Fix grant ledger ownership and truncated-record recovery

This commit is contained in:
dingo
2026-10-02 06:24:56 -04:00
parent c588be237c
commit 7ac587743c
6 changed files with 180 additions and 9 deletions
@@ -41,6 +41,12 @@ The Deployment mounts it optionally (0440). Consumed grants are recorded in
`/objects/dev-session-grants.v1` on the region's persistent volume (fsynced before
each Welcome), so a restarted pod refuses grants its predecessor used. Corrupt or
unwritable ledgers fail closed (`error session-grant-ledger`, or sessions refused).
A sibling `.lock` file enforces one ledger owner across startup and compaction.
Do not delete or replace this lock file while a server might be running; its
presence does not mean a stale lock. Process exit releases the OS lock automatically.
A competing startup fails before modifying the ledger. Incomplete final records and
oversized lines are refused without rewriting; investigate the storage failure
rather than deleting grant history to force startup.
Rotation: create new credentials, replace the Secret, restart evo-region and
redistribute `dev-ca.pem`. Old grants stop verifying with the new key.