Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
>Let's assume that I would like use only manual way for recovering after >split brain (this is what did you say - I should choose more up to date node >and synchronize data with it). It is ok for me. But are there any good >practices in case of disconnecting two primary nodes? I mean for example: >killing node, switching node to read-only mode and so on? I would like to >know how it should be behave in case of disconnecting nodes. I suspect if you read the man page on the conf file, you'll answer your own question and glean some needed insight into how drbd _can_ behave. If one node dies and the other remains/becomes primary, you can control the startup behavior to prevent any issues. If both nodes go primary and receive changes to their data set, you can also configure how it behaves when connectivity is reestablished. There exist quit a few scenarios largely dependent on your configuration in terms of dual pri or not etc. Read that conf file a few times, its enlightening on what you come to understand. jlc