gsadmin/rustfs
1
0
Fork 0
mirror of https://github.com/rustfs/rustfs.git synced 2026-08-31 01:09:23 +00:00
Code Issues Packages Projects Releases Wiki Activity

Expose replication backlog gauges (#5557)

Browse Source 
* fix(replication): count backlog at queue admission

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

* feat(obs): expose bucket replication backlog gauges

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

* fix(obs): report recent backlog from queued work

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

* fix(obs): preserve legacy backlog metric semantics

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

* feat(obs): expose durable MRF backlog gauges

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

* test(obs): cover replication backlog metric scope

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

* fix(obs): keep backlog metrics API-compatible

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

* refactor(obs): streamline replication backlog metrics

Keep MRF backlog accounting and OBS metric collection on a single, cheaper path.

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

* test(kms): update aws capability snapshot

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

---------

Co-authored-by: heihutu <heihutu@gmail.com>
This commit is contained in:
houseme
2026-08-01 16:48:45 +08:00
committed by GitHub
parent 8218248000
commit 62d44d10b8
13 changed files with 1025 additions and 244 deletions
Download Patch File Download Diff File Expand all files Collapse all files
crates/obs/src/metrics/schema/bucket_replication.rs
+50
View File
@@ -33,6 +33,11 @@ const RESYNC_COMPLETED_TOTAL: &str = "resync_completed_total";
const RESYNC_FAILED_TOTAL: &str = "resync_failed_total";
const RESYNC_CANCELED_TOTAL: &str = "resync_canceled_total";
const RESYNC_DURATION_MS_TOTAL: &str = "resync_duration_ms_total";
const CURRENT_BACKLOG_COUNT: &str = "current_backlog_count";
const CURRENT_BACKLOG_BYTES: &str = "current_backlog_bytes";
const DURABLE_MRF_AVAILABLE: &str = "durable_mrf_available";
const DURABLE_MRF_BACKLOG_COUNT: &str = "durable_mrf_backlog_count";
const DURABLE_MRF_BACKLOG_BYTES: &str = "durable_mrf_backlog_bytes";
pub static BUCKET_REPL_LAST_HR_FAILED_BYTES_MD: LazyLock<MetricDescriptor> = LazyLock::new(|| {
new_gauge_md(
@@ -79,6 +84,51 @@ pub static BUCKET_REPL_LATENCY_MS_MD: LazyLock<MetricDescriptor> = LazyLock::new
)
});
pub static BUCKET_REPL_CURRENT_BACKLOG_BYTES_MD: LazyLock<MetricDescriptor> = LazyLock::new(|| {
new_gauge_md(
MetricName::from(CURRENT_BACKLOG_BYTES),
"Current number of bytes admitted to the in-memory replication worker queues for a bucket",
&[BUCKET_L],
subsystems::BUCKET_REPLICATION,
)
});
pub static BUCKET_REPL_CURRENT_BACKLOG_COUNT_MD: LazyLock<MetricDescriptor> = LazyLock::new(|| {
new_gauge_md(
MetricName::from(CURRENT_BACKLOG_COUNT),
"Current number of objects admitted to the in-memory replication worker queues for a bucket",
&[BUCKET_L],
subsystems::BUCKET_REPLICATION,
)
});
pub static BUCKET_REPL_DURABLE_MRF_AVAILABLE_MD: LazyLock<MetricDescriptor> = LazyLock::new(|| {
new_gauge_md(
MetricName::from(DURABLE_MRF_AVAILABLE),
"Whether the durable MRF backlog snapshot is available for this bucket on this node",
&[BUCKET_L],
subsystems::BUCKET_REPLICATION,
)
});
pub static BUCKET_REPL_DURABLE_MRF_BACKLOG_COUNT_MD: LazyLock<MetricDescriptor> = LazyLock::new(|| {
new_gauge_md(
MetricName::from(DURABLE_MRF_BACKLOG_COUNT),
"Current number of objects in the durable MRF backlog file for a bucket on this node",
&[BUCKET_L],
subsystems::BUCKET_REPLICATION,
)
});
pub static BUCKET_REPL_DURABLE_MRF_BACKLOG_BYTES_MD: LazyLock<MetricDescriptor> = LazyLock::new(|| {
new_gauge_md(
MetricName::from(DURABLE_MRF_BACKLOG_BYTES),
"Current bytes in the durable MRF backlog file for a bucket on this node",
&[BUCKET_L],
subsystems::BUCKET_REPLICATION,
)
});
pub static BUCKET_REPL_PROXIED_DELETE_TAGGING_REQUESTS_TOTAL_MD: LazyLock<MetricDescriptor> = LazyLock::new(|| {
new_counter_md(
MetricName::ProxiedDeleteTaggingRequestsTotal,
crates/obs/src/metrics/schema/replication.rs
+1 -1
View File
@@ -128,7 +128,7 @@ pub static REPLICATION_MAX_DATA_TRANSFER_RATE_MD: LazyLock<MetricDescriptor> = L
pub static REPLICATION_RECENT_BACKLOG_COUNT_MD: LazyLock<MetricDescriptor> = LazyLock::new(|| {
new_gauge_md(
MetricName::ReplicationRecentBacklogCount,
"Total number of objects currently in replication backlog (failed plus queued)",
"Legacy replication backlog indicator: failed target objects plus objects currently queued on this node",
&[],
subsystems::REPLICATION,
)