: Because of networking issues and shared drive issues. The issue isn't a problem with SQL Server, but due to the limited hardware support in VMware.
Question: Why is replication not a good way to sychronize Dbs? Is it a good way to recovery a backup?
Answer: It’s a design decision. The types of data and the data structures don’t lend themselves to supportability. Big answer is that's not supported by Microsoft.
Question: Can VMware clone or snapshots be considered a good backup solution ???
Answer: Snapshots for backup, yes. Clones for recovery, yes. Make sure you are tolerant of your original diff area. With external vendors they all have their pros and cons, and I’d recommend serious conversations about high performance.
Question: Recovering a deleted site is my biggest concern at this time. What is the easiest way to do this?
Answer: Easiest way to restore a site is with Quest Recovery Manager. If you don’t have that, the easiest way is to have backed it up with STSADM –o Backup –URL http://sitecollection/name beforehand, but that doesn’t scale and can cause blocking. Having a recovery farm to mount your recovered database and running stsadm – o backup to backup the site collection and doing the –restore on the production farm. If it’s a site and not a site collection, you’d use stsadm –o export on the recovery farm, and –import on the production farm. Pay attention to the options of that command stsadm –help export for more info on what to export such as security. (Note: The export is NOT full fidelity and won’t include things like alerts and workflows)
Question: Do you have experience using SnapManager for Sharepoint (Netapp SAN add-on tool)?
Answer: Yes, but it's probably not appropriate for this webcast. Your best bet is to talk to your NetApp sales and technical contacts. There are pros and cons with any solution.
Question: Compared to other database applications it appears that the transaction log file is very large even though I am doing a full nightly backup and hourly transaction logs. Is this a SharePoint issue? Do I need to change from Full to Simple once in a while to shrink the transaction log file?
Answer: Brent: You want to avoid shrinking the transaction log whenever possible. Here's an article I wrote about that: http://sqlserverpedia.com/blog/sql-server-backup-and-restore/backup-log-with-truncate_only-like-a-bear-trap/
Joel: There is also a whitepaper on SQL database maintenance on Technet I’d recommend.
Question: one great suggestion for DR is vritualizing the entire farm using Hyper-V and failover clustering
Answer: For some high availability requirements this may work with high bandwidth and low latency, but if it's a large solution if you're talking about failing over from your primary datacenter to the disaster recovery datacenter, like from New York City to Chicago. There are a lot of complexities involved with that.
Question: Does Microsoft provide support licenses (free or low cost) to support testing disaster recovery on new farm?
Answer: With SQL Server, you only have to pay for instances that are "live" - you don't pay for disaster recovery instances that are not queried. Joel: With the dev, test, environments you can use MSDN licenses, but if it’s used in production you may have to pay, so it’s best to talk to your MS rep. It isn’t as clear as it is with SQL since there are so many ways of doing DR.
Question: So this is searching, or can search, a SQL database backup directly to allow for Item Level Recovery?
Answer: Yes with Recovery manager it builds an index for quick recovery of items, lists, sites and site collections.
Question: So the quest tool is backing up from the SQL server backups not from it's own backups?
Private Answer: Recovery Manager does not do any backups. It pulls the data and related info from the backups you're already doing. In this case, he was restoring from SQL. Joel: If you were using SharePoint Backup, SQL Backup or Quest Litespeed backup the demo would have looked the same.
Question: So the Quest product doesn't help with backups...just helps with restoring?
Private Answer: This recovery tool is focused on restores, but provides the flexibility of plugging in your existing backup infrastructure for a ton of various backup vendors including the Quest litespeed Database backup tool.
Question: So this tool is *only* for recovery, or does it do backup as well?
Private Answer: Recovery Manager is a recovery tool. It does not do backups.
Question: Would restoring ie. doc library or site colelction keep the metadata?
Private Answer: yes, Recovery Manager does restore the data with related metadata
Question: Can you name some additional benefit Quest software gives over backup exec? Or is there a white paper you might have on that?
Answer: The recovery interface is the key here. You could use Backup Exec to backup your infrastructure and databases and then use this tool for its simple design to recover Site Collections, Sites, lists, and items with full fidelity. With Backup Exec I believe you’d still need the recovery farm, but you should check with them.
Question: Does Quest provide a backup tool in it's suite as well? (thanks for the lolcat!)
Answer: Yes, Quest has a tool called LiteSpeed that performs SQL backups (with compression and FAST!)
Question: When you do a site restore - What happens if the data already exists does it write over?
Answer: Yes, the warning message we saw when he clicked restore stated that overwrites may occur, but you could restore to an alternate location by clicking “restore to” button.
Question: One question I still have about DR is upon a failover to a DR farm. after attaching the databases (wether using SAN replication or DB mirroring) and the web apps, would MOSS always trigger a full crawl? if so, while crawling, would I still be able to search the old (invalidated?!) search index?
Answer: If the GUIDS change then it would require a full crawl. If you stop/pause your indexing schedules before the index notices the changed GUIDS you might be ok. You should test this. If the Query Servers think they have old data (what’s on the index vs. what’s on the Query) they will choke and spew errors until after a crawl is performed.
Question: Using STSADM, is it possible to get site level back up. Or is there a workaround to achieve that? We dont want to restore the whole site collection just for saving a site.
Answer: You can use STSADM Import/export from a recovery farm, but note this isn’t full fidelity.
Question: Does STSADM export/import logical objects (lists/sites/...)?
Answer: Yes, with command line switches for sites yes, but not lists. Importing and exporting lists can be performed with the object model such as content migration API, or with STSADM extensions (Gary Lapoint has some good ones), or with MS Access (some level of fidelity) or Excel (not full fidelity). You will loose created by and created date for most lists when using the client tools. The Quest Recovery Manager does address most of these scenarios.
Question: What are the "pros" of this tool over a tape backup solution such as COMVAULT that can restore at the document level
Answer: This tool (Recovery Manager) is a simple interface with a powerful search/index over all of the sites and site collections including search over the production recycle bins to reduce the time hunting and pecking for the data that needs to be restored across multiple content databases. It can be used in tandem with Comvault.
Multi Part Question: I'm a newbie to SP Administration and have a lot of basic questions. One is, on a failover farm, you would not be replicating the configuration database, correct ?
A: Correct.
So, the configuration would have to be done manually prior to attaching the databases ?
A: You would have a different configuration database on the other farm and yes after you stop log shipping you would reconfigure the farm and then attach the content databases.
Can the re-configuration of the farm be easily scripted ?
A: Yes, almost all of this can be scripted. DNS and the cutover/Stop log shipping is what you want control over.
Also what other databases should not be replicated ?
A: Your SSP stuff including search will require a different method of recovery other than log shipping. Likely a backup restore, but with the new Config DB you’ll have new GUIDS. Simple answer is reindex. Complex answer is 1 or 2 SSP Farms that get paused before the new Config db is created. There are whitepapers that describe some scenarios. See the Mirroring whitepaper.
Question: Can you not replicate (log shipping, mirroring, etc.) all the databases to your DR site? Why only the content db?
Answer: First it’s not supported. Second it’s due to the configuration databases having the server names as well the configuration database keeps track of Webapp settings and a lot of detail about those particular servers. You won't have the same servers in DR.
Question: Follow up to my previous question - Does Sharepoint provides free licenses to test disaster recovery?
Answer: For testing yes. I believe it’s 180 days for a trial. MSDN license vs. production is where you need to talk to your rep.