fix(scanner): fence timed out scan cycles

This commit is contained in:
马登山
2026-08-22 04:43:48 +08:00
parent bc07cfd115
commit 54b77a18b7
12 changed files with 749 additions and 71 deletions
+84 -22
View File
@@ -125,7 +125,10 @@ impl Default for ScannerRuntimeConfig {
cycle_interval_source: ScannerRuntimeConfigSource::Default,
bitrot_cycle: Some(Duration::from_secs(DEFAULT_HEAL_BITROT_CYCLE_SECS)),
bitrot_cycle_source: ScannerRuntimeConfigSource::Default,
cycle_budget: ScannerCycleBudgetConfig::default(),
cycle_budget: ScannerCycleBudgetConfig {
max_duration: Some(Duration::from_secs(DEFAULT_SCANNER_CYCLE_MAX_DURATION_SECS)),
..Default::default()
},
cycle_max_duration_source: ScannerRuntimeConfigSource::Default,
cycle_max_objects_source: ScannerRuntimeConfigSource::Default,
cycle_max_directories_source: ScannerRuntimeConfigSource::Default,
@@ -436,19 +439,46 @@ fn lookup_max_wait(
Ok((speed.max_sleep(), speed_source))
}
fn lookup_optional_seconds(
kvs: Option<&KVS>,
key: &'static str,
env_key: &'static str,
default: u64,
) -> Result<(Option<Duration>, ScannerRuntimeConfigSource), ScannerRuntimeConfigError> {
if let Some(secs) = rustfs_utils::get_env_opt_u64(env_key) {
return Ok((Some(Duration::from_secs(secs)), ScannerRuntimeConfigSource::Env));
fn lookup_cycle_duration(kvs: Option<&KVS>) -> Result<(Option<Duration>, ScannerRuntimeConfigSource), ScannerRuntimeConfigError> {
match rustfs_utils::get_env_parse_outcome::<u64>(ENV_SCANNER_CYCLE_MAX_DURATION_SECS) {
rustfs_utils::EnvParseOutcome::Parsed(secs) => {
return cycle_duration_from_secs(ENV_SCANNER_CYCLE_MAX_DURATION_SECS, secs)
.map(|duration| (duration, ScannerRuntimeConfigSource::Env));
}
rustfs_utils::EnvParseOutcome::Invalid => {
// Do not include the raw environment value in the typed error:
// deployments occasionally put sensitive material in inherited
// environment snapshots. The key still identifies the control.
return Err(invalid_value(
ENV_SCANNER_CYCLE_MAX_DURATION_SECS,
"<invalid>",
"expected unsigned integer seconds",
));
}
rustfs_utils::EnvParseOutcome::Absent => {}
}
if let Some(value) = config_value(kvs, key, default) {
return parse_config_u64(key, value).map(|secs| (Some(Duration::from_secs(secs)), ScannerRuntimeConfigSource::Config));
if let Some(value) = config_value(kvs, SCANNER_CYCLE_MAX_DURATION, DEFAULT_SCANNER_CYCLE_MAX_DURATION_SECS) {
let secs = parse_config_u64(SCANNER_CYCLE_MAX_DURATION, value)?;
return cycle_duration_from_secs(SCANNER_CYCLE_MAX_DURATION, secs)
.map(|duration| (duration, ScannerRuntimeConfigSource::Config));
}
Ok((None, ScannerRuntimeConfigSource::Default))
Ok((
Some(Duration::from_secs(DEFAULT_SCANNER_CYCLE_MAX_DURATION_SECS)),
ScannerRuntimeConfigSource::Default,
))
}
fn cycle_duration_from_secs(key: &'static str, secs: u64) -> Result<Option<Duration>, ScannerRuntimeConfigError> {
if secs == 0 {
return Ok(None);
}
let duration = Duration::from_secs(secs);
if std::time::Instant::now().checked_add(duration).is_none() {
return Err(invalid_value(key, "<overflow>", "duration exceeds the timer range"));
}
Ok(Some(duration))
}
fn lookup_start_delay(kvs: Option<&KVS>) -> Result<(Option<Duration>, ScannerRuntimeConfigSource), ScannerRuntimeConfigError> {
@@ -553,12 +583,7 @@ pub(crate) fn lookup_scanner_runtime_config(
(speed.cycle_interval(), speed_source)
};
let (cycle_max_duration, cycle_max_duration_source) = lookup_optional_seconds(
scanner_kvs,
SCANNER_CYCLE_MAX_DURATION,
ENV_SCANNER_CYCLE_MAX_DURATION_SECS,
DEFAULT_SCANNER_CYCLE_MAX_DURATION_SECS,
)?;
let (cycle_max_duration, cycle_max_duration_source) = lookup_cycle_duration(scanner_kvs)?;
let (cycle_max_objects, cycle_max_objects_source) = lookup_count_budget(
scanner_kvs,
SCANNER_CYCLE_MAX_OBJECTS,
@@ -863,10 +888,10 @@ mod tests {
use rustfs_config::server_config::{Config as ServerConfig, KVS};
use rustfs_config::{
DEFAULT_DELIMITER, DEFAULT_HEAL_BITROT_CYCLE_SECS, ENV_SCANNER_BITROT_CYCLE_SECS, ENV_SCANNER_CACHE_SAVE_TIMEOUT_SECS,
ENV_SCANNER_CYCLE, ENV_SCANNER_CYCLE_MAX_OBJECTS, ENV_SCANNER_DELAY, ENV_SCANNER_MAX_WAIT_SECS, ENV_SCANNER_SPEED,
HEAL_BITROT_CYCLE, HEAL_SUB_SYS, SCANNER_BITROT_CYCLE, SCANNER_CACHE_SAVE_TIMEOUT, SCANNER_CYCLE,
SCANNER_CYCLE_MAX_DIRECTORIES, SCANNER_CYCLE_MAX_DURATION, SCANNER_CYCLE_MAX_OBJECTS, SCANNER_DELAY, SCANNER_IDLE_MODE,
SCANNER_SPEED, SCANNER_SUB_SYS, ScannerSpeed,
ENV_SCANNER_CYCLE, ENV_SCANNER_CYCLE_MAX_DURATION_SECS, ENV_SCANNER_CYCLE_MAX_OBJECTS, ENV_SCANNER_DELAY,
ENV_SCANNER_MAX_WAIT_SECS, ENV_SCANNER_SPEED, HEAL_BITROT_CYCLE, HEAL_SUB_SYS, SCANNER_BITROT_CYCLE,
SCANNER_CACHE_SAVE_TIMEOUT, SCANNER_CYCLE, SCANNER_CYCLE_MAX_DIRECTORIES, SCANNER_CYCLE_MAX_DURATION,
SCANNER_CYCLE_MAX_OBJECTS, SCANNER_DELAY, SCANNER_IDLE_MODE, SCANNER_SPEED, SCANNER_SUB_SYS, ScannerSpeed,
};
use serial_test::serial;
use std::collections::HashMap;
@@ -943,6 +968,43 @@ mod tests {
});
}
#[test]
#[serial]
fn scanner_unset_budget_uses_safe_default_but_explicit_zero_is_unbounded() {
let config = server_config_with_scanner(&[]);
with_var_unset(ENV_SCANNER_CYCLE_MAX_DURATION_SECS, || {
let resolved = lookup_scanner_runtime_config(Some(&config)).expect("scanner runtime config");
assert_eq!(resolved.cycle_budget.max_duration, Some(Duration::from_secs(1800)));
assert_eq!(resolved.cycle_max_duration_source, ScannerRuntimeConfigSource::Default);
});
let config = server_config_with_scanner(&[(SCANNER_CYCLE_MAX_DURATION, "0")]);
with_var_unset(ENV_SCANNER_CYCLE_MAX_DURATION_SECS, || {
let resolved = lookup_scanner_runtime_config(Some(&config)).expect("scanner runtime config");
assert_eq!(resolved.cycle_budget.max_duration, None);
assert_eq!(resolved.cycle_max_duration_source, ScannerRuntimeConfigSource::Config);
});
}
#[test]
#[serial]
fn cycle_budget_invalid_or_overflow_config_is_rejected() {
with_var(ENV_SCANNER_CYCLE_MAX_DURATION_SECS, Some("invalid"), || {
let error = lookup_scanner_runtime_config(None).expect_err("invalid duration env must be rejected");
assert!(error.to_string().contains(ENV_SCANNER_CYCLE_MAX_DURATION_SECS));
assert!(error.to_string().contains("<invalid>"));
assert!(!error.to_string().contains(": invalid ("));
});
with_var(ENV_SCANNER_CYCLE_MAX_DURATION_SECS, Some("18446744073709551616"), || {
assert!(lookup_scanner_runtime_config(None).is_err());
});
with_var(ENV_SCANNER_CYCLE_MAX_DURATION_SECS, Some("18446744073709551615"), || {
assert!(lookup_scanner_runtime_config(None).is_err());
});
let config = server_config_with_scanner(&[(SCANNER_CYCLE_MAX_DURATION, "not-a-duration")]);
assert!(lookup_scanner_runtime_config(Some(&config)).is_err());
}
#[test]
#[serial]
fn scanner_runtime_config_normalizes_persisted_default_speed() {