6b130c6422
fix(dsraft): preserve pending replica set transitions
...
Otherwise, information about pending replica set transitions taking a
long time to complete could be lost on subsequent target set changes and
node crashes.
2024-07-17 12:17:07 +02:00
af81800aec
chore(dsraft): log a bit more informative messages in shard allocator
2024-07-12 18:24:58 +02:00
8e8b382ec0
chore(dsraft): provide more details when replica is unready
2024-07-12 18:23:23 +02:00
70a760850f
chore(dsraft): correct comment spelling errors
2024-07-12 15:27:29 +02:00
3d296abde9
fix(dsrepl): classify ra error conditions more carefully
...
Most importantly: avoid automatic retries of `shutdown` and `nodedown`
errors as this could easily lead to Raft log entries duplication.
2024-06-26 20:44:34 +02:00
8ff48ac5ea
feat(dsrepl): rely on accumulated log size to decide when to flush
2024-06-26 20:44:34 +02:00
7895e9cc45
feat(dsstore): make WAL-less mode optional
...
And make the upper layer choose when to use it.
2024-06-26 20:44:32 +02:00
0c0757b8c2
feat(dsrepl): enable WAL-less batch writes
2024-06-26 20:44:32 +02:00
2705226eb5
feat(dsrepl): release log entries occasionally
...
Also make tracepoints in `apply/3` callback implementation more
uniform.
2024-06-26 20:44:32 +02:00
b6894c18fa
chore(dsrepl): improve tracepoints usability a bit
2024-06-26 20:44:32 +02:00
9a58d71378
fix(ds): Move DS backend supervision trees to their own apps
2024-06-21 18:22:04 +02:00
3851fc189f
fix(ds): Avoid reverse dependencies from storage to repl. layer
2024-06-21 17:16:24 +02:00
09c3ae795d
refactor(ds_raft): Replace egress server with common emqx_ds_buffer
2024-06-21 17:16:24 +02:00
a8ea0ae4e5
refactor(ds): Extract DS replication layer to a separate application
2024-06-21 17:16:24 +02:00
Andrew Mayorov