ydb/yt, branch mainMirror of YDB github reposhttps://code.mastervirt.ru/ydb/atom?h=main2026-06-07T18:00:46ZReclaim hazard pointers periodically on idle worker threads2026-06-07T18:00:46Zbabenkobabenko@yandex-team.com2026-06-07T17:25:31Zurn:sha1:e211f313308ba508d351979d52091dccd44798fe
Worker threads that park via TNotifyManager::Wait (TThreadPool,
TTwoLevelFairShareThreadPool) could strand hazard-retired objects on an
idle thread: an object still protected at the pre-park reclamation scan
stayed in the thread-local retire list and was never reclaimed while the
thread remained parked, pinning the object (and the memory it transitively
held, e.g. an unmounted tablet's lookup row cache) indefinitely.
Idle workers now reclaim hazard pointers periodically; while any retired
pointers remain pending a worker wakes up at most once per
HazardPointerReclaimPeriod to retry instead of blocking indefinitely. The
throttled per-thread reclamation is unified into a single
ReclaimHazardPointersPeriodically helper shared by TSchedulerThread,
TTwoLevelFairShareQueue and TNotifyManager.
commit_hash:862b70b662e47ca85d55864627e723a8f3feb2f8
YT-22593: Migrate IClient StartQuery query param to std::string2026-06-07T17:57:52Zbabenkobabenko@yandex-team.com2026-06-07T17:25:04Zurn:sha1:c906b7501b1f5f91adc8316e883755aa5ab38dbc
commit_hash:ebe98a0b25f06f6e5d601531325c66f0112dc23c
Intermediate changes2026-06-07T17:56:57Zrobot-pigletrobot-piglet@yandex-team.com2026-06-07T17:24:45Zurn:sha1:1114f4d5d16fe70d6bcd798a9fcc3844a83e88f7
commit_hash:11ffc36e8c033b5cd0c7d1ebc2989a6559fcf4b3
YT-27780: Make push-based shuffle configurable via start_shuffle2026-06-07T14:31:59Zapollo1321apollo1321@yandex-team.com2026-06-07T14:05:58Zurn:sha1:3014c231b26aa13f4f00454705af9ba1041b6afe
commit_hash:109d7b77afce712c68e57058c5f38e52eade292e
Intermediate changes2026-06-07T10:07:12Zrobot-pigletrobot-piglet@yandex-team.com2026-06-07T09:45:03Zurn:sha1:62fc40807071963a5e159369fbb27023f135aadb
commit_hash:d5cb132b1cc8a7c199d30cecf74c056e04a86377
YT-22593: Reduce TString usage in core/misc2026-06-06T22:31:50Zbabenkobabenko@yandex-team.com2026-06-06T22:15:06Zurn:sha1:ffc038d5a916cec181915795ca4b45f07ee2ae05
commit_hash:ba8c42476d6274212745348071682042e780037f
YT-22593: Migrate IClient admin (core dump, maintenance) method strings to std::string2026-06-06T21:53:09Zbabenkobabenko@yandex-team.com2026-06-06T21:35:59Zurn:sha1:d1c1aedb88345a776f41cd481415394ad0f4ca60
YT-22593: Migrate IClient admin (core dump, maintenance) method strings to std::string
commit_hash:7fe1737383174203b9883de719f2ec6fc561ed30
YT-22593: Migrate IClient file cache method params to std::string2026-06-06T21:34:24Zbabenkobabenko@yandex-team.com2026-06-06T20:55:29Zurn:sha1:fa15aff561d5197a2f231e8096db43e638d9d033
YT-22593: Migrate IClient file cache method params to std::string
commit_hash:40ffa51072f5818a0247e2cd071ff7455f0e2d9d
YT-22593: Migrate IClient password/token method params to std::string2026-06-06T21:19:47Zbabenkobabenko@yandex-team.com2026-06-06T20:52:49Zurn:sha1:d496342dde195f723cfa385c5e67a090defba016
YT-22593: Migrate IClient password/token method params to std::string
commit_hash:96d2e22ff2de310a8441a48aa0973e0cc8a7e752
Cache process/thread id getters and use them in TError origin capture2026-06-06T21:16:49Zbabenkobabenko@yandex-team.com2026-06-06T20:52:00Zurn:sha1:f10c7206fb31af8057446bceef9707aabaa9456e
## Motivation
Profiling the YT master Automaton thread showed TOriginAttributes::Capture (run on every non-OK TError) spending ~60% of its time in a getpid() syscall — uncached on glibc >= 2.25. NYT::GetCurrentThreadId() (gettid) feeds hot thread-affinity / log-manager checks on the same thread.
## Changes
- New library/cpp/yt/system/process_id.* with cached GetProcessId(); GetSystemThreadId() now caches the kernel tid in TLS. Both caches reset in the child after fork.
- Moved thread_name.{h,cpp} from misc to system.
- Removed GetCurrentProcessId/GetCurrentThreadId shims from yt/yt/core/misc/proc.{h,cpp}; migrated all callers to NYT::GetProcessId / NYT::GetSystemThreadId.
- TOriginAttributes::Capture uses the cached getters; recorded Tid is now the real kernel tid (matches perf/ps).
- Added microbenchmarks (library/cpp/yt/system/benchmarks, yt/yt/core/benchmarks/error.cpp).
## Microbenchmarks (release)
| | before | after |
|---|---|---|
| getpid | 101 ns | 0.33 ns |
| gettid | 102 ns | 1.64 ns |
| Capture | 161 ns | 50 ns |
| failed TError | 221 ns | 74 ns |
commit_hash:ee37ae57d61a5a2dd33daee935270f4bb93b7ff9