Minor fixes and improvements for progress tracking #311
Conversation
mason-sharp
commented
Jan 12, 2026
mason-sharp
commented
- Refactor apply_worker_get_progress() to return only the timestamp needed. Return value instead of pointer to entire struct.
- Lock sooner in spock_group_resource_dump() because of getting the number of entries
- Properly handle progress updates after COPY operations in replication set data sync and avoid potential memory leak
- Address compiler warnings
danolivo
left a comment
danolivo
left a comment
There was a problem hiding this comment.
It looks correct. Minor fixes should be addressed before the merge.
src/spock_group.c
Outdated
| elog(LOG, "SPOCK: adjust spock.progress %d->%d to " | ||
| "remote_commit_ts='%s' " | ||
| "remote_commit_lsn=%llX remote_insert_lsn=%llX", | ||
| "remote_commit_lsn=%lX remote_insert_lsn=%lX", |
There was a problem hiding this comment.
Both variants are incorrect. Conventional way: use the LSN_FORMAT_ARGS macros instead.
| } | ||
|
|
||
| adjust_progress_info(origin_conn); | ||
| progress_entries_list = adjust_progress_info(origin_conn); |
There was a problem hiding this comment.
Good catch! But call:
progress_entries_list = adjust_progress_info(origin_conn);
should be done right before the ensure_replication_slot_snapshot().
There was a problem hiding this comment.
After fixing this, you may launch the following workflow on your branch: https://github.com/pgEdge/spock/actions/runs/21101567363
It fails now in 100% of cases. With this case, it should pass (about 99% or so).
- Refactor apply_worker_get_progress() to return only the timestamp needed. Return value instead of pointer to entire struct. - Lock sooner in spock_group_resource_dump() because of getting the number of entries - Properly handle progress updates after COPY operations in replication set data sync and avoid potential memory leak - Address compiler warnings
mason-sharp
force-pushed
the
progress-track-minor-improvements
branch
from
679df29 to
a78e2cf
Compare
| origin_conn_repl = spock_connect_replica(sub->origin_if->dsn, | ||
| sub->name, "snap"); | ||
|
|
||
| adjust_progress_info(origin_conn); |
There was a problem hiding this comment.
How is it supposed to work?
I don't see here following spock_group_progress_update_list() ...
