Skip to content

PlanetScale alternative for MySQL: Vitess, branching and what $39 buys

11 min readPlanetScaleMySQLComparisonDatabases

Short version: as of November 2026, PlanetScale has no free plan, and its MySQL product is Vitess. The smallest Vitess cluster is PS-10 at $39 a month: 1/8 vCPU, 1 GiB of RAM, one primary and two replicas, 10 GB of storage included. You get high availability, online schema changes through deploy requests, and a path to sharding. You also get MySQL with things removed: no stored procedures, triggers or events, no LOAD DATA INFILE, and foreign keys only as an opt-in on unsharded databases. If you need one plain MySQL server and do not need HA or sharding, Layerbase runs standard MySQL from Solo at $5 a month, with branches of real data included. If you are going to shard, or your team lives in deploy requests, stay on PlanetScale.

Contents

What PlanetScale sells today

PlanetScale's plans page is direct about the free tier: "PlanetScale does not offer a free plan." Hobby ended in 2024, and I covered the fallout in the post on PlanetScale free tier alternatives. The self-serve plan is now called Base (Scaler Pro was renamed to it), and it covers three products: Vitess, Postgres, and Neki, a sharding layer for Postgres.

Prices from planetscale.com/pricing, checked 2026-11-13, for AWS us-east-1:

ProductEntry SKUMonthlyShape
Vitess (MySQL)PS-10$391/8 vCPU, 1 GiB, 1 primary + 2 replicas
Vitess (MySQL), next sizePS-20$59x86-64
Vitess (MySQL), metalM-160$6092 vCPU, 16 GiB, 110 GiB NVMe
Postgres, single nodePS-5$51/16 vCPU, 512 MiB
Postgres, HAPS-5$15Same size, 3 nodes
Postgres, metalM-10$501/8 vCPU, 1 GiB, 10 GiB, HA (arm64)

Single-node Vitess is not available. The cheapest MySQL you can buy from PlanetScale is a three-node cluster, which is the right design for production and an awkward one for a side project or a staging copy.

Vitess storage, from the plans page: 10 GB included, then $0.50 per instance per additional GB, which on an HA production branch works out to $1.50 per GB (a primary and two replicas). Automatic backups run every 12 hours and do not count against storage; backups you schedule yourself cost $0.023 per GB-month.

Vitess is not plain MySQL

Vitess is the reason PlanetScale can shard a MySQL database across many servers without your application knowing. It sits between your app and MySQL 8, routes queries, and runs schema changes online. That buys a lot. It also means some of MySQL is not there, and PlanetScale's compatibility page lists what:

  • No stored routines. "We do not support any form of stored routines." That covers stored procedures, functions, triggers and events.
  • No LOAD DATA INFILE.
  • No native user management. No CREATE USER or GRANT; credentials come from the dashboard, API or CLI.
  • No binlog replication out. "PlanetScale does not currently support binlog replication to external databases."
  • Fixed settings. The global time zone is UTC and the global SQL mode cannot be changed permanently. Only InnoDB.
  • No direct DDL on production once safe migrations is on; schema changes go through deploy requests.

Foreign keys are the one that used to be on this list and partly came off. According to the foreign key docs, they are now an opt-in per database, "only supported in unsharded environments," and deploy requests "do not validate the referential integrity of existing columns", so a constraint can be added over orphaned rows and they stay orphaned. Reverting a deploy that dropped a foreign key can leave orphans too.

None of that is a defect. It is the price of the architecture. But if your app was written against ordinary MySQL, with a trigger here and a stored procedure there, Vitess is a port rather than a move.

Layerbase MySQL is a standard MySQL server: 9.7, or 8.4 LTS if you want the long-term line. Foreign keys are enforced the way MySQL enforces them, routines and triggers come across with mysqldump --routines --triggers, and ALTER TABLE is just ALTER TABLE. What you give up is everything Vitess was adding: no sharding, no online schema-change workflow, and no HA on the shared plans.

Branching: deploy requests against data branches

Both products branch. They mean different things by it.

PlanetScale Vitess branches are about schema. From the branching docs, a new development branch copies only the schema; you get data by restoring a backup into the branch or by using the separate Data Branching feature. You change the schema on the branch, then open a deploy request into production. PlanetScale copies the affected table, applies the change to the copy, keeps it in sync, and cuts over. You can gate the cutover to a moment you choose, and you can revert for 30 minutes after a deploy while keeping data written in that window.

That workflow is excellent. Schema changes reviewed like pull requests, applied without locking a hot table, with an undo button, is something most teams rebuild badly by hand.

Layerbase branches are about data. A branch is a copy of the real database with its real rows, on 17 engines, MySQL included. It gets its own connection string and credentials and does not take one of your plan's database slots. With instant branching on, the branch shares storage with its parent instead of copying it. The GitHub App and the Vercel integration can create a branch when a git branch appears and remove it when the git branch goes, which gives each preview deploy its own data. Free includes 1 branch per database, Solo 3 and Pro 10. Details are in the branching docs.

There is no deploy request on Layerbase. You test a migration on a branch that holds production-shaped data, then run the same migration against production with whatever tool you already use. If online schema change on large tables is the main thing you need, that is a gap, and gh-ost or pt-online-schema-change against a standard MySQL is the usual answer to it.

Bill one: one small production MySQL

An app with one production MySQL database, a few GB of data, modest traffic.

PlanetScaleLayerbase
PlanPS-10 Vitess, BaseSolo
Monthly$39$5
Compute1/8 vCPU, 1 GiB, primary + 2 replicas768 MB always-on pool, 1 vCPU
Storage10 GB included, then $1.50/GB on HA10 GB included
BackupsEvery 12 hours, plus scheduled backups at $0.023/GB-monthEvery 12 hours, kept 7 days
HAYesNo
Meter on topStorage over 10 GB, scheduled backupsNone

These are not the same product. $39 buys three nodes and a failover story; $5 buys one MySQL server that stays up when pinned always-on. MySQL starts on Solo on Layerbase, and pinning it uses all of Solo's 768 MB pool, so a second database on the same plan can exist but sleeps after six idle hours. If you want two databases awake, Pro is $15 a month for up to 10 databases, a 1.5 GB pool, 25 GB, and backups kept 30 days. Past that, $10/month pool blocks add 1 GB of RAM, 1 vCPU and 25 GB each.

If HA is a requirement, the comparison stops here and PlanetScale wins it. If it is not, you are paying $34 a month for two replicas you did not need.

Bill two: production plus three development branches

Same app, now with three development branches for feature work, each running all month.

On PlanetScale Vitess, development branches bill for the time they run. Each database includes twice the hours in the month of development branch time at no cost, which is 1,440 hours in a 30-day month, so two branches running all month are free. A third adds about 720 hours at roughly $0.014 an hour, about $10, plus $0.50 per GB of branch storage. Call it $49 a month before storage, and the branches start schema-only unless you restore data into them.

On Layerbase Solo, the three branches are included: Solo allows 3 per database, each a copy of production's data, and none of them takes a database slot. The bill stays $5. On Pro, 10 branches per database are included at $15.

The honest caveat on my side: branches on a $5 plan are not production-sized compute. They are for running the migration and the test suite against real data, and the ones labeled preview sleep quickly when idle.

The Postgres side

PlanetScale is not MySQL-only anymore. PlanetScale Postgres starts at $5 for a single PS-5 node or $15 for three. Postgres branching works differently from Vitess: there are no deploy requests ("you make schema changes directly on each branch"), a branch restored from a backup includes schema and data, a branch created from the Branches page includes neither, and the plans page prices each development branch at $5 a month. Neki, the Postgres sharding layer, is in Platform Preview and not covered by an SLA, according to its docs.

Layerbase runs PostgreSQL on every plan, Free included, with the same branching as MySQL and point-in-time recovery on Pro for a database pinned always-on. If you are choosing PlanetScale for Postgres specifically, the comparison looks a lot like bill one: PlanetScale's $15 HA cluster against a flat plan without HA.

When PlanetScale is the better choice

You are going to shard. This is what Vitess exists for, and nothing on a single-server MySQL replaces it. If your growth plan is more shards rather than a bigger box, PlanetScale is the product.

Your team works in deploy requests. Online, gated, revertible schema changes reviewed like code are worth real money on a large table under load. Rebuilding that around gh-ost is possible and is work.

You need HA from day one. Vitess has no single-node option, and the entry PS-10 is already a primary with two replicas. Layerbase does not advertise HA or an uptime SLA on its shared plans.

You need fast local NVMe at scale. PlanetScale Metal starts at $609 for Vitess and $50 for Postgres. That is a hardware tier Layerbase's shared plans do not compete with.

The case for moving is a MySQL that does not need any of those: one database, an app written against ordinary MySQL with triggers or procedures, a few staging and preview copies, and a bill that should be a fixed number.

FAQ

Does PlanetScale have a free tier?

No. As of November 2026 PlanetScale's plans page says it "does not offer a free plan," and the Hobby plan ended in 2024. The cheapest database is a $5 single-node Postgres; the cheapest MySQL is a $39 three-node Vitess cluster.

What is the cheapest PlanetScale MySQL database?

PS-10 on Vitess at $39 a month: 1/8 vCPU, 1 GiB of RAM, a primary and two replicas, with 10 GB of storage included. Single-node Vitess is not offered.

Is PlanetScale MySQL compatible with regular MySQL?

Mostly at the query level, with real exceptions. PlanetScale's compatibility docs list no stored procedures, functions, triggers or events, no LOAD DATA INFILE, no native user management, no binlog replication out, and foreign keys only on unsharded databases with the setting turned on.

Does PlanetScale branching copy my data?

Not by default on Vitess: a new development branch copies only the schema, and data comes from restoring a backup or from the Data Branching feature. Layerbase branches are copies of the real data on 17 engines, MySQL included.

Does Layerbase have deploy requests?

No. Layerbase branches give you a copy of production data to test a migration against, and the migration itself runs against production with your usual tool. If online schema change on very large tables is the core need, PlanetScale's deploy requests are the stronger answer.

How much does Layerbase MySQL cost?

MySQL starts on Solo at $5 a month, which covers two databases and 10 GB, with the MySQL database pinned always-on using the whole 768 MB pool. Pro is $15 a month for up to 10 databases and 25 GB. MySQL is not on the free tier; MariaDB and PostgreSQL are.

Moving

PlanetScale Vitess speaks the MySQL wire protocol, so the data ports directly. layerbase.com/migrate/planetscale copies schema and data using a short-lived read-only password it creates and then deletes, and nothing is written back to PlanetScale. The app-side change is usually swapping @planetscale/database for mysql2 or your usual client, and turning foreign keys back on if you had them off. The full walkthrough is in migrating from PlanetScale to Layerbase.

To try it first, create a MySQL database, load a dump, branch it, and run your next migration against the branch.