The SDC Manatee instance puts daily backups into /var/spool/pg_dumps:
https://github.com/joyent/sdc-manatee/blob/master/pg_dump/pg_backup_common.sh#L228
(The comment linked there is incorrect. hagfish has nothing to do with this. It's hermes that looks for things in /var/spool.) However, if hermes is not configured to upload SDC log files and database backups to a Manta instance, then it also never cleans them up. As a result, database backups pile up and are never cleaned up. It's possible to consume all available disk space in the zone this way.
Former user commented on 2016-06-27T15:16:03.000-0400:
This also applies to hagfish-watcher telemetry files in /var/log/usage on each compute node.
Former user commented on 2018-09-25T15:18:48.586-0400:
This one might relate to work that was being done in AGENT-1089, perhaps.