Skip to content

DAOS-18607 object: increase RPC retry latency #18015

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
Nasf-Fan wants to merge 2 commits into from
Closed

DAOS-18607 object: increase RPC retry latency #18015

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
ab11981
DAOS-18607 object: increase RPC retry latency
Nasf-Fan Apr 15, 2026
d3d854a
Merge branch 'release/2.6' into Nasf-Fan/DAOS-18607_2
mchaarawi Apr 17, 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
11 changes: 10 additions & 1 deletion src/cart/crt_rpc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
/*
* (C) Copyright 2016-2024 Intel Corporation.
* (C) Copyright 2025 Hewlett Packard Enterprise Development LP
* (C) Copyright 2025-2026 Hewlett Packard Enterprise Development LP
*
* SPDX-License-Identifier: BSD-2-Clause-Patent
*/
Expand Down Expand Up @@ -1992,3 +1992,12 @@ crt_req_get_proto_ver(crt_rpc_t *req)
{
return (req->cr_opc & CRT_PROTO_VER_MASK) >> 16;
}

void
crt_rpc_dump(crt_rpc_t *req, const char *msg)
{
struct crt_rpc_priv *rpc_priv;

rpc_priv = container_of(req, struct crt_rpc_priv, crp_pub);
RPC_INFO(rpc_priv, "%s\n", msg);
}
4 changes: 2 additions & 2 deletions src/dtx/dtx_rpc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
/**
* (C) Copyright 2019-2024 Intel Corporation.
* (C) Copyright 2025 Hewlett Packard Enterprise Development LP
* (C) Copyright 2025-2026 Hewlett Packard Enterprise Development LP
*
* SPDX-License-Identifier: BSD-2-Clause-Patent
*/
Expand Down Expand Up @@ -743,7 +743,7 @@ dtx_rpc(struct ds_cont_child *cont,d_list_t *dti_list, struct dtx_entry **dtes,
dca->dca_steps = length;

/* Use helper ULT to handle DTX RPC if there are enough helper XS. */
if (dss_has_enough_helper()) {
if (0 && dss_has_enough_helper()) {
rc = ABT_eventual_create(0, &dca->dca_chore_eventual);
if (rc != ABT_SUCCESS) {
D_ERROR("failed to create eventual: %d\n", rc);
Expand Down
10 changes: 9 additions & 1 deletion src/include/cart/api.h
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
/*
* (C) Copyright 2016-2024 Intel Corporation.
* (C) Copyright 2025 Hewlett Packard Enterprise Development LP
* (C) Copyright 2025-2026 Hewlett Packard Enterprise Development LP
*
* SPDX-License-Identifier: BSD-2-Clause-Patent
*/
Expand Down Expand Up @@ -2286,6 +2286,14 @@ int crt_group_secondary_modify(crt_group_t *grp, d_rank_list_t *sec_ranks,
d_rank_list_t *prim_ranks, crt_group_mod_op_t op,
uint32_t version);

/**
* Dump CaRT RPC information.
*
* \param[in] req The rpc structure.
* \param[in] fmt The message to be logged.
*/
void crt_rpc_dump(crt_rpc_t *req, const char *msg);

/**
* Initialize swim on the specified context index.
*
Expand Down
14 changes: 8 additions & 6 deletions src/object/cli_obj.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1725,7 +1725,7 @@ dc_obj_retry_delay(tse_task_t *task, uint32_t opc, int err, uint32_t *retry_cnt,
/* Randomly delay [1, max_delay - 5] for DER_OVERLOAD_RETRY case. */
if (err == -DER_OVERLOAD_RETRY) {
delay = daos_rpc_rand_delay(timeout_sec) << 20;
} else if (++(*retry_cnt) > 1) {
} else if (++(*retry_cnt) > 1 || obj_is_modification_opc(opc)) {
/* Randomly delay [31 ~ 1023] us if it is not the first retried object RPC. */
delay = (d_rand() | ((1 << 5) - 1)) & ((1 << 10) - 1);
/* Rebuild is being established on the server side, wait a bit longer */
Expand All @@ -1739,16 +1739,18 @@ dc_obj_retry_delay(tse_task_t *task, uint32_t opc, int err, uint32_t *retry_cnt,
delay <<= 8;
break;
case DAOS_OBJ_RPC_CPD:
/* 8 times of the delay for compounded RPC. */
delay <<= 3;
delay <<= (*retry_cnt + 3);
break;
default:
if (obj_is_modification_opc(opc))
delay <<= (*retry_cnt + 1);
else
delay <<= (*retry_cnt - 1);
break;
}

/* Increase delay after multiple times retry. */
if (*retry_cnt >= 5)
delay <<= 1;
if (*retry_cnt > 10 || delay > 3000000)
Copy link
Copy Markdown
Collaborator

@gnailzenh gnailzenh Apr 15, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we can just 10X each time and start from 10us, so we eventually retry after each 10 seconds after 6 tries.
Let's see if it can help first.

delay = 3000000 + ((d_rand() | ((1 << 5) - 1)) & ((1 << 10) - 1));
}
}

Expand Down
8 changes: 6 additions & 2 deletions src/object/srv_obj.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2883,14 +2883,18 @@ ds_obj_tgt_update_handler(crt_rpc_t *rpc)
* Pre-allocate DTX entry for handling resend under such case.
*/
rc = obj_local_rw(rpc, &ioc, dth);
if (rc != 0)
if (rc != 0) {
DL_CDEBUG(
rc == -DER_INPROGRESS || rc == -DER_TX_RESTART || rc == -DER_ALREADY ||
(rc == -DER_EXIST &&
(orw->orw_api_flags & (DAOS_COND_DKEY_INSERT | DAOS_COND_AKEY_INSERT))) ||
(rc == -DER_NONEXIST &&
(orw->orw_api_flags & (DAOS_COND_DKEY_UPDATE | DAOS_COND_AKEY_UPDATE))),
DB_IO, DLOG_ERR, rc, DF_UOID, DP_UOID(orw->orw_oid));
DB_IO, DLOG_ERR, rc, "tgt_update " DF_UOID " with TX " DF_DTI,
DP_UOID(orw->orw_oid), DP_DTI(&orw->orw_dti));
if (unlikely(rc == -DER_AGAIN))
crt_rpc_dump(rpc, "tgt_update");
}

out:
if (dth != NULL)
Expand Down
6 changes: 4 additions & 2 deletions src/vos/vos_dtx.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
/**
* (C) Copyright 2019-2024 Intel Corporation.
* (C) Copyright 2025 Hewlett Packard Enterprise Development LP
* (C) Copyright 2025-2026 Hewlett Packard Enterprise Development LP
* (C) Copyright 2025 Google LLC
*
* SPDX-License-Identifier: BSD-2-Clause-Patent
Expand Down Expand Up @@ -1523,8 +1523,10 @@ vos_dtx_validation(struct dtx_handle *dth)
/* The DTX has been ever aborted. Return -DER_AGAIN to make related client to retry sometime
* later without triggering dtx_abort().
*/
if (dth->dth_aborted || rc == DTX_ST_ABORTED || rc == DTX_ST_ABORTING)
if (dth->dth_aborted || rc == DTX_ST_ABORTED || rc == DTX_ST_ABORTING) {
D_WARN("Current DTX " DF_DTI " is aborted: %d\n", DP_DTI(&dth->dth_xid), rc);
return -DER_AGAIN;
}

return rc > 0 ? 0 : rc;
}
Expand Down
5 changes: 4 additions & 1 deletion src/vos/vos_obj_cache.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,6 @@
/**
* (C) Copyright 2016-2024 Intel Corporation.
* (C) Copyright 2026 Hewlett Packard Enterprise Development LP
*
* SPDX-License-Identifier: BSD-2-Clause-Patent
*/
Expand Down Expand Up @@ -360,8 +361,10 @@ vos_obj_hold(struct daos_lru_cache *occ, struct vos_container *cont,
obj = container_of(lret, struct vos_object, obj_llink);
}

if (obj->obj_zombie)
if (obj->obj_zombie) {
D_WARN("Hit zombie obj " DF_UOID ", need to retry\n", DP_UOID(oid));
D_GOTO(failed, rc = -DER_AGAIN);
}

if (intent == DAOS_INTENT_KILL && !(flags & VOS_OBJ_KILL_DKEY)) {
if (obj != &obj_local) {
Expand Down
Loading