Skip to content

DAOS-19591 container: two fixes for EC aggregation boundary reporting #19284

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
mchaarawi wants to merge 2 commits into from
Closed

DAOS-19591 container: two fixes for EC aggregation boundary reporting #19284

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6e8824a
DAOS-19591 container: two fixes for EC aggregation boundary reporting
liuxuezhao Sep 4, 2026
93ed6e0
DAOS-19591 container: re-send EC agg boundary to ranks waiting for it…
mchaarawi Oct 16, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
40 changes: 35 additions & 5 deletions src/container/srv_container.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1712,10 +1712,12 @@ cont_ec_agg_alloc(struct cont_svc *cont_svc, uuid_t cont_uuid,
ec_agg->ea_servers_num = rank_nr;
ec_agg->ea_current_eph = 0;
ec_agg->ea_rdb_eph = 0;
ec_agg->ea_start_ts = daos_gettime_coarse();
ec_agg->ea_warn_slug_ts = ec_agg->ea_start_ts;
for (i = 0; i < rank_nr; i++) {
ec_agg->ea_server_ephs[i].rank = doms[i].do_comp.co_rank;
ec_agg->ea_server_ephs[i].eph = 0;
ec_agg->ea_server_ephs[i].ee_update_ts = daos_gettime_coarse();
ec_agg->ea_server_ephs[i].ee_update_ts = ec_agg->ea_start_ts;
}
d_list_add(&ec_agg->ea_list, &cont_svc->cs_ec_agg_list);
*ec_aggp = ec_agg;
Expand Down Expand Up @@ -1776,6 +1778,22 @@ ds_cont_leader_update_agg_eph(uuid_t pool_uuid, uuid_t cont_uuid,
if (ec_agg->ea_server_ephs[i].eph < eph) {
ec_agg->ea_server_ephs[i].eph = eph;
ec_agg->ea_server_ephs[i].ee_update_ts = daos_gettime_coarse();
} else if (eph == 0 && ec_agg->ea_current_eph != 0) {
/*
* A rank reports zero only while its EC aggregation is paused
* waiting for the boundary (sc_ec_agg_eph_valid == 0), e.g. a
* reintegrated/extended rank that got the VOS container after the
* last refresh. Ask cont_agg_eph_sync() to re-send the current
* boundary even if the minimum epoch did not change, otherwise
* the rank never starts and pins the boundary forever.
* (ea_current_eph == 0 refreshes every round anyway.)
*/
if (!ec_agg->ea_need_refresh)
D_INFO(DF_CONT ": rank %u waits for EC agg boundary, "
"current " DF_X64 "\n",
DP_CONT(pool_uuid, cont_uuid), rank,
ec_agg->ea_current_eph);
ec_agg->ea_need_refresh = 1;
}
break;
}
Expand Down Expand Up @@ -1850,9 +1868,9 @@ ds_cont_tgt_refresh_agg_eph(uuid_t pool_uuid, uuid_t cont_uuid,
uuid_copy(arg.cont_uuid, cont_uuid);
arg.min_eph = eph;

rc = ds_pool_task_collective(pool_uuid, PO_COMP_ST_NEW | PO_COMP_ST_DOWN |
PO_COMP_ST_DOWNOUT, cont_refresh_vos_agg_eph_one,
&arg, DSS_ULT_FL_PERIODIC);
rc = ds_pool_task_collective(pool_uuid,
PO_COMP_ST_NEW | PO_COMP_ST_DOWN | PO_COMP_ST_DOWNOUT,
cont_refresh_vos_agg_eph_one, &arg, DSS_ULT_FL_PERIODIC);
if (rc) {
DL_ERROR(rc, DF_CONT ": refresh ec_agg_eph " DF_X64 " failed.",
DP_CONT(pool_uuid, cont_uuid), eph);
Expand Down Expand Up @@ -2074,7 +2092,14 @@ cont_agg_eph_sync(struct ds_pool *pool, struct cont_svc *svc)
if (min_eph < ec_agg->ea_rdb_eph)
min_eph = ec_agg->ea_rdb_eph;

if (min_eph == ec_agg->ea_current_eph && ec_agg->ea_current_eph != 0)
/*
* Skip the refresh when nothing changed, unless some rank asked for the
* boundary (see ds_cont_leader_update_agg_eph()). Such a rank has eph 0,
* so min_eph is pinned to ea_rdb_eph == ea_current_eph and it would never
* get the boundary it needs to start reporting.
*/
if (min_eph == ec_agg->ea_current_eph && ec_agg->ea_current_eph != 0 &&
!ec_agg->ea_need_refresh)
continue;

/**
Expand All @@ -2094,6 +2119,7 @@ cont_agg_eph_sync(struct ds_pool *pool, struct cont_svc *svc)
ec_agg->ea_warn_slug_ts = cur_ts;

} else if (cur_eph && new_eph > cur_eph && (new_eph - cur_eph) >= 600 &&
(cur_ts - ec_agg->ea_start_ts) >= (new_eph - cur_eph) &&
(cur_ts - ec_agg->ea_warn_slug_ts) >= 600) {
ec_agg->ea_warn_slug_ts = cur_ts;
D_WARN(DF_CONT ": Sluggish EC boundary reporting. "
Expand All @@ -2120,6 +2146,10 @@ cont_agg_eph_sync(struct ds_pool *pool, struct cont_svc *svc)
}
}

/* clear before the refresh (which yields), so a request arriving meanwhile
* triggers another refresh in the next round.
*/
ec_agg->ea_need_refresh = 0;
rc = cont_iv_ec_agg_eph_refresh(pool->sp_iv_ns, ec_agg->ea_cont_uuid, min_eph,
svc->cs_ec_leader_ephs_req);
if (rc) {
Expand Down
11 changes: 10 additions & 1 deletion src/container/srv_internal.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,11 +69,20 @@ struct cont_ec_agg {
uuid_t ea_cont_uuid;
daos_epoch_t ea_current_eph;
daos_epoch_t ea_rdb_eph;
/* time when this leader started tracking the container. A boundary jump larger
* than the tracking age cannot have been observed by this leader, it comes from
* the stale epoch loaded from rdb at leader step-up, so it is not a real stall.
*/
uint64_t ea_start_ts;
uint64_t ea_warn_slug_ts;
struct ec_eph *ea_server_ephs;
d_list_t ea_list;
int ea_servers_num;
uint32_t ea_deleted : 1;
uint32_t ea_deleted : 1,
/* some rank reported eph 0 (its EC aggregation waits for the boundary), re-send the
* boundary in the next cont_agg_eph_sync() even if the minimum epoch did not change.
*/
ea_need_refresh : 1;
};

/*
Expand Down
Loading