Minimizing cluster freeze snapshot in scale out solution

A non-transitory computer readable medium (CRM) comprising computer readable program code, which when executed by a computer processor enables the computer processor to perform a method for creating a snapshot, the method comprising:

• receiving, at a first node, a first request for a snapshot of a partition of a distributed system;
• requesting, by the first node, an exclusive lock mode in the first node with a distributed lock manager (DLM) server;
• making a first determination, in response to the requesting an exclusive lock mode, identifying which nodes in a distributed system are on the partition and in a shared lock mode using the DLM server;
• sending, from the DLM server, a revoke request to the identified nodes based on the first determination;
• flushing, in response to the revoke request, dirty data on the identified nodes to a remote storage system;
• after flushing the dirty data:
• revoking the shared lock mode from the identified nodes;
• granting the exclusive lock mode to the first node after revoking the shared lock mode from the identified nodes;
• and
• creating, by the first node, the snapshot of the partition.

A method for creating a snapshot, comprising:

• receiving, at a first node, a first request for a snapshot of a partition on a distributed system, wherein the first node and a second node are in a shared global flush lock and a shared partition lock;
• requesting, in response to the first request for a snapshot, by the first node, an exclusive global flush lock with a distributed lock manager (DLM) server;
• after requesting the exclusive global flush lock, beginning an asynchronous flush in the first node, wherein the asynchronous flush is performed in a shared global flush end lock state;
• sending, in response to requesting the exclusive global flush lock, a global flush lock revoke request to the second node from the DLM server;
• beginning, in response to the global flush lock revoke request, a second asynchronous flush in the second node, wherein the second asynchronous flush is performed in the shared global flush end lock state;
• after beginning the second asynchronous flush:
• receiving, at the first node, the exclusive global flush lock from the DLM server, wherein the first node downgrades the exclusive global flush lock to shared mode to allow modifications to data on the partition;
• making a first determination that that first node has completed the asynchronous flush based on the first node leaving the shared global flush end lock state;
• requesting, based on the first determination, an exclusive global flush end lock for the first node;
• sending, in response to requesting the exclusive global flush end lock, a second revoke request for a shared global flush end lock from the DLM server;
• confirming, in response to the second revoke request, the second node has completed the second asynchronous flush based on the second node leaving the shared global flush end lock state;
• receiving, based on the confirming, the exclusive global flush end lock at the first node;
• requesting, by the first node, an exclusive partition lock;
• in response to requesting the exclusive partition lock:
• revoking, with the DLM server, the shared partition lock from the second node;
• after revoking the shared partition lock from the second node, flushing any dirty metadata and dirty data in the second node;
• after flushing the second node, granting, with the DLM server, an exclusive partition lock to the first node;
• after granting the exclusive partition lock, flushing any dirty metadata and dirty data in the first node;
• performing, after flushing the first node, the snapshot by the first node;
• and
• releasing, in response to the snapshot, the exclusive partition lock.