Disaster recovery in a distributed file system
A snapshot is received at the MDS from a backup storage location at a point-in-time subsequent to taking the snapshot, wherein the MDS comprises a map identifying an object store location for a plurality of data objects and a queue identifying individuals of the plurality of data objects that only reside on the transient object store. Individuals of the plurality of data objects identified in the queue are located on the distributed file system. The map is updated to identify the object store location for the individuals of the plurality of data objects identified in the queue.
1. A method for writing a data object in a distributed file system, the method comprising:
transmitting a write request associated with the data object from a client to a metadata server;
receiving a site cache location at the client from the metadata server;
transmitting the data object to the site cache such that the data object is stored in a memory at the site cache;
maintaining a location of the data object in a task queue at the metadata server; and
flushing the data object, by the metadata server, to an authoritative object store by:
transmitting a write request by the site cache to the metadata server;
receiving an authoritative object store location from the metadata sever; and
transmitting the data object to the authoritative object store.
2. The method of claim 1 , further comprising managing the task queue by the metadata server.
3. The method of claim 1 , further comprising updating the metadata server after flushing the data object to an authoritative object store.
4. The method of claim 1 , further comprising determining that the metadata server requires repair.
5. The method of claim 4 , further comprising restoring the metadata server from a backup and correcting location of data objects identified in the snapshot that are incorrect in the restored metadata server.
6. The method of claim 5 , wherein the backup corresponds to a point in time prior to flushing the data object from the site cache.
7. The method of claim 6 , wherein the task queue of the restored metadata server indicates that data objects that have been moved to the authoritative object store are stored at the site cache.
8. The method of claim 7 , further comprising correcting the locations of the data objects in the task queue of the restored metadata server.
9. The method of claim 8 , further comprising locating the data objects identified in the task queue and updating the restored metadata server such that locations of the data objects identified in the task queue are correct.
10. The method of claim 1 , further comprising taking a backup of the metadata server, the backup including a snapshot.
11. A non-transitory storage medium comprising computer executable instructions for performing operations, the operations including:
transmitting a write request associated with the data object from a client to a metadata server;
receiving a site cache location at the client from the metadata server;
transmitting the data object to the site cache such that the data object is stored in a memory at the site cache;
maintaining a location of the data object in a task queue at the metadata server; and
flushing the data object, by the metadata server, to an authoritative object store by:
transmitting a write request by the site cache to the metadata server;
receiving an authoritative object store location from the metadata sever; and
transmitting the data object to the authoritative object store.
12. The non-transitory storage of claim 11 , the operations further comprising managing the task queue by the metadata server.
13. T The non-transitory storage of claim 11 , the operations further comprising updating the metadata server after flushing the data object to an authoritative object store.
14. The non-transitory storage of claim 11 , the operations further comprising determining that the metadata server requires repair.
15. The non-transitory storage of claim 14 , the operations further comprising restoring the metadata server from a backup and correcting location of data objects identified in the snapshot that are incorrect in the restored metadata server.
16. The non-transitory storage of claim 15 , wherein the backup corresponds to a point in time prior to flushing the data object from the site cache.
17. The non-transitory storage of claim 16 , wherein the task queue of the restored metadata server indicates that data objects that have been moved to the authoritative object store are stored at the site cache.
18. The non-transitory storage of claim 17 , the operations further comprising correcting the locations of the data objects in the task queue of the restored metadata server.
19. The non-transitory storage of claim 18 , the operations further comprising locating the data objects identified in the task queue and updating the restored metadata server such that locations of the data objects identified in the task queue are correct.
20. The non-transitory storage of claim 11 , the operations further comprising taking a backup of the metadata server, the backup including a snapshot.