1: Scrimping on SharePoint's RAM or Hard Disk Space
A poor, defenseless SharePoint server working as hard as it can to keep users happy, but having its hands tied because of limited resources. This situation is usually a casualty of aggressive virtualization. Virtualization itself isn't bad, but it must be done intelligently and without sacrificing SharePoint's ability to do its job.
If SharePoint finds itself starved for RAM, it starts shutting off functionality so that it can fit into the available space. It also caches less in the web application pools and recycles those pools more often. Less caching and more recycles result in a degraded end-user experience, as SharePoint must compile the same ASP.NET code over and over. And no one likes unhappy users, not even their mothers.
The solution to this particular issue is easy: Add RAM. Microsoft has published the hardware requirements for SharePoint 2010 in the TechNet article "Hardware and software requirements (SharePoint Server 2010)." These requirements state that at the very least, each SharePoint 2010 production server should have 8GB of RAM and a C drive with at least 80GB. In many cases, that won't be enough. If your servers are in production, you can watch their memory utilization to see whether they use the entire 8GB of RAM. If so, they could use more. If your servers are not yet in production, you can use a variety of load-testing tools to simulate your intended load and see how the servers hold up. For example, you can download the Microsoft Load Testing Kit, part of the SharePoint Administration Toolkit.
As for your C drive, SharePoint itself doesn't need much space, but Windows does. After all, your server has several years of Windows patches to look forward to. While you're adding drive space to your machine, consider adding a secondary drive as well. This drive is a great place to store all the files that you use when you install SharePoint. All the third-party installation files can go there too. You can also have SharePoint put its log and Search index files on this drive. This approach takes some pressure off the C drive. Happy C drive and happy end users equal a happy SharePoint server administrator.
2: Using Virtualized Microsoft SQL Server
Virtualization isn't bad. But virtualization allows administrators to make mistakes on a much grander scale. Take virtualizing SQL Server. In the context of SharePoint, this process can be especially painful. The main mistake I see when virtualizing SQL Server is overcommitting the host, be it through RAM, CPU, or drive space. Because everything in SharePoint is stored in SQL Server, if SQL Server is slow, SharePoint is slow.
The obvious solution is to move SQL Server to a physical box, on which it doesn't need to share resources. Moving SharePoint's SQL Server instance is easy, thanks to aliases. This process outlined with pictures, at www.toddklindt.com/sqlalias.
If you can't get a physical SQL Server box, then at the very least ensure that your virtualized SQL Server instance has a fighting chance. First, make sure that its virtual drives aren't thin provisioned. I/O is one of the areas in which virtualized SQL Server struggles the most, and thin-provisioned drives exacerbate that problem. Also try to put the SQL Server guests' virtual drives on their own spindles on the host. Doing so should improve I/O by preventing SQL Server from fighting other guests for time with the drives. Finally, you shouldn't allow the virtualization host to overcommit its RAM. If the host must swap to meet its RAM obligations, then it's slowing down SQL Server.
Brent Ozar has recorded a brilliant video on how best to virtualize SQL. Go get some wine and pizza, invite your fellow SharePoint admins, dim the lights, and watch that video. You'll learn a lot.
3: Using the Farm Configuration Wizard
Using the Farm Configuration Wizard was a pretty common mistake when SharePoint 2010 first came out but thankfully has diminished as our familiarization with SharePoint 2010 has increased. The wizard's list of atrocities is long, so I'll just cover some of the better known ones. First, and maybe most heinous, is that all the databases that the wizard creates have nasty globally unique identifiers (GUIDs) at the end of their names. The wizard also creates a content web app, at http://servername, that just doesn't scale well. To add insult to injury, the wizard creates your My Site host on that same web app, at http://servername/my. Finally, the wizard encourages you to create service applications that you might not actually use. It's tough to resist the siren song of those check boxes, I know.
The Farm Configuration wizard leaves its dirty handprints all over SharePoint, and it can be a challenge to clean up all of them. However, a few places can be easily fixed. Start with your web apps. Create a web app for My Site and give it a Fully Qualified Domain Name (FQDN), such as mysites.company.com. Create a My Site host at the web app's root. Use the Windows PowerShell cmdlet Move-SPSite to move any My Site to one content database, and then attach that content database to your new web app. You'll also need to adjust your User Profile Service and tell it about your new My Site location.
Next, clean up your service applications. Go through your list of service applications and delete any that you aren't using. You gain no benefit from having a service application that you aren't going to use for another six months. After you've deleted unnecessary service applications, stop the associated service instances (also called services on server) that power them. If possible, remove the GUIDs from the service application database names. The technique for completing these tasks varies among service application; the Microsoft article "Rename or Move Service Application Databases (SharePoint Server 2010)" has directions for all the service applications. Of course, take good backups before doing any of this.
4: Using an Incorrect URL when Creating a Content Web App
Like any relationship, SharePoint and Microsoft IIS have communication problems from time to time. Web app creation is one of those times. SharePoint doesn't tell IIS about changes that you might make to a web app after it is created. For instance, if you create an Alternate Access Mapping (AAM) for a web app in Central Administration, you still need to go into IIS and add the host header for the new address.
The issue is compounded when SharePoint farms that you never thought would need to be accessible from the Internet suddenly need to be accessible from the Internet. Budding SharePoint administrators commonly give their web apps short URLs, such as http://portal, to save users some typing. Of course, that URL doesn't route across the Internet, so the web app needs a fully qualified URL added to its stable of AAMs. Not only is this new URL not written to the IIS host headers, but it's also missing from all the alerts, workflows, and anything else that saves URLs -- all those items have the old URL hard-coded in. Because SharePoint didn't write any additional URLs to IIS when they were created, it won't write them to any new SharePoint servers that are added to the farm. Nor will SharePoint write these changes to IIS if the Microsoft SharePoint Foundation Web Application service instance is stopped and started.
This issue might not seem like a big deal, but it has bitten many people at the worst possible time: during an outage. In a few cases, administrators have lost or needed to rebuild a SharePoint server and forgotten about the host headers that they put in manually months earlier. SharePoint is up and going, but when browsing to SharePoint, end users get the blue IIS 7 splash page instead of the SharePoint page that they were expecting. Again, unhappy users usually mean unhappy administrators.
Because SharePoint writes host headers only when a web app is created, you can't fix problematic web apps. You'll need to recreate them. That's good news and bad news. The good news is that you won't lose any of the content that your users worked so hard to create. The bad news is that you will lose all the settings that you worked so hard to create.
The first step is to make notes of all your web app settings. In most cases, there won't be many, and you'll be familiar with any changes that you made. Then, detach the content databases from your web app. Keep them safe; you're going to need them. Next, make a copy of the web.config file for that web application. Some settings, such as forms-based authentication (FBA) and BLOB cache settings, are in that file. Finally, go into Central Administration and delete the web app. Tell SharePoint to delete the extra stuff. The scary part is over.
Now, recreate the web app, but do it right this time. First, enter the correct, fully qualified URL in the Host Header box. Do your end users a favor, and put the web app on port 80, as Figure 1 shows. Under the Security Configuration settings, accept all the defaults, even if you're going to use Kerberos or SSL. You can change those settings later, and you want to make sure that the web app works correctly before you apply fancy security settings. Doing so helps in any troubleshooting that you might need to do. Under the Application Pool settings, pick an existing application pool.
It is important to give your content databases distinct names. You should be able to look at a content database name and know exactly which web app that database goes with. This is another one of those things that doesn't usually seem important but is priceless in a disaster-recovery situation. If the content databases that you detached from the web app before you deleted it didn't have such names, then take this opportunity to right that wrong when you recreate the web app. Give the new content database a good name, then use the PowerShell cmdlet Move-SPSite to move the site collections to that new database. If your content database already has a good name, enter it during the creation of the new web app. If you had multiple content databases, attach the rest after the web app is created.
After the web app is created, you can tweak settings as needed. Most settings can be changed in Central Administration. If you made any changes to the web.config file of the original web app, now is the time to copy those changes to the newly created web.config file. You can use a program such as Notepad++ to compare the two files. You should now have a well-created web application that you can trust in times of crisis.
5: Running Web Apps or Service Apps in Separate App Pools
Web apps and service applications run inside of an application pool, which is a W3WP.exe process that runs on your server. Unless you have reason to do otherwise, you should run all SharePoint web apps inside one application pool; the same goes for the service applications. Running each web app in its own application pool makes inefficient use of the server's memory. Each application pool has a minimum overhead of more than 100MB, and its memory footprint increases as it caches content that's rendered frequently. Figure 2 shows multiple W3WP.exe processes running as sp_webapps, the result of web apps running in separate application pools. We've all experienced SharePoint slowing first thing in the morning because the app pools recycle overnight and need to warm up and cache that content again. Well, multiple application pools mean that the same content is cached multiple times. Most users are impatient. I'm sure that studies would show that they spend the time waiting for SharePoint to respond by thinking of ways to punish us for SharePoint's poor performance.
For service applications, this problem is easy to fix. First, make sure that you have a good service application pool to use. I recommend calling this pool Default SharePoint Service App Pool so that it floats to the top of all your drop-down lists. Use a dedicated sp_serviceapps account for the pool's identity. Most service applications allow you to assign them to a new service application pool by modifying their properties in Central Administration. If the option is unavailable there, look for it in PowerShell.
Web applications are a tougher matter. There's no easy, out-of-the-box way to change the application pool that a web app is using. Fortunately, we have PowerShell at our disposal.
6: Using One Account for Everything
Security is complicated, and SharePoint doesn't necessarily make it any easier. Using just one account -- maybe even the coveted Domain Administrator account -- is so easy. We've all done it, even though it's a bad idea. When you use an existing account, you open up SharePoint to several security issues. Anyone who knows the account password can do anything in SharePoint, so you can't separate duties. You also lose the ability to audit who made which changes. And if that common account password is compromised or needs to be changed, you jeopardize SharePoint's uptime as well. Even if you use one dedicated account for SharePoint, you leave yourself vulnerable to attack. If that account is compromised via a security exploit, the bad guys will have access to everything in SharePoint.
To fix this mistake, start by creating the accounts. Add the sp_webapps and sp_serviceapps accounts as managed accounts. Use the techniques that describe in Mistake No 5 to fix your web app and service application accounts. You can change the default content access account for the Search service application at the Search Service Application page. Under Central Administration, Security, Configure Service Accounts, you can change the accounts that other processes use as well. (You can even change the Farm Account there. I've done so in test environments but haven't been brave enough to do it in production.) If you're using the User Profile Service, make sure that your new sp_userprofile account has the correct permissions in Active Directory (AD), and recreate your AD connection in the User Profile Service.
7: Keeping Default SharePoint Database Settings
When SharePoint creates its multitudes of databases, it makes some bad assumptions. Take the autogrow settings: The database files grow by 1MB at a chunk, almost ensuring that they're going to autogrow with every upload. Not only does this slow down SQL Server (which slows down SharePoint), but it also results in database files that are spread all over your drives in itty-bitty 1MB chunks.
SharePoint also creates most of its databases, notably the Config and Content databases, with the recovery model set to Full. Although this is great if you want to recover data, you must manage the process correctly or those sneaky .ldf files will slowly, methodically fill your hard disk. If you think users get upset when SharePoint is slow because of fragmented databases, you should see how angry they get when SharePoint stops completely because the SQL Server drives are full.
To fix this mistake, set your databases' autogrow settings in such a way that they don't need to grow frequently. For most farms, I recommend changing the 1MB autogrow to something like 500MB or 1GB. Autogrow should also be a last resort. Someone, either the SharePoint administrator or a dedicated DBA, should pregrow your databases so that autogrow is unnecessary.
Your recovery model setting needs to be consistent with your disaster recovery plans. If you need your transaction logs, make sure you're performing routine log backups to keep those .ldf files in check. If you don't need your transaction logs, then consider switching your databases to the simple recovery model. Doing so will keep your .ldf files from swelling up like a nasty bee sting.
8: Not Enabling BLOB Caching
We all want SharePoint to get files to the users as quickly as possible. However, more often than not, I see SharePoint farms without BLOB caching enabled. BLOB caching is one of the easiest and least expensive ways to improve SharePoint performance. Not only does it help to get files to users more quickly, but it also eases the burden on SQL Server. Everybody wins.
This might be the easiest solution so far: Enable BLOB caching, of course! BLOB caching is actually a function of IIS; SharePoint just takes advantage of it. Therefore, to enable BLOB caching requires a change to each web app's web.config file on each server. Fortunately, the setting already exists and just needs to be enabled. By default, the web.config files are in a directory under C:\inetpub\wwwroot\wss\virtualdirectories. Each web app has a directory and a web.config file. Open one of these files and look for the following line:
<BlobCache location="C:\blobcache\14" path="\.(gif|jpg|jpeg|jpe|jfif|
bmp|dib|tif|tiff|ico|png|wdp|hdp|css|js|asf|avi|flv|m4v|mov|mp3|mp4|
mpeg|mpg|rm|rmvb|wma|wmv)$" maxSize="10" enabled="false" />
To enable BLOB caching, replace "false" with "true" and save the web.config file. You should also move the file to a directory on a drive other than the C drive. The maxSize parameter is measured in gigabytes, with a default of 10GB. If the space is available, you might want to increase this size.
If editing this file in Notepad on all your servers isn't your idea of fun, you can use PowerShell to automate the process. You still need to perform the process on each server, but using PowerShell is quicker and reduces the chances of a mistake.
9: Not Installing a PDF iFilter
Most organizations have a tremendous number of PDF files in their SharePoint farms, and those files represent a wealth of information. End users want to be able to discover that information by using SharePoint Search. Getting users excited about SharePoint Search is a great way to get them excited about SharePoint in general.
Installing a PDF iFilter is fairly easy. Adobe has a free PDF iFilter that you can install. You can find the download link and detailed installation instructions in the Microsoft article "SharePoint 2010 - Configuring Adobe PDF iFilter 9 for 64-bit platforms." You need to install the iFilter only on those SharePoint servers that run the Search Index role, although installing it on the rest of your SharePoint servers doesn't hurt. If you have a large farm and want to reduce the time needed to index your PDF files, you can use the PDF iFilter from Foxit. This product has better performance than the Adobe iFilter but isn't free.
10: Not Pointing Your SharePoint Servers at Themselves
When SharePoint works, it is magnificent. When it doesn't work, it can be a nightmare to fix. For this reason, anything you can do to ease troubleshooting is time well spent. To that end, I make sure that every server in the SharePoint farm points to itself for all web apps. If I get sporadic reports about SharePoint not responding, I can easily use RDP to log in to each server and try to pull up SharePoint. If this attempt works, then I know that the server is working. If SharePoint does not come up, then I know in exactly which Microsoft User Location Server (ULS) logs to look for the relevant errors. No worrying about which web front end the load balancer sent my request to. The quicker you get to the correct log files, the quicker the problem is resolved.
Pointing your Search indexer at itself has another advantage: It improves performance for your end users. If you don't point your Search server at itself, then when it starts to perform a crawl, it lets DNS do its work and then starts crawling whichever web front end DNS points it to. That server is most likely the same one that is sending pages to your end users. Making the server do double-duty means that everyone waits longer. Pointing the Search server at itself means that your web front end doesn't need to handle that traffic and can get back to doing its #1 job: keeping users happy.
There is a simple fix for this mistake: Open the hosts file (C:\windows\system32\drivers\etc\hosts) on each SharePoint box, and add all the URLs that SharePoint knows about. Point those URLs to 127.0.0.1, which translates to "this computer." Figure 3 shows how this file looks in a typical SharePoint environment.
Posts
