Skip to content
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

DAOS-17111 cart: Use only swim ctx for outage #15924

Draft
wants to merge 1 commit into
base: master
Choose a base branch
from
Draft

DAOS-17111 cart: Use only swim ctx for outage #15924

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
4 changes: 2 additions & 2 deletions src/cart/crt_hg_proc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,6 @@
/*
* (C) Copyright 2016-2023 Intel Corporation.
* (C) Copyright 2025 Hewlett Packard Enterprise Development LP
*
* SPDX-License-Identifier: BSD-2-Clause-Patent
*/
Expand Down Expand Up @@ -450,8 +451,7 @@ crt_hg_unpack_header(hg_handle_t handle, struct crt_rpc_priv *rpc_priv,

/* Sync the HLC. Clients never decode requests. */
D_ASSERT(crt_is_service());
rc = d_hlc_get_msg(rpc_priv->crp_req_hdr.cch_hlc,
&ctx->cc_last_unpack_hlc, &clock_offset);
rc = d_hlc_get_msg(rpc_priv->crp_req_hdr.cch_hlc, NULL /* hlc_out */, &clock_offset);
if (rc != 0) {
REPORT_HLC_SYNC_ERR("failed to sync HLC for request: opc=%x ts="
DF_U64" offset="DF_U64" from=%u\n",
Expand Down
2 changes: 0 additions & 2 deletions src/cart/crt_internal_types.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -413,8 +413,6 @@ struct crt_context {

/** timeout per-context */
uint32_t cc_timeout_sec;
/** HLC time of last received RPC */
uint64_t cc_last_unpack_hlc;

/** Per-context statistics (server-side only) */
/** Total number of timed out requests, of type counter */
Expand Down
43 changes: 21 additions & 22 deletions src/cart/crt_swim.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,6 @@
/*
* (C) Copyright 2019-2024 Intel Corporation.
* (C) Copyright 2025 Hewlett Packard Enterprise Development LP
*
* SPDX-License-Identifier: BSD-2-Clause-Patent
*/
Expand Down Expand Up @@ -376,6 +377,14 @@ crt_swim_lookup_id(swim_id_t id)
return grp_ver;
}

static void crt_swim_update_last_unpack_hlc(struct crt_swim_membs *csm, uint64_t hlc)
{
crt_swim_csm_lock(csm);
if (csm->csm_last_unpack_hlc < hlc)
csm->csm_last_unpack_hlc = hlc;
Copy link
Contributor

Choose a reason for hiding this comment

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

csm->csm_last_unpack_hlc = MAX(csm->csm_last_unpack_hlc, hlc):

Copy link
Contributor Author

Choose a reason for hiding this comment

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

If the current style does not have any real problem, I'd like to keep it as it is. Also, seems like the current style is more consistent with similar code in the rest of the file.

Copy link
Contributor

Choose a reason for hiding this comment

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

it's just better code. After this PR https://github.com/daos-stack/daos/pull/15929/files is merged, I believe most of places will be fixed.

crt_swim_csm_unlock(csm);
}

static void crt_swim_srv_cb(crt_rpc_t *rpc)
{
struct crt_rpc_priv *rpc_priv = container_of(rpc, struct crt_rpc_priv, crp_pub);
Expand All @@ -396,6 +405,8 @@ static void crt_swim_srv_cb(crt_rpc_t *rpc)

D_ASSERT(crt_is_service());

crt_swim_update_last_unpack_hlc(csm, hlc);

from_id = rpc_priv->crp_req_hdr.cch_src_rank;

/* Initialize empty array in case of error in reply */
Expand Down Expand Up @@ -605,6 +616,14 @@ static void crt_swim_cli_cb(const struct crt_cb_info *cb_info)
if (to_id == ctx->sc_target)
ctx->sc_deadline = 0;
swim_ctx_unlock(ctx);
} else {
struct crt_swim_membs *csm = &grp_priv->gp_membs_swim;

/*
* Although some errors also suggest incoming messages, we keep
* it simple for now.
*/
crt_swim_update_last_unpack_hlc(csm, hlc);
}

reply_rc = cb_info->cci_rc ? cb_info->cci_rc : rpc_out->rc;
Expand Down Expand Up @@ -987,24 +1006,6 @@ static void crt_swim_new_incarnation(struct swim_context *ctx,
state->sms_incarnation = incarnation;
}

static void crt_swim_update_last_unpack_hlc(struct crt_swim_membs *csm)
{
struct crt_context *ctx = NULL;
d_list_t *ctx_list;

D_RWLOCK_RDLOCK(&crt_gdata.cg_rwlock);

ctx_list = crt_provider_get_ctx_list(true, crt_gdata.cg_primary_prov);
d_list_for_each_entry(ctx, ctx_list, cc_link) {
uint64_t hlc = ctx->cc_last_unpack_hlc;

if (csm->csm_last_unpack_hlc < hlc)
csm->csm_last_unpack_hlc = hlc;
}

D_RWLOCK_UNLOCK(&crt_gdata.cg_rwlock);
}

static void
crt_metrics_sample_delay(crt_context_t crt_ctx, uint64_t delay, bool glitch)
{
Expand Down Expand Up @@ -1048,10 +1049,8 @@ static int64_t crt_swim_progress_cb(crt_context_t crt_ctx, int64_t timeout_us, v
} else if (rc == -DER_TIMEDOUT || rc == -DER_CANCELED) {
uint64_t now = swim_now_ms();

crt_swim_update_last_unpack_hlc(csm);

/*
* Check for network idle in all contexts.
* Check for network idle in swim context.
* If the time passed from last received RPC till now is more
* than 2/3 of suspicion timeout suspends eviction.
* The max_delay should be less suspicion timeout to guarantee
Expand All @@ -1064,7 +1063,7 @@ static int64_t crt_swim_progress_cb(crt_context_t crt_ctx, int64_t timeout_us, v
uint64_t max_delay = swim_suspect_timeout_get() * 2 / 3;

if (delay > max_delay) {
D_ERROR("Network outage detected (idle during "
D_ERROR("SWIM network outage detected (idle during "
Copy link
Contributor

Choose a reason for hiding this comment

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

I tend to think we should never update csm->csm_last_unpack_hlc in this code block, since that variable is the ground truth to measure how long the current node has been isolated. If we need to calculate progressive delay, let's introduce another variable for that.

if (csm->csm_alive_count > 2) {

Do you know what's the purpose of this check?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I don't know about csm_alive_count.

Agreed on "should never update csm_last_unpack_hlc". But I wonder if we just remove the update below, would this error be printed too frequently? Perhaps this should be addressed in next PR that changes the handling of outages?

Copy link
Contributor Author

@liw liw Feb 19, 2025 •
edited
Loading

Choose a reason for hiding this comment

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

It seems that csm_alive_count intends to track the number of alive members. (This tracking might have been broken by an old change of mine. Actually, I think it was problematic even before my old change. I'll work on a fix.) The number must be > 2 here perhaps because if there's only one alive member, that is, myself, then there's no use in detecting outage, whereas if there are two alive members, then we need to detect the death of the peer?

Copy link
Contributor

@jxiong jxiong Feb 19, 2025 •
edited
Loading

Choose a reason for hiding this comment

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

yeah I understand what it tracks, but I don't know why this is required, like why it is critical to have at least 2 alive members.

In my local fix, I just removed this check and it fixes all the issues for me in my 3-node cluster. I tend to just remove it.

"%lu.%lu sec > expected %lu.%lu sec).\n",
delay / 1000, delay % 1000,
max_delay / 1000, max_delay % 1000);
Expand Down