[SERVER-19287] Support unacknowledged operations via OP_COMMAND Created: 06/Jul/15 Updated: 03/Jan/18 Resolved: 02/Oct/17 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | Networking |
| Affects Version/s: | None |
| Fix Version/s: | None |
| Type: | New Feature | Priority: | Major - P3 |
| Reporter: | Andrew Morrow (Inactive) | Assignee: | Backlog - Tuning Team |
| Resolution: | Duplicate | Votes: | 0 |
| Labels: | PM-315 | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Issue Links: |
|
||||||||||||||||
| Participants: | |||||||||||||||||
| Case: | (copied to CRM) | ||||||||||||||||
| Description |
|
Many drivers fall back to legacy writes to perform w:0 inserts since commands are acknowledged. During the 3.2 dev cycle, all operations are becoming available as commands, and we have introduced a new mechanism (OP_COMMAND) for issuing commands. Over the next few server releases, the non-OP_COMMAND mechanisms for communicating with the server will be deprecated and then removed. If we intend to continue to support un-ack'ed operations, we will need to extend OP_COMMAND (or invent a wholly new facility) to support these operations, and drivers will need to be updated to conform. |
| Comments |
| Comment by J Rassi [ 08/Dec/16 ] |
|
Andy, what do you think of David's suggestion of dropping this in favor of |
| Comment by Andy Schwerin [ 08/Dec/16 ] |
|
Yeah, I'm coming around to this idea because of what Spencer mentioned, above. I think an OP_COMMAND_NO_REPLY might actually be appropriate. |
| Comment by Spencer Brody (Inactive) [ 08/Dec/16 ] |
|
Do we expect to ever want to be able to send metadata along with fire-and-forget writes? For instance as part of the lamport clock maintenance? That's the only other reason I could think of for wanting to replace the existing protocol with something new. |