[MCOL-4286] MCS processes should reestablish connections when appropriate Created: 2020-08-31 Updated: 2020-12-02 Resolved: 2020-12-02 |
|
| Status: | Closed |
| Project: | MariaDB ColumnStore |
| Component/s: | DMLProc |
| Affects Version/s: | 1.0.0 |
| Fix Version/s: | 5.5.1 |
| Type: | Task | Priority: | Major |
| Reporter: | Patrick LeBlanc (Inactive) | Assignee: | Jose Rojas (Inactive) |
| Resolution: | Duplicate | Votes: | 0 |
| Labels: | None | ||
| Epic Link: | ColumnStore Failover Improvements |
| Sprint: | 2020-8 |
| Description |
|
Right now network connections aren't re-established in case of connection loss. New cluster management API depends on this capability b/c the services are restarted by systemd in case of a crash. There are different kind of network connections used by network services some of them must be re-established(DMLProc/DDLProc -> WriteEngine) some not (ExeMgr -> PrimProc). |
| Comments |
| Comment by Patrick LeBlanc (Inactive) [ 2020-08-31 ] |
|
This is a clone of mcol-3899 targeting version 1.5.5 instead of 1.5.4. |
| Comment by David Hall (Inactive) [ 2020-10-09 ] |
|
All connections, even ExeMgr<->PrimProc should be re-establised if one process goes down. THe industry standard way of dealing with this is to use heartbeat (or stay-alive) messages. |
| Comment by Jose Rojas (Inactive) [ 2020-12-02 ] |
|
Ticket is duplicate. |