DRBD notes: Linux platform to achieve high-availability scenarios

DRBD is a block device implementation, mainly being used for the Linux platform of high availability (HA) programmes.

He is a kernel modules and related procedures, through network communications to synchronous mirroring the entire device, somewhat like a network RAID functionality. That is when you write data to the DRBD device on the local file system, the data will also be sent to the network of another host, and in exactly the same form of records in a file system (in fact file system was created by the DRBD synchronously accomplished). Local node (host) and the remote node (host) can guarantee real-time data synchronization, and ensure the consistency of the IO. So when the local node's host fails, the remote node hosts also retains an identical data, you can continue to use, to achieve high availability. In high-availability (HA) solution to use DRBD features, you can instead use a shared disk array storage device. Because data at the same time exist on the local host and the remote host, to meet a need to switch, the remote host to which you only need to use it that backs up data, you can continue to provide services. The underlying device supported DRBD need to build upon the underlying device, and then build a block device. For users, a DRBD device, just like a physical disk, you can create a file system of the geographical position. DRBD supports the underlying device has the following categories: 1, a disk or disk partition of a; 2, a softraid equipment; 3, an LVM logical volume; 4, an EVMS (Enterprise volume management EnterpriseVolumeManagementSystem) volume; 5, any other block devices. Configuration profile 1, the global configuration (global) basically we can do is to configure usage-count is yes or no, usage-count parameter is really just to let the linbit company collects the current use of drbd. When drbd in installation and upgrades when they are sending information via the HTTP protocol to linbit company servers. 2. public configuration items (common) common here, refers to the drbd manages multiple resources in common. Configuration item is configured inside the main drbd all resource can be set to the same parameters, such as protocol, the syncer and so on. 3. resource allocation (resource) resource entry is configured in the management of drbd all resources, including the node's IP information, the underlying storage device name, device size, meta information, drbd device name external offers and so on. Each resource will need to configure each node information, rather than individual information of this node. In fact, the entire cluster drbd, each node above drbd.conf file needs. In addition, there are many other resource internal configuration item: net: network configuration-related content, you can set whether to allow double master (allow-two-primaries), etc. Startup: starts when the relevant settings, such as setting startup who as a primary (or both is primary: become-primary-onbot) syncer: synchronization-related settings. You can set the "back" synchronization (re-synchronization) speed (rate) setting, or you can set whether the online validation data consistency between nodes (verify-alg detection algorithm is md5, sha1, and crc32, etc.). Data validation may be a more important things, open an online checking function, we can pass the related command (drbdadmverifyresource_name) to start the online validation. In the checkout process, drbd will record the inconsistencies between the node block, but does not block any behavior, even in the inconsistent block IO request above. When inconsistent block, drbd requires re-synchronization, syncer rate items inside settings, mainly used for re-synchronization, because if you have a lot of inconsistent data, we may not have all the bandwidth assigned to drbd re-synchronization, so do affect the external references to the provision of services. Rate setting and also need to consider the impact of IO capabilities. If we would have a Gigabit network outlet, but the ability of our disk IO per second and only 50M, then the actual processing capacity is only 50M, generally speaking, the ability to set up a network and disk IO IO capabilities in a minimum of 30% of the bandwidth compared to re-synchronization is appropriate (official). In addition, drbd also provides a temporary rate change commands, you can temporary syncer rate of change: drbdsetup/dev/drbd0syncer-r100M. This temporary setting the speed to 100M re-synchronization. But in the end, re-synchronization you need to let the drbd drbdadmadjustresource_name as configured to work in the rate.