| 1 |
Does the Cyrus Murder support High Availability configurations? |
| 2 |
--------------------------------------------------------------- |
| 3 |
|
| 4 |
The :ref:`Cyrus Murder <murder>` clustering solution drastically |
| 5 |
increases system-wide reliability as compared to a traditional |
| 6 |
standalone IMAP server. However, this configuration is not a true High |
| 7 |
Availability/Redundancy solution (nor is such a beast trivial to |
| 8 |
implement given the constraints on traditional Email systems). |
| 9 |
|
| 10 |
The frontends of the CyrusMurder are, in every sense, 100% redundant. No |
| 11 |
server-specific data is maintained on them, and one is as good as any |
| 12 |
other from the point of view of a user. |
| 13 |
|
| 14 |
The mupdate server is indeed a single point of failure, however a |
| 15 |
failure at this point only prohibits message delivery (via the LMTP |
| 16 |
proxies that are querying it for up-to-date mailbox data) and certain |
| 17 |
mailbox operations (create, delete, setacl, and so on). Mail can still |
| 18 |
be accessed by all users of the system, and incoming mail will receive |
| 19 |
temporary failures, so the sending MTA should continue to queue the |
| 20 |
mail. |
| 21 |
|
| 22 |
The failure of a backend will result in a partial system wide failure -- |
| 23 |
the mailboxes that were stored on that backend will no longer be |
| 24 |
available for use by any users. Delivery of new mail will again be |
| 25 |
tempfailed, and queued by the sending MTA. |
| 26 |
|
| 27 |
Some people have suggested shared storage and the use of hot spare |
| 28 |
backends to combat this solution. The Cyrus authors do not generally |
| 29 |
recommend the use of shared storage in the Cyrus environment. However |
| 30 |
people have reported success in various configurations. The shared |
| 31 |
filesystem still needs to support file locking, and it is highly |
| 32 |
recommended that it have an efficient mmap implementation. |
| 33 |
|
| 34 |
|