Synchronized Files

Details synchronization methods between the active and passive SMWs.

For files not located on shared storage (boot RAID), the SLEHA Extension software includes the csync2 utility to synchronize (sync) important files between the two SMWs. When a file changes on the active SMW, it is automatically synchronized to the passive SMW. The csync2 utility synchronizes the required files and directories for the SMW HA cluster, such as /etc/passwd and /opt/cray/hss/*/etc/*.

File synchronization is automatically configured during initial installation. The file /etc/csync2/csync2_cray.cfg lists the Cray-specific files and directories that must be synchronized, as well as small files that are convenient to keep in sync.

File synchronization happens in one direction only: from the active SMW to the passive SMW. If a synchronized file changes on the passive SMW, the change will not be propagated to the active SMW in the course of normal operations and could be overwritten on the passive SMW later if there is a subsequent change to the corresponding file on the active SMW. However, if a failover occurs, the previously passive SMW becomes the active SMW. If the change is still in place, the changed file becomes a candidate for propagation to the other SMW (subject to the rules of file conflict resolution).

The fsync resource controls file synchronization operations. Every 100 seconds, fsync checks for files that need to be synchronized.

Very large files are explicitly excluded from synchronization (such as /opt/cray/hss-images/master). The csync2 utility is designed to synchronize small amounts of data. If csync2 must monitor many directories or synchronize a large amount of data, it can become overloaded and failures may not be readily apparent. Cray recommends that sites do not change the list of synchronized files (or add only small files); instead, copy large files and directories manually to the other SMW.