Database indexes missing on foreign keys

Description

UPDATE: I’ve added notes about specific indexes that need to be created below

SQL server does not enforce that indexes are created on FK columns but in many (most) cases they should have indexes since quite often is the case that these columns are queried against.

One important one that is missing is:

CREATE NONCLUSTERED INDEX TEST_IX_umbracoContentVersion_nodeId ON dbo.umbracoContentVersion (nodeId)

We query against this column a lot!

So we definitely need that Index created but there’s probably lots of others…

I found an SQL script that lists all FK columns that are missing indexes from this site: https://www.mssqltips.com/sqlservertip/5004/script-to-identify-all-nonindexed-foreign-keys-in-a-sql-server-database/

For reference, here’s an article describing the benefits of indexing FK columns: https://sqlperformance.com/2012/11/t-sql-queries/benefits-indexing-foreign-keys

Here are the results of that script:

Table_Name	Column_Name
cmsContentType2ContentType	childContentTypeId
cmsContentTypeAllowedContentType	AllowedId
cmsDictionary	parent
cmsDocumentType	templateNodeId
cmsLanguageText	languageId
cmsLanguageText	UniqueId
cmsMember2MemberGroup	MemberGroup
cmsMemberType	NodeId
cmsPropertyType	contentTypeId
cmsPropertyType	dataTypeId
cmsPropertyType	propertyTypeGroupId
cmsPropertyTypeGroup	contenttypeNodeId
cmsTagRelationship	propertyTypeId
cmsTagRelationship	tagId
cmsTask	nodeId
cmsTask	parentUserId
cmsTask	taskTypeId
cmsTask	userId
umbracoAccess	loginNodeId
umbracoAccess	noAccessNodeId
umbracoAccessRule	accessId
umbracoContent	contentTypeId
umbracoContentSchedule	languageId
umbracoContentSchedule	nodeId
umbracoContentVersion	nodeId
umbracoContentVersion	userId
umbracoContentVersionCultureVariation	availableUserId
umbracoDocumentVersion	templateId
umbracoDomain	domainRootStructureID
umbracoLog	userId
umbracoNode	nodeUser
umbracoRedirectUrl	contentKey
umbracoRelation	childId
umbracoRelation	relType
umbracoUser2NodeNotify	nodeId
umbracoUser2UserGroup	userGroupId
umbracoUserGroup	startContentId
umbracoUserGroup	startMediaId
umbracoUserLogin	userId
umbracoUserStartNode	startNode
umbracoUserStartNode	userId

What to fix

We definitely need the umbracoContentVersion.nodeId indexed but we should review all of the above and determine if adding indexes to any of these will be beneficial. We only really want to add indexes if they are used in queries in our codebase. Adding too many indexes will slow down data insertion which we don’t want.

In most cases you can find the “DTO” object for a table such as: Umbraco.Core.Persistence.Dtos.ContentVersionDto and then you can righ click on it’s column that you want to search for to see if it’s used in queries. For example, for the ContentVersionDto.NodeId, i right click this member and say find references in VS, then i can see how many queries it’s used in

How to fix

For indexes that should exist, the DTO classes will need to be updated to declare the index and then a migration will need to be written to add all of the new indexes.

---

This item has been added to our backlog AB#4131

[liamlaverty]

Hey, I took a look through this last night and found the following uses on those properties. Anything that I could see definitely did a JOIN, SELECT, WHERE, ORDERBY I’ve listed as needs index. There’s a few things I couldn’t find (CMSTask for example), and then I’ve listed anything that’s just used for inserting as “no” (I assume they don’t get indexed). There’s also a handful that I didn’t know enough about the codebase to work out, where the property is handed off to another method which may or may not do querying somewhere else.

Hopefully it’s a useful start though. I found all of the dtos using the class here

If these make sense, I could maybe look at updating the DTOs to include the relevant attributes. Can you give an example of a DTO that already has an index attribute, and then I can follow that style.

Table_Name	Column_Name	Num Uses	Needs index?	Notes
cmsContentType2ContentType	childContentTypeId	2	maybe?	ContentTypeCommonRepository.cs orders a query by this param
				ContentTypeRepositoryBase.cs uses this in 2 inserts
cmsContentTypeAllowedContentType	AllowedId	5	no	only used in inserts
cmsDictionary	parent	4	yes	DictionaryRepository.cs uses this in Where SqlSyntax.GetQuotedColumnName("Parent")
cmsDocumentType	templateNodeId	2	no	only used in inserts
cmsLanguageText	languageId	5	yes	DictionaryRepository.cs uses this in ConvertFromDto, Where LanguageID > 0
cmsLanguageText	UniqueId	4	yes	DictionaryRepository.cs uses this on the right hand side of a join
cmsMember2MemberGroup	MemberGroup	7	yes	MemberGroupRepository.cs uses this as the On part of a join. MemberRepository.cs uses this in two Where statements to find all members in a role. MemberRepository.cs uses this in two Where statements to GetByMemberGroup
cmsMemberType	NodeId	1	no	insert only
cmsPropertyType	contentTypeId	25	yes	as part of joins all over the place
cmsPropertyType	dataTypeId	37	yes	as part of joins all over the place
cmsPropertyType	propertyTypeGroupId	23	yes	as part of joins, orderbys, wheres, all over the place
cmsPropertyTypeGroup	contenttypeNodeId	8	yes	ContentTypeCommonRepository.cs and MemberTypeRepository.cs use this for joins
cmsTagRelationship	propertyTypeId	10	yes	TagRepository.cs and ContentTypeRepository.cs use this for joining
cmsTagRelationship	tagId	9	yes	TagRepository.cs and ContentTypeRepositoryBase.cs use this for joining
cmsTask	nodeId	can’t find in codebase
cmsTask	parentUserId	can’t find in codebase
cmsTask	taskTypeId	can’t find in codebase
cmsTask	userId	can’t find in codebase
umbracoAccess	loginNodeId	3	unsure	insert only
umbracoAccess	noAccessNodeId	3	unsure	insert only
umbracoAccessRule	accessId		yes	DocumentRepository.cs uses for joins
umbracoContent	contentTypeId	34	yes	as part of joins, orderbys, wheres, all over the place
umbracoContentSchedule	languageId	2	unsure	used in a methd “GetIsoCodeById”
umbracoContentSchedule	nodeId	7	yes	Used in Wheres in DocumentRepository
umbracoContentVersion	nodeId	27	yes	Used in lots of wheres all over the place
umbracoContentVersion	userId	7	yes	used in a join in DocumentRepository.cs
umbracoContentVersionCultureVariation	availableUserId	1	unsure	inserts only (column name also has a “rename this” comment on it)
umbracoDocumentVersion	templateId	8	unsure
umbracoDomain	domainRootStructureID	3	no	just used for creating new objects
umbracoLog	userId	5	yes	used in a join for AuditRepository.GetBaseQuery
umbracoNode	nodeUser	67	yes	used in a join on ContentRepositoryBase.ApplySystemOrdering, selects in EntityRepository
umbracoRedirectUrl	contentKey	5	yes	used in a where RedirectUrlRepository.Get
umbracoRelation	childId	3	unsure	appears to only be inserts
umbracoRelation	relType	6	maybe?	used in by RelationRepository.PerformGetAll in an orderby and PerformGetByQuery in an orderby
umbracoUser2NodeNotify	nodeId	6	yes	used in joins and wheres in NotificationsRepository
umbracoUser2UserGroup	userGroupId	8	yes	used in joins in UserGroupRepository and UserRepository
umbracoUserGroup	startContentId	10	yes	UserGroupRepository uses it in a GroupBy statemetn
umbracoUserGroup	startMediaId	11	yes	UserGroupRepository uses it in a GroupBy statemetn
umbracoUserLogin	userId	2	yes	used in a where on UserRepository.ClearLoginSessions
umbracoUserStartNode	startNode	6	yes	used in selects in UserFactory, UserRepository
umbracoUserStartNode	userId	6	yes	used in WhereIns in UserRepository

[Shazwazza]

Hey @liamlaverty, i think it would be easiest for now to just make the migration simple, don’t worry for now about trying to be fancy and making a single migration to do all of the indexes. Might be easiest to just start with the basics, make one migration for this one index. Don’t worry about reflection, etc…

Maybe later we can combine this into a more fancy approach to multiple indexes at once. In which case we don’t need to use reflection like is being done there, you can get a tables metadata by doing this:

var tableInfo = Context.Database.PocoDataFactory.ForType(typeof(ContentNuDto));

which will give you all of the table data you need.