Managing counters in a distributed file system

A non-transitory computer readable storage medium having instructions for causing a system to:

• divide a clustered file system into plural segments;
• independently check each of the plural segments for inconsistency in a corresponding local counter associated with the corresponding segment, wherein the plural segments are associated with the corresponding local counters included in respective distributed computer nodes of the clustered file system;
• allocate the local counter in a particular one of the computer nodes in response to the particular computer node having to modify distributed count information and the distributed count information not being located at the particular computer node;
• use a file system checker to locate and correct an inconsistency in the local counter associated with a particular one of the plural segments while the clustered file system remains online with access available to file system data;
• store non-distributed global count information in a global tree in a global file manager separate from the computer nodes, the non-distributed global count information describing states of the local counters associated with the plural segments, wherein the global tree includes a root node linked to child nodes that describe the states of the respective local counters;
• and
• collapse information of the allocated local counter to the global count information at the global file system manager in response to contention of the allocated local counter no longer existing.

A server, comprising:

• a file system manager including a processor to execute instructions stored in memory to:
• divide a clustered file system having multiple distributed computer nodes into plural segments;
• use a file system checker to independently check each of the plural segments for inconsistencies in distributed local count information provided by local counters in the respective computer nodes of the clustered file system, the check for the inconsistencies being performed while the clustered file system remains online, wherein the file system manager is separate from the computer nodes;
• allocate the local counter in a particular one of the computer nodes in response to the particular computer node having to modify distributed count information and the distributed count information not being located at the particular computer node;
• store non-distributed global count information that describes a state of the local count information of each of the local counters, wherein the global count information is in a global tree of the file system manager, the global tree including a root node linked to child nodes that describe states of the respective local counters;
• and
• collapse information of the allocated local counter to the global count information at the global file system manager in response to contention of the allocated local counter no longer existing.