Consistent snapshot points in a distributed storage service

A system, comprising:

• one or more hardware processors with associated memory that implement a distributed storage service that includes a storage subsystem, a plurality of access nodes, and a snapshot coordinator,
• wherein the plurality of access nodes is configured to issue client writes to data stored as a file system in the storage subsystem in response to requests from clients of the distributed storage service, and
• wherein the snapshot coordinator is configured to:
• determine to create a snapshot of the data stored in the storage subsystem;
• cause the access nodes to pause issuing the client writes to the data in the storage subsystem to identify a point in time during the pause to establish as a consistent snapshot point of the data to be snapshotted;
• set a preserve flag in a superblock of the file system to change a write mode of the file system to a copy-on-write mode, wherein the setting of the preserve flag causes the access nodes to resume issuing client writes to the file system as copy-on-write operations;
• during the issuing of the client writes as copy-on-write operations, cause the snapshot of the data to be generated as of the consistent snapshot point;
• and
• subsequent to a determination that the snapshot is completed, reset the preserve flag in the superblock to stop the file system from being written in the copy-on-write mode, wherein the resetting of the preserve flag causes the access nodes to stop issuing further client writes to the file system as copy-on-write operations.

One or more non-transitory computer-readable storage media storing program instructions that when executed on or across one or more processors cause the one or more processors to implement a snapshot coordinator for a distributed storage service and cause the snapshot coordinator to:

• determine to create a snapshot of data stored by a storage subsystem of the distributed storage service, wherein the data is stored in a file system and written by a plurality of access nodes of the distributed storage service that issues client writes to the data in response to requests from clients of the distributed storage service;
• cause the access nodes to pause issuing the client writes to the data in the storage subsystem to identify a point in time during the pause to establish as a consistent snapshot point of the data to be snapshotted;
• set a preserve flag in a superblock of the file system to change a write mode of the file system to a copy-on-write mode, wherein the setting of the preserve flag causes the access nodes to resume issuing client writes to the file system as copy-on-write operations;
• during the issuing of the client writes as copy-on-write operations, cause the snapshot of the data to be generated as of the consistent snapshot point;
• and
• subsequent to a determination that the snapshot is completed, reset the preserve flag in the superblock to stop the file system from being written in the copy-on-write mode, wherein the resetting of the preserve flag causes the access nodes to stop issuing further client writes to the file system as copy-on-write operations.