`orca.maintain`

Table maintenance for a project's datasets: snapshot expiry.

The orca binary doesn't link iceberg, so expiry runs in a binary of the project's own that does: bin/orca_maintain.mojo,

from orca.maintain import maintain_main

def main() raises:
    maintain_main()

which orca expire and the scheduler's periodic expiry run (it is found as $ORCA_MAINTAIN, beside the orca binary, or on PATH). Run by hand it takes the snapshots to keep besides the policy's with --retain; orca expire works those out from Postgres (orca.expiry.retained) and passes them, which is how a run that can still resume keeps the snapshots it read and the ones it committed.

orca_maintain expire <dataset> [--older-than 30d|12h|90m|<ms>]
    [--keep N] [--retain ID,...] [--if-policy] [--dry-run] [--report]

Without --older-than and --keep the table's own history.expire.max-snapshot-age-ms and history.expire.min-snapshots-to-keep decide (iceberg's defaults where unset: five days, one snapshot); --if-policy expires nothing when the table has no history.expire.max-snapshot-age-ms, which is how the scheduler leaves tables orca didn't create alone. An expiry gives a table orca's metadata defaults (write.metadata.delete-after-commit.enabled, write.metadata.previous-versions-max) when it lacks them, so its own commit already deletes the old metadata files it drops from the log. --report lists metadata/ before and after, to count its files.

Structs

`Expiry`

struct Expiry

What one dataset's expiry did, for the line orca_maintain prints.

Implements Copyable, Movable, Writable

Fields: dataset: String; skipped: String: Why nothing was decided (no policy), or empty.; result: ExpireResult; snapshots_before: Int; snapshots_after: Int; metadata_files_before: Int: -1 when not counted (no --report).; metadata_files_after: Int; log_after: Int; seconds: Float64

`write_to`

def write_to(self, mut w: T)

Functions

`parse_duration_ms`

def parse_duration_ms(text: String) -> Int64

30d, 12h, 90m, 45s or plain milliseconds.

`expire_dataset`

def expire_dataset(dataset: String, older_than_ms: Int64 = Int64(-1), keep: Int = Int(-1), retain: List[Int64] = List(), if_policy: Bool = False, dry_run: Bool = False, report: Bool = False) -> Expiry

Expires dataset's snapshots: older than older_than_ms ago (or by the table's properties when that is -1), keeping at least keep (or the table's history.expire.min-snapshots-to-keep) and every snapshot in retain. The metadata is committed first, then the files nothing retained points at are deleted.

`maintain_main`

def maintain_main()

The orca_maintain command (see the module doc).

Aliases

comptime USAGE = "usage:\n  orca_maintain expire <dataset> [--older-than 30d|12h|90m|<ms>] [--keep N]\n      [--retain ID,...] [--if-policy] [--dry-run] [--report]"