This is an attempt to document the cyrus database formats. It should not be considered authoritative and is subject to change.
No external tools should make use of this information. The only supported method of access to the databases is via the API specific to the database and the underlying cyrusdb interface.
This database contains the master list of all mailboxes on the system. The database is indexed by mailbox name and each data record contains the mailbox type, the partition on which the mailbox resides and the ACL on the mailbox. The format of each record is as follows:
Key: <Mailbox Name> Data: <Type Number>SP<Partition>SP<ACL (space-separated userid/rights pairs)>
This database contains mailbox and server annotations. The database is indexed by mailbox name (empty for server annotations) + annotation name + userid (empty for shared annotations) and each data record contains the value size, value data, content-type of the data and timestamp of the record. The format is each record is as follows:
Key: <Mailbox Name>\0<Annotation Name>\0<Userid>\0 Data: <Value Size (4 bytes)><Value>\0<Content-Type>\0<Timestamp (4 bytes)>
This database contains the master list of quotaroots on the system. The database is indexed by quota root and each data record contains the current usage of all mailboxes under the quota root and the limit of the quota root. The format of each record is as follows:
Key: <Quota Root> Data: <Usage (in bytes)>SP<Limit (in Kbytes)>
The legacy quota database uses a distributed system in which each quota root is stored in a separate file named by quota root and the contents has the following format:
<Usage (in bytes)>\n <Limit (in Kbytes)>\n
The translation to/from this data record format is handled by the quota_legacy cyrusdb backend.
This database is used for duplicate delivery suppression, retrieving usenet articles by message-id, and tracking Sieve redirects and vacation responses. The database is indexed by message-id + recipient (either mailbox or email address) and each data record contains the timestamp of the record and the UID of the message within the mailbox (if delivered locally). The format of each record is as follows:
Key: <Message-ID>\0<Recipient>\0 Data: <Timestamp (4 bytes)><Message UID (4 bytes)>
This database caches SSL/TLS sessions so that subsequent connections using the same session-id can bypass the SSL/TLS handshaking, resulting is shorter connection times. The database is indexed by session-id and each data record contains the timestamp of the record and the ASN1 representation of the session data. The format of each record is as follows:
Key: <Session-ID (multi-byte)> Data: <Timestamp (4 bytes)><Session Data (multi-byte)>
This database caches authentication state records, resulting in shorter authentication/canonicalization times. The database is indexed by userid and each data record contains an authentication state for the userid. The format of each record is as follows:
Key: <Userid> Data: <Auth State (multi-byte)>
Key: <Mailbox UID> Data: <Version>SP<Last Read Time>SP<Last Read UID>SP<Last Change Time>SP<List of Read UIDs>
This database is a per-user database and contains the list of mailboxes to which the user has subscribed. The database is indexed by mailbox name and each data record contains no data. The format of each record is follows:
Key: <Mailbox Name> Data: None