Aurora separates a shared cluster volume from writer and reader instances. Reader scaling, Serverless compute scaling and Global Database recovery address different capacity and availability needs.
Aurora basics
Amazon Aurora is AWS’s MySQL- and PostgreSQL-compatible relational database, managed through Amazon RDS.
How it works
- Create an Aurora cluster with the MySQL or PostgreSQL edition.
- Applications connect as they would to MySQL or PostgreSQL.
- AWS manages the database.
Exam cues
- If the question says MySQL- or PostgreSQL-compatible managed database built by AWS, think Aurora.
Remember
- Compatible with MySQL and PostgreSQL.
Trap: Aurora is relational; DynamoDB is the NoSQL choice.
Da documentação da AWS (3)
- Amazon Aurora is a fully managed relational database engine compatible with MySQL and PostgreSQL.
- Amazon Aurora is a fully managed relational database engine compatible with MySQL and PostgreSQL.
- Aurora applies operating system and database engine patches during the cluster maintenance window, so the customer does not patch servers.
Storage, replicas, and auto scaling
An Aurora cluster separates database instances from a shared storage volume. The writerThe database instance that modifies the shared cluster data. handles changes; readerA read-only database instance that shares the writer storage and can be promoted after writer failure. instances serve read-only work and can take over if the writer fails. Storage copies protect data, while readers in other Availability Zones provide replacement compute capacity.
How it works
- Store six copies of the data across three Availability Zones. This storage replication exists even with only one database instance. Storage grows with data, subject to engine-version limits.
- Add up to 15 Aurora Replicas, meaning read-only instances sharing the same volume. Adding a reader does not copy all table data again.
- Send read/write connections to the cluster endpointThe connection address that follows the current writer in an Aurora cluster., the address that follows the writer. Send read-only connections to the reader endpointThe address that balances new read-only connections across available readers., which balances connections across readers. It does not redistribute queries within a connection.
- For automatic reader counts, use Aurora Auto Scaling target trackingA scaling policy that adjusts reader count to keep a metric near a chosen value. and configure reader limits. The average includes every reader and excludes the writer. At least one reader must exist for these metrics to drive scaling.
- Auto Scaling removes only the readers it created, even though manually added readers contribute to its metrics.
Exam cues
- If the question says durable storage but only one database instance, think add a reader in another zone, because storage copies alone are not spare compute.
- If the question says automatically adjust reader count, think Aurora Auto Scaling, because it uses target tracking within configured limits.
- If the question says a single read session overloads one reader, think connection balancing, because the reader endpoint does not balance each query.
Remember
- Six data copies across three zones; one writer and up to 15 readers.
- Aurora is MySQL and PostgreSQL compatible; AWS patches it in the maintenance window.
- The cluster volumeThe shared storage holding data for the writer and readers of an Aurora cluster. shrinks when data is removed, on supported versions.
Trap: Manually created readers affect Auto Scaling metrics but are not removed by its scale-in actions. With no readers, the reader endpoint reaches the writer and can accept writes.
The cluster endpoint selects the writer.
The reader endpoint balances new connections among available readers.
The writer modifies data in the shared volume.
Readers use the same volume without making another full copy of the tables.
Da documentação da AWS (23)
- An Aurora cluster has one writer instance using a shared cluster volume.
- An Aurora cluster supports up to 15 reader instances that share the writer cluster volume.
- Aurora stores six copies of cluster data across three Availability Zones, independent of the number of database instances.
- Adding an Aurora instance connects it to the existing cluster volume without making a new copy of the table data.
- Aurora cluster storage automatically grows as data increases; maximum size and resizing behavior depend on engine version.
- Aurora Replicas serve reads and can be promoted when the writer fails; placing instances in different Availability Zones improves instance availability.
- The cluster endpoint follows the current writer and supports reconnection after failover.
- The reader endpoint balances new connections among available Aurora Replicas, not individual queries within a connection.
- With no Aurora Replicas, the reader endpoint connects to the writer and can accept writes.
- Aurora Auto Scaling uses target tracking to add and remove reader instances within configured limits; it does not scale the writer.
- Aurora Auto Scaling metric averages include all readers, including manually added readers and custom-endpoint members, and exclude the writer.
- If an Aurora cluster has no readers, reader metrics are not calculated and scaling alarms remain inactive; at least one reader is needed for metric-driven Auto Scaling.
- Aurora Auto Scaling removes only readers it created.
- RDS Proxy MaxConnectionsPercent caps database connections as a percentage of the Aurora max_connections setting; it does not open all allowed connections in advance.
- RDS Proxy reserves internal connections, so MaxDatabaseConnectionsAllowed can be lower than the value calculated from MaxConnectionsPercent.
- An Aurora custom endpoint points to a set of DB instances you choose and balances connections among them, for example to send reporting queries to specific replicas.
- A MySQL 5.7 or 8.0 database can be migrated physically to Aurora MySQL by copying Percona XtraBackup files to Amazon S3 and restoring from them.
- If an Aurora cluster has no Aurora Replicas, a failure recreates the primary instance in the same Availability Zone, which typically takes under 10 minutes; promoting a replica is much faster.
- An Aurora cluster volume grows automatically up to 256 TiB on supported engine versions, and you pay only for the space used.
- Aurora MySQL Backtrack rewinds the same cluster to an earlier time in minutes without a restore; it must be enabled when the cluster is created or restored, and the window is at most 72 hours.
- Aurora machine learning lets SQL in Aurora MySQL or Aurora PostgreSQL call Amazon Bedrock, Amazon Comprehend or SageMaker AI models.
- To move from RDS for MySQL to Aurora MySQL with little downtime, create an Aurora read replica of the RDS instance, wait for replica lag 0, point clients to it and promote it.
- On supported versions, an Aurora cluster volume shrinks automatically when data is removed, such as by dropping or truncating a table, which lowers storage charges.
Global database switchover and failover
Aurora Global Database keeps copies of a database in multiple Regions using asynchronous replicationCopying writes to another Region after they can already be committed on the primary., meaning a write can commit before a secondary receives it. SwitchoverA planned primary-Region change that synchronizes the secondary before promotion. moves the primary role during a planned operation; failoverPromoting a replacement writer or primary cluster to recover from failure. recovers from an outage that prevents normal synchronization.
Recovery point objectiveThe acceptable amount of recent data loss expressed as a time interval. (RPO) measures how much recent data can be lost, expressed as time. Recovery time objectiveThe time allowed for a system to return to service after an outage. (RTO) measures how long service takes to resume.
How it works
- Prepare a secondary with database instances and compatible engine versions. Managed operations require matching major and minor versions; permitted patch differences depend on the version. A headless secondaryA secondary cluster that has storage but no database instances., meaning storage without database instances, needs an instance before promotion.
- For a planned move, use switchover. Aurora synchronizes the target, makes the old primary read-only and promotes a writerThe database instance that modifies the shared cluster data. in the target Region. The topology stays intact and RPO is zero, though the role change briefly interrupts service.
- For an outage, stop application writes to the old primary and select the least-lag secondary. Managed failover does not wait for synchronization and can lose unreplicated writes. Write fencingAn attempt to stop writes to the old primary during global failover., Aurora’s attempt to block writes to the old primary, is best effort; split-brainA condition where old and new primary clusters can both accept writes. means both sides might accept writes.
- Use the global writer endpointA stable connection hostname that follows the primary Region of a global database., the stable hostname that follows the primary Region. Confirm DNSThe name resolution mechanism through which an endpoint follows a new primary. changes, meaning the hostname resolves to the new destination, before resuming writes. With RDS Proxy, use the read/write proxy endpointA database connection address that selects the appropriate instance or group of instances. associated with the new primary.
- Align parameter groupsA collection of database configuration settings applied to a cluster or instance., monitoring and regional service integrations. They are not inherited. Managed failover rebuilds the other secondaries and adds the recovered old primary back as a secondary; later switchover can return the primary role.
| Recovery choice | Data and operation behavior |
|---|---|
| Switchover | Synchronizes first; planned role change with RPO zero |
| Managed failover | Promotes without waiting for synchronization; possible data loss |
| Copied cluster snapshotA saved database storage image from which a new cluster can be restored. | Restores backup data in another Region |
| AWS Backup cross-Region copy | Produces a copied recovery point; continuous copies become snapshots without PITRRestoring a database to a chosen time within its retained backup history. |
Point-in-time recovery (PITR) means restoring to a selected time within retained backup history. Aurora does not support native cross-Region automated backup replication. Use copied cluster snapshots, supported AWS Backup copies or database replication for regional recovery.
For event notifications, configure an Aurora event subscriptionA configuration that publishes selected database resource events to an SNS topic. with resources, event categories and an SNS topic. The topic must permit RDS event publishing; its policy can restrict source resources or allow cross-account notifications.
Exam cues
- If the question says a planned primary-Region move must avoid data loss, think switchover, because the target synchronizes first.
- If the question says a primary Region is unavailable, think failover to the least-lag secondary, because waiting for the failed Region to synchronize is not the recovery path.
- If the question says regional recovery from backup, think copied cluster snapshots or supported AWS Backup copies, because Aurora lacks native automated backup replication across Regions.
- If the question says database events should notify recovery workflows, think event subscriptions to SNS, because they publish selected event categories from selected resources.
Remember
- Switchover RPO: 0. Failover can lose unreplicated writes; –allow-data-loss accepts this.
- Parameter groups, alarms and service integrations need regional alignment.
- Unplanned failover RPO: typically seconds. Aurora PostgreSQL
rds.global_db_rpopauses primary commits when every secondary lags past it.
Trap: Global replication is asynchronous. An unchanged endpoint name does not remove replication loss or DNS propagation concerns during failover.
Aurora synchronizes the selected secondary before a planned switchover.
The old primary becomes read-only and the target assumes the primary role.
The global writer endpoint points to the new primary; observe its DNS change before resuming writes.
Da documentação da AWS (21)
- Aurora Global Database replicates asynchronously from a primary Region to secondary Regions for regional reads and disaster recovery.
- Recovery point objective measures acceptable data loss as a time interval.
- Recovery time objective measures the time needed to resume service after an outage.
- Switchover, formerly managed planned failover, synchronizes the selected secondary before changing primary Region and has RPO zero.
- Switchover makes the old primary read-only, promotes a writer in the selected secondary and preserves the existing regional topology, with a brief interruption.
- Managed failover handles an unplanned outage without waiting for synchronization; asynchronous replication lag can cause lost writes.
- The failover-global-cluster CLI operation uses --allow-data-loss to request failover that can lose unreplicated writes.
- Managed switchover and failover require equal major and minor engine versions; allowed patch-level differences depend on the engine version.
- Choose the secondary with the least replication lag to minimize failover data loss.
- AWS recommends taking applications offline before managed failover to prevent writes to the old primary; automatic write fencing is best effort and split-brain remains possible.
- The global writer endpoint keeps the same name and points to the new primary after switchover or failover; applications must observe the DNS change before resuming writes.
- A promoted secondary does not inherit the old primary parameter groups, monitoring settings or service integrations; these must be aligned for the new Region.
- After Global Database failover, RDS Proxy users must direct writes to the read/write proxy endpoint associated with the new primary cluster.
- Managed failover rebuilds other secondary clusters and adds the recovered old primary back as a secondary; switchover can later return the primary role to the original Region.
- A secondary without database instances needs an instance added before switchover or failover.
- Aurora DB clusters do not support native cross-Region automated backup replication; cluster snapshots can be copied to another Region for recovery.
- AWS Backup supports Aurora cross-Region backup copies; copies of continuous backups become snapshot recovery points without point-in-time recovery.
- Aurora event subscriptions select database resources and event categories and publish notifications to an SNS topic.
- RDS event publishing requires permissions on the SNS topic; a custom topic policy can restrict source resources or allow cross-account notification.
- An Aurora global database replicates asynchronously, so the RPO of an unplanned cross-Region failover is typically measured in seconds rather than zero.
- For an Aurora PostgreSQL global database, the rds.global_db_rpo parameter sets a target RPO; when every secondary lags beyond it, the primary pauses commits until one catches up.
Serverless v2 capacity and promotion
Aurora Serverless v2 changes database compute capacity as load varies. Each writerThe database instance that modifies the shared cluster data. or readerA read-only database instance that shares the writer storage and can be promoted after writer failure. uses Aurora capacity unitsA compute capacity measure combining approximately 2 GiB of memory with CPU and networking. (ACUs), which combine about 2 GiB of memory with corresponding CPU and networking resources. The cluster’s stored data stays separate from compute scaling.
How it works
- Set the cluster’s minimum and maximum capacity range before adding Serverless instances. Each instance has its own current capacity within that range.
- Aurora adjusts capacity during active connections and transactions without waiting for the database to become idle.
- Provisioned and Serverless instances can coexist. A Serverless reader in promotion tierA priority from 0 to 15 that influences which reader takes over as writer. 0 or 1 uses the ACU equivalent of a provisioned writer’s memory as its capacity floor.
- To move a provisioned writer to Serverless v2, add or convert a Serverless reader and fail over to promote it. The existing cluster and data remain in place.
- When an instance auto-pauses, its compute-capacity charge stops, but storage and other cluster charges continue. A new connection waits for the instance to resume.
Exam cues
- If the question says unpredictable database load with active transactions, think Serverless v2, because it scales instance compute without waiting for idle connections.
- If the question says a provisioned writer must move to Serverless with little downtime, think add or convert a Serverless reader and promote it through failoverPromoting a replacement writer or primary cluster to recover from failure., because the cluster and data remain in place.
- If the question says a Serverless reader must be ready to replace a provisioned writer, think tier 0 or 1, because its capacity floor uses the ACU equivalent of the writer’s memory.
Remember
- One ACU includes about 2 GiB of memory with corresponding CPU and networking.
- Provisioned and Serverless instances can coexist in a cluster.
- Paused instances have no compute-capacity charge; storage and other cluster charges continue.
- Serverless capacity: ACUs billed per second.
Trap: Normal capacity scaling can occur during active transactions, but a connection to a paused instance must wait for it to resume.
A tier 0 or 1 reader keeps its minimum tied to writer capacity within the configured maximum.
Detalhes das caixas (1)
- Tier 2 to 15
Independent readers remain within the cluster range and might need to scale up after promotion.
Da documentação da AWS (13)
- Aurora Serverless v2 scales compute capacity for each writer or reader within a cluster-level minimum and maximum capacity range.
- An Aurora capacity unit represents approximately 2 GiB of memory with corresponding CPU and networking resources.
- A cluster must define ServerlessV2ScalingConfiguration before instances of class db.serverless can be added.
- Aurora Serverless v2 can scale during open connections and active transactions without waiting for a quiet point or disrupting current database operations.
- Aurora clusters can combine provisioned and Serverless v2 instances.
- With a provisioned writer, a tier 0 or 1 Serverless v2 reader uses the ACU equivalent of writer memory for its capacity floor.
- Serverless v2 uses the same six-copy, three-Availability-Zone Aurora storage as provisioned capacity; compute scaling does not resize stored data.
- Setting minimum capacity to 0.5 ACU or greater turns off auto-pause.
- Paused Serverless v2 instances have no instance-capacity charge, but storage and other cluster charges continue.
- A connection to an auto-paused instance must wait for it to resume, so applications must tolerate resume latency.
- Aurora Auto Scaling reader-count policies are unsupported for Serverless v2; create readers in advance and let them scale compute.
- A provisioned Aurora DB instance can be converted to Serverless v2 by modifying its instance class; to keep writer downtime low, convert or add a Serverless v2 reader and then fail over to it.
- Aurora Serverless measures capacity in Aurora Capacity Units (ACUs) and bills them per second.
Storage configurations and Babelfish
Two Aurora choices that come down to money and migration effort: the storage configuration decides how I/O is billed, and Babelfish lets SQL Server applications talk to Aurora PostgreSQL.
How it works
- Pick Aurora Standard (pay per I/O) or Aurora I/O-Optimized (no I/O charges) per cluster.
- Compare I/O cost to total Aurora cost: 25% is the break-even rule of thumb from AWS.
- For SQL Server apps, enable Babelfish on an Aurora PostgreSQL cluster; it opens a TDS endpointA database connection address that selects the appropriate instance or group of instances. next to the PostgreSQL one.
Exam cues
- If the question says I/O-heavy Aurora with unpredictable I/O bills, think Aurora I/O-Optimized.
- If the question says migrate SQL Server apps to PostgreSQL with minimal code changes, think Babelfish (plus AWS DMS for the data).
Remember
- I/O-Optimized at 25% or more of spend on I/O; Standard below that.
- Standard to I/O-Optimized once every 30 days.
- Babelfish: SQL Server clients over TDS to Aurora PostgreSQL.
Trap: I/O-Optimized isn’t a performance tier you always want. With low I/O, Standard costs less.
Da documentação da AWS (4)
- Aurora I/O-Optimized charges only for instance usage and storage with no charge for read and write I/O, and is the best choice when I/O is 25% or more of total Aurora spending.
- Aurora Standard adds a charge per million I/O requests and is the best choice when I/O is less than 25% of total Aurora spending.
- You can switch from Aurora Standard to I/O-Optimized once every 30 days and back to Standard at any time, without downtime on non-NVMe instances.
- Babelfish for Aurora PostgreSQL accepts connections from SQL Server clients over the TDS wire protocol, so SQL Server applications work with Aurora PostgreSQL with few code changes and no driver changes.
Scaling, connections and integrations
Aurora offers separate controls for readerA read-only database instance that shares the writer storage and can be promoted after writer failure. capacity and database connections, plus ways to clone data and invoke Lambda. Choose the feature that matches the problem: scaling readers, reusing connections, creating a test cluster or calling a function.
How it works
- Register the cluster with Application Auto Scaling under the
rdsnamespace and therds:cluster:ReadReplicaCountdimension. Set a target and limits so Aurora Auto Scaling can add or remove readers. The predefined targets use average reader CPU or connections. See the registration details. - Connect clients to the RDS Proxy endpointA database connection address that selects the appropriate instance or group of instances.. The proxy manages reusable database connections separately from client connections. Set MaxConnectionsPercent to cap database connections as a percentage of
max_connections. - Create an Aurora clone for a test environment. It initially shares data pages with the source using copy-on-write, which uses little extra space.
- Give an Aurora MySQL cluster access to Lambda, then call a function from SQL using
lambda_syncorlambda_async.
Exam cues
- If the question says read demand changes, think Aurora Auto Scaling, because it adjusts the reader count within configured limits.
- If the question says many clients need a cap on database connections, think RDS Proxy and MaxConnectionsPercent, because the proxy reuses database connections.
- If the question says create a fast, space-efficient test cluster from production data, think Aurora cloning, because the clone initially shares data pages.
- If the question says Aurora MySQL must call Lambda from SQL, think
lambda_syncorlambda_async, because these native functions invoke Lambda after the cluster has access.
Remember
- The predefined reader metrics are
RDSReaderAverageCPUUtilizationandRDSReaderAverageDatabaseConnections. MaxDatabaseConnectionsAllowedcan be lower than the percentage calculation because RDS Proxy reserves internal connections. The proxy does not open all allowed connections in advance.- Automated backup storage up to the latest cluster volumeThe shared storage holding data for the writer and readers of an Aurora cluster. size is free.
Trap: Aurora Auto Scaling does not scale the writerThe database instance that modifies the shared cluster data. and cannot add readers to a Global Database secondary or an Aurora Serverless v2 cluster. Pre-created Serverless readers scale compute capacity instead.
Da documentação da AWS (12)
- Aurora Auto Scaling uses target tracking to add and remove reader instances within configured limits; it does not scale the writer.
- Application Auto Scaling registers an Aurora cluster with service namespace rds and scalable dimension rds:cluster:ReadReplicaCount.
- RDSReaderAverageCPUUtilization is average CPU utilization across Aurora readers.
- RDSReaderAverageDatabaseConnections is the average connection count across Aurora readers.
- RDS Proxy manages reusable database connections separately from client connections; clients connect to the proxy endpoint.
- RDS Proxy MaxConnectionsPercent caps database connections as a percentage of the Aurora max_connections setting; it does not open all allowed connections in advance.
- RDS Proxy reserves internal connections, so MaxDatabaseConnectionsAllowed can be lower than the value calculated from MaxConnectionsPercent.
- Aurora Auto Scaling is unsupported on Global Database secondary clusters.
- Aurora Auto Scaling is unsupported for Serverless v2; pre-created Serverless readers scale compute capacity instead of being added by a reader-count policy.
- Aurora cloning creates a new cluster that initially shares the source's data pages using copy-on-write, so it is fast and uses little extra space; it suits test environments built from production data.
- Aurora MySQL can invoke a Lambda function from SQL with the native functions lambda_sync and lambda_async, once the cluster has access to Lambda.
- Aurora provides automated backup storage equal to the latest cluster volume size at no charge; only usage beyond that allowance is billed.
Flashcards
Nada para revisar aqui agora. Escolha outros tópicos ou outro filtro.
·
Todos os 37 flashcards em lista
What kind of database is Amazon Aurora?
Amazon Aurora is a fully managed relational database engine compatible with MySQL and PostgreSQL.
A company runs MySQL and PostgreSQL applications and wants a managed relational database that AWS patches for it. Which AWS database fits?
Amazon Aurora.
It is compatible with MySQL and PostgreSQL, and AWS applies OS and engine patches in the maintenance window.
What separates Aurora storage durability from database instance availability?
Aurora stores six data copies across three Availability Zones regardless of instance count. One writer and up to 15 readers share that volume; readers in other zones provide promotion targets.
An Aurora application sends every read over one persistent reader-endpoint connection. Why does one reader remain busy?
The endpoint balances connections, not individual queries. A persistent connection stays with its selected reader.
An Aurora writer fails and the application must reconnect without selecting an instance hostname. Which endpoint applies?
The cluster endpoint. It follows the current writer, while Aurora can promote a reader to take over.
An Aurora cluster needs to handle more read traffic and provide a failover target. What should the team add?
Add an Aurora Replica. Readers serve read traffic and can be promoted if the writer fails.
An Aurora cluster has variable read demand. How can it adjust read capacity automatically?
Use Aurora Auto Scaling to add or remove reader instances within configured limits. It does not scale the writer.
Analytics queries must go only to two large Aurora replicas, while the app uses the others. Which Aurora feature fits?
A custom endpoint that includes just those two replicas.
How can an Aurora cluster improve database instance availability across Availability Zones?
Place Aurora Replicas in different Availability Zones. A reader can be promoted if the writer fails.
The writer of a single-instance Aurora cluster fails. What does Aurora do, and how long can it take?
It recreates the primary instance in the same AZ, typically within 10 minutes. Add an Aurora Replica in another AZ so failover becomes a fast promotion.
An Aurora application needs read scaling without copying table data for each database instance. How do replicas use storage?
Aurora readers share the cluster volume with the writer. Adding an instance does not make a new copy of the table data.
An Aurora workload needs to distribute read connections across replicas. Which endpoint should it use?
Use the reader endpoint. It balances new connections among available Aurora Replicas.
A production RDS for MySQL database must move to Aurora MySQL with minimal downtime. What path does AWS recommend?
Create an Aurora read replica of the RDS instance; when replica lag reaches 0, switch clients and promote the replica.
A team drops large tables in an Amazon Aurora cluster. Does it keep paying for the storage those tables used?
No, on supported versions. Aurora shrinks the cluster volume automatically when data is removed, lowering storage charges.
An Aurora Global Database must move its primary Region during planned maintenance without data loss. How does this differ from disaster failover?
Use switchover: it synchronizes the target and has RPO zero. Unplanned failover does not wait for synchronization and can lose writes; --allow-data-loss explicitly accepts that risk.
What should an Aurora Global Database team prepare before and after a managed failover?
Stop old-primary writes, choose the least-lag secondary and use the global writer endpoint. Check engine compatibility, DNS changes, parameter groups, monitoring and regional integrations before resuming writes.
An Aurora cross-Region recovery design proposes native automated backup replication. Is that supported?
No. Use copied cluster snapshots, supported AWS Backup copies or Global Database replication. A copied continuous backup is a snapshot recovery point without PITR.
An Aurora team needs database event notifications for recovery workflows. What should it configure?
An event subscription selecting resources and event categories, targeting an SNS topic with permission for RDS to publish.
An Aurora Global Database failed over, but clients use an RDS Proxy in the old primary Region. What connection change is required?
Direct writes to the read/write endpoint of the proxy associated with the new primary cluster.
Can an Aurora Global Database immediately promote a headless secondary during switchover or failover?
No. Add a database instance to that secondary first. Headless means the secondary has storage but no database instance.
Does an Aurora global database guarantee replication lag below one second, so an unplanned Region failover loses no data?
No. Replication is asynchronous; the RPO of an unplanned failover is typically some seconds.
A switchover synchronizes first and has RPO zero.
An Aurora PostgreSQL global database must never lose more than a set amount of recent data in a Region failover. Which setting enforces that, and what is the trade-off?
rds.global_db_rpo. When every secondary lags beyond it, the primary pauses commits until one catches up.An Aurora workload has unpredictable bursts. What does Serverless v2 scale, and does ordinary capacity scaling require idle connections?
It scales each writer or reader compute capacity within the cluster ACU range. An ACU includes about 2 GiB of memory with CPU and networking. Capacity scaling can occur during open connections and active transactions.
Can Aurora Serverless v2 mix with provisioned instances, and what keeps a high-priority Serverless reader ready for a provisioned writer?
Yes. Mixed clusters are supported; a tier 0 or 1 Serverless reader uses the ACU equivalent of the writer memory as its capacity floor.
Does auto-pausing an Aurora Serverless v2 instance stop every cluster charge and allow immediate connections?
No. Instance-capacity charges pause, but storage and other charges continue. New connections wait for resume.
A provisioned Aurora cluster must move its writer to Serverless v2 for traffic spikes, with little downtime. What is the usual path?
Add (or convert) a Serverless v2 reader, then fail over so it becomes the writer. No new cluster or data migration is needed.
How is Amazon Aurora Serverless v2 compute billed?
In Aurora Capacity Units (ACUs), billed per second as capacity scales.
An Aurora cluster's I/O charges are 40% of its bill and fluctuate month to month. Which change lowers and stabilizes the cost?
Switch the cluster to Aurora I/O-Optimized: no per-I/O charges, best when I/O is 25% or more of spend.
A SQL Server application needs to move to Aurora PostgreSQL with few code changes. Which Aurora feature supports this migration?
Babelfish for Aurora PostgreSQL accepts SQL Server client connections without changing database drivers.
A company wants to move a SQL Server application to Aurora PostgreSQL with minimal code changes and without changing database drivers. Which feature fits?
Babelfish for Aurora PostgreSQL, which understands the SQL Server TDS protocol and T-SQL.
An Aurora cluster has light, steady I/O that is under 10% of its spend. Should it use I/O-Optimized?
No. Aurora Standard is cheaper when I/O is under 25% of total Aurora spending.
How does an Aurora cluster automatically add read-serving instances as demand changes?
Aurora Auto Scaling uses Application Auto Scaling target tracking on rds:cluster:ReadReplicaCount. Predefined metrics are RDSReaderAverageCPUUtilization and RDSReaderAverageDatabaseConnections.
A serverless application opens many short-lived connections to an Amazon Aurora database. What reduces the connection load on the database?
Amazon RDS Proxy. It pools and reuses database connections, and clients connect to the proxy endpoint.
Can Aurora Auto Scaling add readers to a Global Database secondary or an Aurora Serverless v2 cluster?
No. Reader-count Auto Scaling does not support those configurations. Pre-created Serverless readers can scale their compute capacity.
Developers need a full copy of a 5 TB production Aurora database for testing, created in minutes and at low storage cost. Which feature fits?
Aurora cloning (copy-on-write: shares pages until either side changes them).
When a row is deleted in Aurora MySQL, a Lambda function must run and forward the change. How can the database call Lambda directly?
With the Aurora MySQL native functions lambda_sync or lambda_async, after granting the cluster access to Lambda.
How much Amazon Aurora automated backup storage is included at no extra charge?
An amount equal to the latest cluster volume size. Only backup usage beyond it is billed.