Names, limits and statuses

Names

Name	Rules
Cluster	Lowercase letters, digits and hyphens, not starting or ending with a hyphen; unique in the resource group. The portal takes 3 to 50 characters, the API up to 63 — and up to 53 when PgBouncer is on, because the pooler's name adds -pooler-rw. Cannot be changed.
Database	Lowercase letters, digits and underscores, starting with a letter, at most 63 characters (the API also accepts a leading underscore and $). Also the name of the owning user. Cannot be changed.
Schema, table, column (Data Explorer forms)	Lowercase letters, digits and underscores, starting with a letter or an underscore, at most 63 characters

A cluster created through the API without a database name gets a database and a user named app.

Versions

	Versions
Offered by the portal's wizard	13, 14, 15, 16
Accepted by the API and the CLI	12, 13, 14, 15, 16, 17
Default	16

The version is the major version; the operator's image for it is used. A major version can only be raised — see Upgrade the major version.

Limits

What	Limit
Instances	1 to 10
PgBouncer instances	1 to 5
Pool modes	session, transaction, statement
Storage size	A whole number of Gi in the portal; Mi, Gi or Ti through the API. Only grows.
Memory per instance	At least 256Mi, request and limit
CPU and memory	Limit at least the request
Backup schedule	Five-field cron, UTC; default 0 3 * * *
Backup Retention Policy	A number followed by d, w or m; default 30d
Allowed CIDR Blocks	IPv4 or IPv6 ranges with a prefix length, comma-separated

Data Explorer limits

What	Limit
Statement run time	30 seconds (the API accepts up to 5 minutes)
Lock wait	5 seconds
Rows returned per batch	5,000 (the API can ask for more)
Rows per page on Browse	100 in the portal; up to 1,000 through the API
Rows per group of row changes	500
Row editing	Tables with a single-column primary key
Dry run before a DDL commit	At most 2 minutes old, same user, same SQL

Statuses

The Status of a cluster is the operator's own description of it, with two replaced:

Status	Means
Running	The operator reports the cluster healthy; Ready Instances on the Overview says how many instances serve. The API and the CLI show the operator's words, Cluster in healthy state.
Stopped	The cluster was stopped; no instance runs
Anything else	What the operator is doing, in its own words — such as Setting up primary while the first instance is prepared, or during a restore. It lasts until the cluster is Running again.