Add num_member_sups and member_shutdown pool config options#109
Merged
Conversation
Internal refactor: positional tuple destructuring (`{MRef, Status, Time, ExpTs}`)
is replaced by `#member{}' record fields throughout the pool gen_server. Each
call site only references the fields it cares about, and partial updates
preserve the rest of the record automatically — making future field additions
mechanical instead of a fan-out of pattern rewrites.
External API stability: `pool_stats/1' still returns the legacy 4-tuple shape
via the new `member_to_info_tuple/1' helper; the `member_info()' exported type
is unchanged. Tests that pattern-match `{_, free, _, _}` continue to work.
Hot upgrade: bump -vsn 4 → 5; add `do_upgrade_to_v5' that wraps existing
v4 4-tuples in `#member{}` records. The v3 → v4 step now chains into v4 → v5
so upgrades from 1.5.x / 1.6.0 / 1.7.0 all converge on the new record shape.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Each pool's single pooler_pooled_worker_sup serialises all start_child and
terminate_child calls from parallel starters, async stoppers and cull events.
With slow start_mfa (e.g. a monolithic third-party start_link that opens a
network connection) this becomes the bottleneck even when individual workers
are healthy.
Mirrors the approach used by Ranch 2.0 (num_conns_sups): split the single
supervisor into N parallel shards. New starts are distributed round-robin
across shards; each #member{} records its shard_idx so terminations route to
the correct supervisor in O(1). Defaults to 1 — identical to the legacy layout.
Shard 1 always keeps the legacy unsuffixed name (pooler_<pool>_member_sup)
for backward compatibility; additional shards use _2, _3, ... suffixes.
Changes:
- pool_config() gains num_member_sups (pos_integer, default 1)
- #pool{} replaces member_sup with member_sups tuple + next_shard counter
- #member{} gains shard_idx for O(1) termination routing without leaking
supervisor names across every member record
- pooler_pool_sup starts N member supervisors; add_member_sups/4 supports
runtime increase via pool_reconfigure/2 (decrease refused)
- pooler_starter: tagged start/stop spec tuples; new '$pooler_member_sup' and
'$pooler_pool' placeholders; '$pooler_pool_name' kept as deprecated alias
- code_change(4,...) extends do_upgrade_to_v5 to also migrate #pool{} shape
(member_sup -> member_sups 1-tuple, next_shard=1, shard_idx=1 on all members)
- pooler.appup.src: note for future version bump to include
{update, pooler_pool_sup, supervisor}
- Tests: 7 EUnit + 1 PropEr property covering distribution, routing, cull,
and reconfigure scenarios
Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
Replaces the hardcoded brutal_kill in pooler_pooled_worker_sup with a configurable member_shutdown option. Defaults to brutal_kill (unchanged behaviour); a pos_integer() value maps directly to the OTP supervisor child spec Shutdown field, sending exit(Pid, shutdown) and waiting up to that many milliseconds for the worker's terminate/2 to complete. Effective only when the worker has called process_flag(trap_exit, true) in its init/1 — exit(Pid, shutdown) kills a non-trapping process immediately. Ignored by custom stop_mfa that bypasses supervisor:terminate_child/2. Pairs with num_member_sups: stops across different shards proceed in parallel, reducing total teardown time from O(N x timeout) to O(N x timeout / num_member_sups). Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Related to #104
Each pool's single
pooler_pooled_worker_supserialises allstart_child/terminate_childcalls from parallel starters, async stoppers, and cull events. Under load with a slowstart_mfathis becomes the bottleneck even when workers themselves are healthy.Mirrors the approach used by Ranch 2.0: split the supervisor into N parallel shards (
num_member_sups, default 1). New starts are distributed round-robin; each worker records its shard index so terminations route in O(1). Shard 1 keeps the legacy supervisor name for backward compatibility.Also adds
member_shutdownto replace the hardcodedbrutal_kill— allows gracefulterminate/2shutdown. With multiple shards, concurrent graceful stops proceed in parallel, reducing total teardown time fromO(N × timeout)toO(N × timeout / num_member_sups).Preparatory refactor converts all_members tuples to a
#member{}record. Hot-upgrade compatible (code_change v4→v5 handles the full state migration).