Update set collision resolution

A collision is an update that has a newer local update.

The platform detects collisions by comparing the values in the Name and Updated fields of the customer update record from each update set. If the name matches but there are different update date values, then there’s a collision.

When a customer update is moved from one instance to another, it may be rewritten to match the target instance. The rewrite can involve changing the update name of the customer update and one or more sys_id within the update. The rewrites are done when the record or the reference field is for a table that uses a coalesce strategy. This verifies that the customer update will be applied to the correct record. For example, if the sys_dictionary record for tablename.fieldname has sys_id 123 on instance A and sys_id 456 on instance B, when a customer update that refers to that record is retrieved from instance A and recorded in the sys_update_xml table on instance B, references to 123 are updated to read 456.

Coalesce strategies

Coalescing refers to the process of identifying a record based on one or more fields that are used as unique identifiers during the update process. Update sets can detect collisions between identical records that you independently create on separate instances. To detect such collisions, the record must have a coalesce strategy based on coalescing columns. Because collision detection depends on the uniqueness of tables, the tables must be unique when the coalescing columns are combined. Records that aren’t listed here won’t collide if the same record is created separately on different instances.

Type	Coalescing Columns
atf_input_variable	name, element
atf_output_variable	name, element
dp_data_pattern	source_sys_id
dynamic_attribute	namespaced_name
dynamic_category	namespaced_name
dynamic_category_member	category, attribute
dynamic_choice_override	choice, category, attribute
dynamic_namespace	name
sys_analytics_bucket	sys_scope, bucket_document_id, bucket_table_name
sys_attachment	(uses custom matching logic)
sys_choice_set	name, element
sys_collection	collection, name, join_field
sys_db_object	name
sys_df_data_dictionary	name, element
sys_dictionary	name, element
sys_documentation	name, element, language
sys_index	logical_table_name, col_name_string
sys_module	path
sys_notification_category	name
sys_package	source
sys_package_dependency_m2m	dependency, sys_package
sys_properties	name
sys_report_chart_color	name, element, value
sys_scope_script_access	source_scope, target_scope, script_name
sys_scope_table_access	source_scope, target_scope, table_name
sys_script_validator	internal_type, ui_type
sys_translated	name, element, value, language
sys_translated_text	tablename, fieldname, documentkey, language
sys_ui_form	name, view, sys_domain
sys_ui_list	name, view, sys_domain, element, relationship, parent
sys_ui_message	key, language, code
sys_ui_related_list	name, view, related_list, sys_domain
sys_ui_section	name, view, caption, sys_domain
sys_ui_view	name
sys_user_group	name
sys_user_role	name
sys_wizard	name
ua_table_licensing_config	name

How customer update record names affect collisions

To understand coalescing, it helps to understand how records that don’t coalesce work. For most record types, when a customer update is moved to a new instance, the system doesn’t detect collisions for the following reason:

• When you create a record, it receives a unique sys_id. For most record types, the sys_id becomes part of the customer update record name. For example: sysevent_email_template_9e1998c078b71100a92ecacd80df1d39.
• Creating an identical record in the same table on another instance produces a customer update record name with a different sys_id. For example: sysevent_email_template_10b958c8653311005840134572f8e020

As a result, even though the records might be otherwise identical, the records have different names so the system doesn’t detect the collision.

Coalescing records, in contrast, use the following approach to naming records and determining collisions: The following customer update record types use some or all of their coalescing columns instead of the sys_id in their names.

• sys_dictionary
• sys_documentation
• sys_choice_set
• sys_ui_list
• sys_ui_related_list

The resulting identical record name in each instance helps the system to identify collisions even if the records have different sys_ids.

When a customer update is moved from one instance to another, it may be rewritten to match the target instance. The rewrite can involve changing the update name of the customer update and one or more sys_ids within the update. The rewrites are done when the record or the reference field is for a table that uses a coalesce strategy. This confirms that the customer update will be applied to the correct record. For example, if the sys_dictionary record for tablename.fieldname has sys_id "123456789" on instance A and sys_id "987654321" on instance B, when a customer update that refers to that record is retrieved from instance A and recorded in the sys_update_xml table on instance B, references to "123456789" are updated to read "987654321".

Preventing duplicate records

• Transfer data with update sets rather than recreating it on separate instances to verify the records have the same sys_id.
• Export and import records as XML files to verify the records have the same sys_id. See Export and import XML files.
• Enable a unique index for the table from the system dictionary. See Table administration.

Note:

The default records included in the baseline system will always have the same sys_id because the instance imports the records as XML files during instance provisioning.