gsadmin/rustfs
1
0
Fork 0
mirror of https://github.com/rustfs/rustfs.git synced 2026-07-26 08:18:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(cluster): clarify peer health and listing timeouts (#5086)

Browse Source 
* fix(cluster): surface observed peer health

Refs rustfs/backlog#1387

Co-Authored-By: heihutu <heihutu@gmail.com>

* fix(admin): distinguish unreported peer health

Refs rustfs/backlog#1388

Co-Authored-By: heihutu <heihutu@gmail.com>

* fix(ecstore): decouple metacache peek timeout

Refs rustfs/backlog#1389

Co-Authored-By: heihutu <heihutu@gmail.com>

* docs(ops): diagnose metacache listing timeouts

Refs rustfs/backlog#1390

Co-Authored-By: heihutu <heihutu@gmail.com>

* refactor(cluster): clarify observed peer health

Co-Authored-By: heihutu <heihutu@gmail.com>

* fix(ecstore): route capability state through contract

Co-Authored-By: heihutu <heihutu@gmail.com>

---------

Co-authored-by: heihutu <heihutu@gmail.com>
This commit is contained in:
houseme
2026-07-21 19:11:24 +08:00
committed by GitHub
parent bb7bba3237
commit 7805cf5ae6
10 changed files with 349 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/operations/drive-timeout-tuning.md
+5
View File
@@ -56,6 +56,7 @@ Explicit per-operation overrides always win over the profile, so you can select
| Environment variable | Default | `high_latency` default | Bounds |
|---|---:|---:|---|
| `RUSTFS_DRIVE_WALKDIR_STALL_TIMEOUT_SECS` | `5` | `60` | Max time a single walk filesystem call may go without answering during a listing walk. **The knob for listing failures on large prefixes.** |
| `RUSTFS_DRIVE_WALKDIR_PEEK_TIMEOUT_SECS` | `10` | `120` | Max time the metacache merge consumer waits for the next visible entry from a walk reader. Values below the resolved stall timeout are clamped up to the stall timeout. |
| `RUSTFS_DRIVE_WALKDIR_TIMEOUT_SECS` | `5` | `60` | Total wall-clock timeout for a `walk_dir`. Retained for non-foreground callers; the foreground listing path no longer uses it (see below). |
| `RUSTFS_DRIVE_LIST_DIR_TIMEOUT_SECS` | `5` | `60` | Timeout for a standalone `list_dir` metadata listing. |
| `RUSTFS_DRIVE_METADATA_TIMEOUT_SECS` | `5` | `60` | Timeout for metadata reads such as `read_metadata`. |
@@ -118,6 +119,10 @@ Raise the walk stall budget, or select the high-latency profile:
# widen just the listing walk budget
-e RUSTFS_DRIVE_WALKDIR_STALL_TIMEOUT_SECS=60
# widen the metacache reader wait budget when drives keep progressing but do
# not publish visible entries quickly enough for the merge consumer
-e RUSTFS_DRIVE_WALKDIR_PEEK_TIMEOUT_SECS=120
# or raise every profile-aware drive default at once
-e RUSTFS_DRIVE_TIMEOUT_PROFILE=high_latency
```