Comparing two Clusters

/compare shows two Clusters side by side: why does staging behave differently from production, which of two Clusters drifted from the template. Open it from the cluster list (tick two rows, then Compare), from any Cluster (Compare with…, which fills in the first side), from the top bar or from the command palette. The two Clusters are in the URL (/compare?left=team-a/orders&right=team-a/billing), so a comparison can be shared.

How a comparison is made

One request answers the whole page (Compare in internal/console/status/compare.go, behind the /api/v1/compare handler). Both sides are authorized before either is read, and both are read from the console's informer cache, never from the API server at request time.

What is compared

Section	Rows
Versions and image	PostgreSQL major, status.image, spec.imageName, spec.imageCatalogRef, the detected CloudNativePG version.
Instances	spec.instances, ready instances, primary update strategy and method, failoverDelay, switchoverDelay, whether it is a replica Cluster.
PostgreSQL parameters	Every key of spec.postgresql.parameters on either side, and shared_preload_libraries.
Resources	CPU and memory requests and limits first, then any other resource (huge pages included).
Storage	Data and WAL volume size and storage class, tablespaces.
Backup configuration	Method (plugin or in-tree), object store, retention policy, backup target, scheduled backups.
Topology	Synchronous replication, pod anti-affinity (enabled, type, topology key), node selector, tolerations, poolers.

The values are what the Cluster declares (its spec), plus the few status fields named in the rows; they are not what PostgreSQL is running with. The page says so in its blind spots.

Same, differs, cannot tell

The server decides every row; the page only draws it.

• Same: both sides were read and are equal. A memory parameter written in two spellings that PostgreSQL reads as the same number of bytes (2GB and 2048MB, 131072 and 1GB for shared_buffers) is the same, with the note "the same value, spelled differently". Units are parsed exactly as PostgreSQL does: case-sensitive kB, MB, GB, TB, the parameter's base unit for a bare number.
• Differs: both sides were read and are not equal; the row is highlighted. "Not set" (the default applies) on one side and a value on the other differs.
• Cannot tell: one side could not be read by the console, including a list that failed this time (a failed read is never "nothing declared"). It is never counted as the same: the total of differences includes it, and Show only differences keeps it.

Values are compared in full; a long value is shortened for display only, after the comparison. Parameter names are matched case-insensitively, as PostgreSQL matches them (Work_Mem and work_mem are one row, with a note); of two spellings on one side, the later in sorted order wins, as the later line of postgresql.conf does. At most 400 parameter names are compared; any beyond that are counted in a blind spot.

Show only differences hides the rows that are the same.

Authorization

GET /api/v1/compare?leftNamespace=&leftName=&rightNamespace=&rightName= runs the same access review as every per-Cluster view (get clusters) for each side, before reading either. If you may not see one of them, the whole comparison is refused with 403; a missing Cluster gives 404. The comparison is never half shown.