CSA domain - 30% of the exam

Database Management and Platform Security

Database Management and Platform Security is 30% of the CSA exam. These are the objectives it covers, each with practice questions, with every answer explained.

The domain in numbers

102
Practice questions
6
Objectives
17
Medium
85
Hard

Where people go wrong

  • Picking a server-side control at the wrong level

    The wrong answers pick a server-side mechanism that works at the wrong level. A before query business rule filters whole records, so it cannot hide one field, and a read only dictionary setting still displays the value. A read access control rule filters rows already retrieved, while a before query business rule changes the query itself. Access control rules test roles and conditions, never a list of named users, and a condition comparing Assigned to with the signed in user is what limits a write to one record's assignee.

    See it in a sample question
  • Reading an access control rule as passing on any one part

    The wrong answers treat the parts of one access control rule as alternatives, or let a script or condition be skipped. A rule lists required roles, a condition and sometimes a script, and all of them must evaluate true before the rule grants access. A user holding the itil role still fails when the condition is false or the script sets the answer to false. Within the role list, any one listed role is enough, and an empty role list adds no role requirement while the condition still applies.

    See it in a sample question
  • Editing the parent or the form instead of using a dictionary override

    To make an inherited field mandatory on one child table, the wrong answers edit the parent dictionary entry, add a UI policy, or create a second field on the child. Editing the Task entry makes the field mandatory on every table that extends Task. A UI policy runs on the form and leaves the field definition unchanged. A new field duplicates data and splits it across two columns. A dictionary override on the child table changes the attribute for that table alone.

    See it in a sample question
  • Giving coalesce behaviour it does not have

    A matched row is updated rather than skipped or rejected, so a batch of rows sharing one key leaves a single record holding the value from the last row processed. An empty coalesce value matches nothing, so those rows insert again on every run. Several coalesce fields combine into one query that needs all of them to point at the same record, with no first match order and no fallback key. A miss always inserts.

    See it in a sample question
  • Offering a reference, extension or choice where many-to-many is needed

    When records on each table can link to many records on the other, the wrong answers offer a reference field, table extension or a choice field. A reference field stores one value, so it models one to many at best. Extension is an inheritance link that shares columns, and a table has a single parent. A choice field holds static strings, not records. A many-to-many relationship creates an intermediate table with one row per pairing and a related list on both sides.

    See it in a sample question

Objectives in this domain

What to study

The official documents this domain's practice questions cite most.

Sample questions from this domain

Free sampleDatabase Management and Platform Securityhard

A user holding the itil role opens an incident record and every field renders except a custom field named u_cost_centre, which is absent from the form, absent from the incident list view and empty in a CSV export of the same records. The read access control rule on the incident table itself evaluates to true for this user. What explains the behaviour?

  • AA field level read access control rule on incident.u_cost_centre requires a role the user does not hold, and both the table rule and the field rule must pass before the value is returned. Correct
  • BA UI policy on the Incident form hides u_cost_centre for users without a finance role, so the field is suppressed on the form and in every other view of the data.
  • CA before query business rule on the incident table adds a condition that strips u_cost_centre from the result set returned to users without a finance role.
  • DThe dictionary entry for u_cost_centre is configured as read only for the itil role, so the stored value stays hidden until the user is granted write access to the field.
Field level access control rules are evaluated in addition to the table rule, so a passing table rule alone does not reveal a field. Access control is evaluated from most specific to least specific, and a read of a field requires the field level rule and the table level rule to both grant access. Because that evaluation happens on the server for every request, a denied field disappears from forms, lists and exports alike, which is what separates it from a form only mechanism such as a UI policy.

Why A is correct: Correct: table level and field level read rules are both evaluated, so passing the table rule still leaves the field rule able to deny the single field wherever the data is requested.

Why B is wrong: Tempting because a UI policy is a common way to hide a field and does render it invisible on the form, but a UI policy acts only on the form at runtime and cannot remove a field from a list view or from an export.

Why C is wrong: Tempting because a before query business rule really does filter what a user gets back, but it filters whole records rather than individual fields, so it would hide incidents, not one column of a visible incident.

Why D is wrong: Tempting because a read only dictionary setting does restrict what a user may do with a field, but read only still displays the value; it blocks editing rather than reading.

Free sampleDatabase Management and Platform Securityhard

A write access control rule on the incident table lists itil in its required roles, carries a condition of Active is true, and runs a script that sets the answer to false whenever the Caller field is empty. A user who holds the itil role edits an active incident that has no caller populated, and no other write rule on the table matches. What is the outcome and why?

  • AThe write succeeds, because the user satisfies the required role and the condition is true, and the script part of a rule is consulted only when the required roles list has been left empty.
  • BThe write succeeds, because the platform evaluates the parts of a rule in sequence and stops at the first check that grants access, which in this case is the successful role check.
  • CThe write is denied, because a rule that contains a script is skipped as ambiguous and the platform falls back to the wildcard write rule on the table, which grants nothing.
  • DThe write is denied, because the required role, the condition and the script must all evaluate true for the rule to grant access, and the empty caller makes the script return false. Correct
Within one access control rule the roles, the condition and the script are combined with logical and, so any failing part denies the operation. A single access control rule is satisfied only when the user holds one of the required roles, the condition evaluates true against the record, and the script sets the answer to true. Candidates often reason that holding the named role is sufficient, but the role check is one of three gates inside the same rule, so a script returning false denies a user who otherwise qualifies.

Why A is wrong: Tempting because a role check is the part candidates think of first, but the script is evaluated for every rule that has one, whether or not roles are also listed on that rule.

Why B is wrong: Tempting because evaluation order does matter across rules of differing specificity, but within one rule there is no first match wins behaviour; every part of it is required.

Why C is wrong: Tempting because falling back to a less specific rule is real behaviour when no rule matches, but a scripted rule is not skipped; it matched here and its script produced the denial.

Why D is correct: Correct: the parts of a single rule are combined with logical and, so one failing part denies the operation even when the role and the condition are both satisfied.

Free sampleDatabase Management and Platform Securityhard

The Description field is defined on the Task table and inherited by Incident, Problem and Change Request. It must become mandatory on Change Request while staying optional on the other two tables. Which mechanism delivers that at the data layer?

  • ACreating a dictionary override on the Change Request table for the Description field, which alters the mandatory attribute for that one table alone. Correct
  • BEditing the Dictionary entry for Description on the Task table and marking it mandatory, because each extending table can then relax the setting back locally.
  • CCreating a second Description field directly on the Change Request table and marking that copy mandatory, while leaving the inherited field in place unchanged.
  • DMarking the Description field mandatory through the Change Request form layout, which records the requirement for that table without touching the other two.
Recognise the dictionary override as the way to change an inherited field's attribute on one extending table without affecting the parent or its siblings. An extending table shares the parent's field definitions rather than copying them, so editing the parent dictionary entry propagates everywhere. A dictionary override exists precisely to break that symmetry: it stores a table specific value for one attribute of one inherited field, so Change Request enforces the mandatory setting while Incident and Problem continue to read the parent definition.

Why A is correct: Correct, because a dictionary override records a table specific value for an attribute of an inherited field, leaving the parent definition and the other child tables untouched.

Why B is wrong: Changing the parent dictionary entry is the direct route to the attribute, which is why it tempts, but the setting is inherited by every child table, so Incident and Problem would become mandatory as well.

Why C is wrong: Adding a field on the child table is possible, so this looks workable, but it produces two columns holding the same kind of data and splits the values written by users and integrations.

Why D is wrong: Form layout governs which fields appear and in what order on a form, so it is a real configuration surface, but it does not carry field attributes such as the mandatory setting.

Free sampleDatabase Management and Platform Securityhard

An administrator loads a staff CSV into an import set and transforms it to the user table. The field map for Employee number carries the coalesce flag, and no other field map coalesces. Of the 400 source rows, twelve carry the employee number 88231 with a different Department value on each row. What is the state of the user table after one transform run?

  • ATwelve separate user records are created for employee number 88231, because each source row is transformed into a target record of its own.
  • BOne user record exists for 88231, holding the Department value from the last of the twelve rows processed, and the import log shows updates. Correct
  • CThe transform halts at the second row carrying 88231 and the import log marks the whole run as failed against a duplicate key violation.
  • DThe first row carrying 88231 is transformed and the remaining eleven rows are skipped, so the earliest Department value survives the run intact.
Recognise that a coalesce match updates the existing target record, so repeated source rows overwrite one another and the last row processed wins. Coalesce turns the field map into a lookup key. For each staging row the transform queries the target table on the coalesce value; a hit updates that record with the mapped values and a miss inserts one. Twelve rows sharing 88231 therefore produce one insert followed by eleven updates against the same record, and the Department value left behind is the one carried by the final row transformed.

Why A is wrong: This is what happens with no coalesce field at all, which makes it a tempting reading, but a coalesce field forces the transform to look for an existing target record before inserting, so only the first of the twelve inserts.

Why B is correct: Correct: the first row inserts the record and each later row coalesces onto it and overwrites the mapped fields, so the final Department value is whichever row was transformed last.

Why C is wrong: Duplicate keys do abort a load in many relational tools, which makes this plausible, but a coalesce match on the platform is an update instruction rather than a constraint, so the run continues without error.

Why D is wrong: Skipping later matches sounds like sensible de-duplication, but coalesce never skips a matched row: it updates the matched record, so the earliest value is overwritten rather than protected.

Free sampleDatabase Management and Platform Securityhard

A requirement states that one problem record can be linked to many incident records, and that one incident can equally be linked to many problem records, with the links visible from both sides. Which data model construct supports that requirement?

  • AA reference field on the Incident table targeting Problem, since a reference field holds several target records whenever more than one of them matches the incident.
  • BExtending the Incident table from the Problem table, so that each incident inherits the fields of the problem records that it happens to relate to.
  • CA dictionary override on the Problem table, which alters the inherited reference field so that it accepts several target records at once instead of one.
  • DA many to many relationship, which creates a separate table whose records each pair one incident with one problem and surfaces as a related list on both tables. Correct
Explain that a many to many relationship uses an intermediate table of paired records, so both tables can hold any number of links to each other. A single reference field can store only one target record, so it cannot express links running in both directions without limit. A many to many relationship solves this with a separate table whose every record names one member from each side. Because the pairing lives in its own row, either table can be the starting point and the association appears as a related list on both.

Why A is wrong: A reference field is the obvious way to connect two tables, which makes this tempting, but a single reference stores exactly one target record, so it models one to many at best and cannot be read from both sides.

Why B is wrong: Extension does create a link between two tables, but it is an inheritance link that shares column definitions with a parent, not an association between individual records at run time.

Why C is wrong: Dictionary overrides do change how an inherited field behaves on one table, so the mechanism is real, but they adjust attributes such as mandatory or default value rather than converting a reference into a multiple value field.

Why D is correct: Correct, because the intermediate table stores one row per pairing, so each side can hold any number of links and the related list renders the association from either record.

Other domains in this exam

See also the CSA cert hub, the study guide, and the cheat sheet.

Examworthy is not affiliated with or endorsed by ServiceNow. Original, blueprint-aligned practice material only.