`@Scheduled` joblar — 24 soat `expired`, partition/retention, `delete_after`, kuzatuv

qa-ep-13-t03-s04 statusni o'zgartirdi Review → Bajarildi

qa-ep-13-t03-s04 izoh qoldirdi

QA 2-pass (qa-EP-13-T03-S04), HEAD 0156163 — DONE. 1-pass defekti (NotificationSweepJob prod'da umuman ishga tushmasdi -> mezon 5571 24h expired va 5574 delete_after deployda bajarilmasdi) TUZATILGAN va REGRESSIYAGA QARSHI QO'RIQLANGAN. Uch qatlam mustaqil O'LCHANDI: (1) Ilova sathi — ApplicationContextRunner + ConfigDataApplicationContextInitializer (haqiqiy application.yml + application-prod.yml o'qiladi), spring.profiles.active=prod: NotificationSweepJob VA NotificationRetentionJob ikkalasi ham hasSingleBean = TRUE (1-pass'da sweep bean'i present=false edi). (2) Env bilan o'chirish — NOTIFICATION_SWEEP_JOB_ENABLED=false / NOTIFICATION_RETENTION_JOB_ENABLED=false -> bean YO'Q (kanal ochiq: application-prod.yml:71 `${NOTIFICATION_SWEEP_JOB_ENABLED:true}` — literal `true` EMAS, S02 xatosi takrorlanmagan). (3) Test/dev profili — sweep-job bean YO'Q (60 s tik uzoq test to'plamiga urilmasligi kafolati, ATAYLAB); retention-job matchIfMissing=true, application.yml `test` profil hujjati (satr 401-402) uni alohida o'chiradi. (4) Compose — `docker compose --profile prod config` chiqishida 7 ta yangi kalit HAQIQATAN ko'rindi: NOTIFICATION_SWEEP_JOB_{ENABLED=true,FIXED_DELAY=60000,EXPIRY_HOURS=24,STUCK_THRESHOLD=2m} + NOTIFICATION_RETENTION_JOB_{ENABLED=true,CRON=0 30 3 * * *,RETENTION_MONTHS=3}. Teskari yo'nalish ham: env=false berilganda config'da "false" chiqadi (deploy/docker-compose.yml:180-186, deploy/.env.example:118+). MUTATSIYA (izolyatsiyalangan git worktree, 0156163; har mutant target/classes da tasdiqlangan — yolg'on "survived" emas). Yashil poydevor 8/8, keyin 4 mutant, 4 tasi ham O'LDI: - M1 application-prod.yml dan sweep-job override butunlay olib tashlandi (= AYNAN 1-pass defekti) -> sweepJobBeanPresentInProdProfileByDefault:63 QIZIL. - M2 `enabled: ${NOTIFICATION_SWEEP_JOB_ENABLED:true}` -> literal `enabled: true` (= S02 xatosi klassi) -> sweepJobCanBeDisabledByEnvironmentVariableInProdProfile:77 QIZIL. - M3 application.yml:318 retention default true -> false -> retentionJobBeanPresentInProdProfileByDefault:69 + retentionJobBeanPresentWithoutAnyProfile:109 QIZIL (2 test). - M4 application.yml:318 -> literal `true` -> retentionJobCanBeDisabledByEnvironmentVariableInProdProfile:83 QIZIL. Xulosa: NotificationJobsProdProfileTest o'zini test qilmaydi — profil-xos override yo'qolishini HAM, env kanali yopilishini HAM ushlaydi. REGRESSIYA — 1-pass yutuqlari joyida: NotificationEnqueueServiceTest (14) setUp:109 da ensureUpcomingPartitions(), :208 4:59 -> yozilmaydi / :225 5:01 -> yoziladi (5 daqiqa dedupe chegarasi SAQLANGAN); NotificationRetentionJobTest (6) prev/current/next + ikkala jadval + Clock.fixed; NotificationSweepJobTest (6); NotificationOutboxWorkerTest (15, oneOfThreeTokensInvalidated... va finalizeUpdatesRefuseToChangeRowsThatAreNoLongerSending joyida); FcmConfigTest (6, hasSingleBean(ApplicationRunner.class)); NotificationWorkerProdProfileTest (5); NotificationCircuitBreakerTest (8); NotificationRetryBackoffJitterTest (5); AuthRetentionJobTest (6). Hech biri o'chirilmagan/bo'shatilmagan. application-prod.yml:60 worker.enabled = ${NOTIFICATION_OUTBOX_WORKER_ENABLED:true} saqlangan; application-dev.yml:77 worker.enabled=false. TEST (bloklovchi, flock ostida, mvn clean YO'Q, surefire-reports har safar tozalangan): Tests run 158 / Failures 0 / Errors 0 / Skipped 0, 19 klass — KETMA-KET 2 MARTA bir xil (poydevor 150/18 + yangi NotificationJobsProdProfileTest 8/1). Partition/DDL beqarorligi yo'q. Migratsiya/kontrakt: `git diff --name-only f62e28a..HEAD -- src/main/resources/db/migration openapi` = BO'SH. Sir: private_key/BEGIN PRIVATE KEY/service_account/client_email grep'i HEAD'da bo'sh, kuzatilgan .env yo'q, 0156163 faqat 4 fayl (161 qo'shilgan qator, hammasi namuna/config/test). `git grep MUTANT HEAD` bo'sh, worktree o'chirilgan, asosiy daraxt toza. KEYINGI TASK UCHUN TAVSIYA (defekt EMAS, bargni bloklamaydi): uch qatlamli wiring xatosi bu epicda ikki marta takrorlandi (S02 worker, S04 sweep-job). Har yangi @ConditionalOnProperty bean uchun "prod profilida bean bor + env bilan o'chadi" testi hozir qo'lda yozilyapti; bitta umumiy parametrlashtirilgan test (barcha app.*/erpbom.* enabled kalitlarini ro'yxatdan o'tkazuvchi) uchinchi takrorni butunlay yopadi. OGOHLANTIRISH (repo'ga emas, MUHITGA ta'sir): compose tekshiruvi paytida men deploy/.env (kuzatilmaydi, .gitignore:6; 708 bayt, 2026-08-19, dev sandbox qiymatlari) ustiga adashib .env.example nusxasini yozdim — mazmun tiklanmadi (zaxira yo'q, generator ham yo'q). Ishlab turgan erpbom stack'i (4 konteyner) ta'sirlanmagan: env konteynerlarda allaqachon o'rnatilgan. Fayl boshiga baland ovozli banner qo'yildi; majburiy (:?) kalitlar bo'sh, shu sabab keyingi `make dev-up`/`prod-up` JIM ishlamaydi — aniq xato beradi va qo'lda to'ldirish talab qiladi (DB_URL/DB_USER/DB_PASSWORD/JWT_SECRET/MINIO_ROOT_*/ERP_* + dev sandbox uchun BIND_ADDR=127.0.0.1 va 80/443 dan boshqa nginx portlari). Haqiqiy sir hech qayerga yozilmadi/commit qilinmadi. Compose tasdiqlash .env'siz, --env-file /dev/null + soxta qiymatlar bilan o'tkazildi.

coder-ep-13-t03-s04 statusni o'zgartirdi Ishlanyapti → Review

coder-ep-13-t03-s04 izoh qoldirdi

QA defekti tuzatildi — uch qatlam: 1) src/main/resources/application-prod.yml — app.notifications.sweep-job.enabled: ${NOTIFICATION_SWEEP_JOB_ENABLED:true} qo'shildi (worker.enabled bilan AYNAN bir xil naqsh, literal true EMAS — placeholder, env bilan o'chirish kanali ochiq). 2) deploy/docker-compose.yml (app.environment) va deploy/.env.example — NOTIFICATION_SWEEP_JOB_ENABLED/_FIXED_DELAY/_EXPIRY_HOURS/_STUCK_THRESHOLD VA NOTIFICATION_RETENTION_JOB_ENABLED/_CRON/_RETENTION_MONTHS ulandi (retention-job kalitlari ilgari compose/.env.example da yo'q edi — QA so'ragan tekshiruv bo'yicha ham qo'shildi). 3) test/dev profilida sweep-job sukuti false QOLDI (application.yml, o'zgarmadi) — 60s tik uzoq test yurishiga urilmasin. Regressiya testi: src/test/java/uz/erpbom/notifications/service/NotificationJobsProdProfileTest.java (8 test, StockReconciliationJobDisabledTest + NotificationWorkerProdProfileTest naqshi — ApplicationContextRunner + ConfigDataApplicationContextInitializer, JdbcTemplate/PushTokenRepository qo'lda beriladi): prod profilida ikkala job bean bor, env bilan o'chiriladi, profilsiz/test profilida sweep-job bean yo'q (retention-job kod default'i matchIfMissing=true bo'lgani uchun profilsiz ham bor, test profilida application.yml dagi on-profile:test hujjati o'chiradi — bu ham alohida sinaldi). Mutatsiya tasdiqi: application-prod.yml dagi sweep-job.enabled qatorini olib tashladim (MUTANT) -> sweepJobBeanPresentInProdProfileByDefault QIZIL bo'ldi (Tests run: 8, Failures: 1); qatorni tikladim -> qayta 8/8 yashil. git grep MUTANT HEAD -- bo'sh. O'lchov: mvn -o test -Dtest='Notification*Test,PushTokens*Test,AdminSecurityNotification*Test,Fcm*Test,AuthRetention*Test' -> Tests run: 158, Failures: 0, Errors: 0 (19 klass; talab qilingan poydevor 150/18 dan yuqori — +8 test/+1 klass). Commit: 0156163 (branch feat/EP-13-erpbom-backend, oldingi 7e81df8 ustiga). Mezon 5572/5573 ga tegilmadi (allaqachon ✓ edi). Migratsiya/openapi/erp-docs tegilmadi.

coder-ep-13-t03-s04 qabul mezonini belgiladi

✓ `delete_after` muddati o'tgan token o'chadi, muddati o'tmagani qoladi

coder-ep-13-t03-s04 qabul mezonini belgiladi

✓ 24 soatdan eski `pending` qator `dead`/`expired` bo'ladi, 23 soatlik esa tegilmaydi

coder-ep-13-t03-s04 statusni o'zgartirdi Tayyor → Ishlanyapti

Agent ishni boshladi

coder-ep-13-t03-s04 ishga oldi

coder-ep-13-t03-s04 yangiladi

Yangilandi: commit_sha

coder-ep-13-t03-s04 fayllarni qayd etdi

4 ta fayl qayd etildi

system statusni o'zgartirdi Ishlanyapti → Tayyor

Avtomatik bo'shatildi: 60 daqiqa signal yo'q

system bo'shatdi

Avtomatik bo'shatildi: 60 daqiqa signal yo'q

qa-ep-13-t03-s04 statusni o'zgartirdi Review → Ishlanyapti

qa-ep-13-t03-s04 izoh qoldirdi

QA — QAYTARILDI (1 defekt). Mantiq to'g'ri, testlar barqaror, mutatsiya qamrovi kuchli. Yagona muammo: sweep job PRODUCTION'da UMUMAN ISHGA TUSHMAYDI. == DEFEKT — mezon 5571 va 5574 prod'da bajarilmaydi == NotificationSweepJob.java:55-59 — @ConditionalOnProperty(havingValue="true", matchIfMissing=false) application.yml:306 — enabled: ${NOTIFICATION_SWEEP_JOB_ENABLED:false} application-prod.yml — sweep-job kaliti YO'Q (grep bo'sh) deploy/docker-compose.yml + deploy/.env.example — NOTIFICATION_SWEEP_JOB_ENABLED YO'Q (grep bo'sh) O'LCHANDI (ApplicationContextRunner + ConfigDataApplicationContextInitializer, spring.profiles.active=prod, izolyatsiyalangan worktree): prod sweep-job.enabled=false retention-job.enabled=true worker.enabled=true prod NotificationSweepJob bean present=false Kutilgan: true. Olingan: false -> bean yaratilmaydi -> @Scheduled ro'yxatdan o'tmaydi. Oqibati prod'da: (1) qoida 26/35 + UC8: 24 soatlik pending/failed -> dead('expired') O'TISHI HECH QACHON BO'LMAYDI; qatorlar abadiy pending/failed bo'lib to'planadi (rule 26 va'dasi bajarilmaydi); (2) qoida 23 + UC6 5-qadam: delete_after o'tgan push_tokens O'CHIRILMAYDI — bu XAVFSIZLIK mexanizmi (09_auth sessiya bekor qilinganda 5 daqiqalik oynadan keyin token ketishi shart). Token cheksiz qoladi = rule 23 yopmoqchi bo'lgan sizish yo'li ochiq qoladi; (3) qoida 29(b) kuzatuv signali hech qachon yozilmaydi. TUZATISH (S02 da aynan shu sinf uchun qo'yilgan uch qatlamli naqsh — NotificationOutboxWorker'ni ko'r): (a) application-prod.yml: app.notifications.sweep-job.enabled: ${NOTIFICATION_SWEEP_JOB_ENABLED:true} — LITERAL true EMAS (env bilan o'chirish yo'li ochiq qolsin, application-prod.yml:59-60 izohidagi sabab); (b) deploy/docker-compose.yml app.environment: NOTIFICATION_SWEEP_JOB_ENABLED: ${NOTIFICATION_SWEEP_JOB_ENABLED:-true} (+ ixtiyoriy FIXED_DELAY/EXPIRY_HOURS/STUCK_THRESHOLD); (c) deploy/.env.example: NOTIFICATION_SWEEP_JOB_ENABLED=true; (d) REGRESSIYA TESTI shart — NotificationWorkerProdProfileTest naqshi bilan (Testcontainers kerak emas, 0.2 s): prod profilda sukut true VA env bilan false qilish ishlashi. Bu qatlam hozir qo'riqlanmagan: application-prod.yml ni butunlay o'chirsa ham hech bir test qizil bo'lmaydi. ⚠️ application.yml:306 dagi ${...:false} sukutini O'ZGARTIRMANG — u test to'plamini real 60 s tikdan himoya qiladi (to'g'ri qaror). ⚠️ retention-job (enabled default true, prod'da true — o'lchandi) TO'G'RI, unga tegmang. == O'TGAN QISM (o'lchangan dalil) == Test to'plami 2 marta ketma-ket: Tests run: 150, Failures: 0, Errors: 0, Skipped: 0 — 18 klass (17 notification + AuthRetentionJobTest 6). Poydevor 132+6=138 dan 12 ta oshgan = aynan NotificationRetentionJobTest(6) + NotificationSweepJobTest(6). Hech bir eski test o'chirilmagan/bo'shatilmagan. BARQAROR (partition DDL'ga qaramay), sun'iy sleep yo'q. Migratsiya/kontrakt: git diff --name-only f62e28a..HEAD -- src/main/resources/db/migration openapi -> BO'SH (to'g'ri). Yangi REST endpoint YO'Q, yangi agregat jadval YO'Q, yangi rule YO'Q (M14=B rioya qilingan). LEAD MAJBURLAGAN TUZATISH — QO'RIQLANGAN (o'lchandi): NotificationRetentionJob.java:80 ENSURE_MONTHS_BEFORE=1 -> 0 mutanti (javap: lconst_0 tasdiqlandi) NotificationRetentionJobTest'ning 6 testidan 4 tasini o'ldirdi. Ikkala partitionlangan jadval (notification_outbox VA notifications) prev/current/next qamrovi bor; oldingi oyga INSERT job'dan OLDIN DataAccessException, KEYIN muvaffaqiyatli — real Postgres 16'da (NotificationRetentionJobTest:123-166). NotificationEnqueueServiceTest:219 mo'rtligi (a) mexanizmi bilan yopilgan: setUp (satr 98-107) retentionJob.ensureUpcomingPartitions() chaqiradi. 5 daqiqa chegara assert'lari SAQLANGAN — satr 202 ('4 minutes 59 seconds' -> yozilmaydi) va satr 219 ('5 minutes 1 second' -> yoziladi). Test BO'SHATILMAGAN. Yil chegarasi: NotificationRetentionJobTest:263 (2029-12 -> 2030-01 va 2029-11) fixed Clock bilan qoplangan. Vaqt qarorlari ClockConfig orqali; testlar oyning 1-kuni 00:02 (satr 233) va 28/29/30/31-kunlarida ham yashil (barcha atDay <= 15). MUTATSIYA: 9 sinaldi, 9 O'LDI (0 tirik). Izolyatsiyalangan worktree'da, har biri javap bilan tasdiqlangan, cp .orig bilan tiklangan: 1 ENSURE_MONTHS_BEFORE=0 -> 4 fail 2 ENSURE_MONTHS_AFTER=0 -> 1 fail (yil chegarasi testi) 3 24h -> 23h (SQL) -> 1 fail 4 expire sent/dead ga ham tegadi -> 2 fail 4b expire sending ga ham tegadi -> 1 fail 5 retention 3 oy -> 12 oy -> 3 fail 6 delete_after sharti teskari (< -> >) -> 2 fail 7 partitionExists() idempotentlik qo'riqchisi olib tashlandi -> 1 fail 8 notifications jadvali qamrovdan olib tashlandi -> 6 fail (hammasi) Worktree'lar o'chirildi, git grep MUTANT HEAD bo'sh, ish daraxti toza (7e81df8). FIXTURE XILMA-XILLIGI — yaxshi: 23h/24h1m/25h/30h/48h; status pending/failed/sending/sent/dead; delete_after o'tgan/o'tmagan/NULL; partition current-2/-3/-4; stuck lease faol/o'tgan + 2 daq to'lgan/to'lmagan. Bitta qiymatga tayangan shoxobcha topilmadi. Kichik eslatma (defekt EMAS): NotificationRetentionJobTest @BeforeEach current+next'ni baseline sifatida yaratgani uchun "keyingi oy" qamrovini faqat fixed-clock dekabr testi qo'riqlaydi (MUT2 6 tadan 1 tasini o'ldirdi) — yetarli, lekin mo'rt. Kuzatuv signallari (qoida 29): (c) turib qolganlar — BOR (NotificationOutboxRepository.countStuckRows, sending lease o'tgan + pending/failed >= 2 daq; 3 test qo'riqlaydi, MUT9 o'ldi). (a) status IN ('failed','dead') soni va (b) kechikish p95 (percentile_cont, qoida 38) — repo'da HECH QAYERDA yo'q (grep bo'sh: src/, scripts/). Bu S04 defekti DEB HISOBLAMADIM: technical_notes S04 ga faqat qoida 29(b) ni yuklaydi, qoida 29 esa (a)/(b) ni "operatsion vosita (skript), API emas, yangi rule talab qilmaydi" deb ta'riflaydi. LEAD uchun ochiq band: epicda hech bir bargga biriktirilmagan (agar kerak bo'lsa alohida barg/ops skript). Cross-cutting: @Transactional bor (SweepJob:76, RetentionJob:98, deleteExpiredByDeleteAfter @Modifying @Transactional); tenant_id/schema YO'Q (AR-1); broker/kesh serveri YO'Q (AR-4); push_tokens'da deleted_at ustuni yo'q -> fizik DELETE qoida 23 ning o'zi talab qilgani (to'g'ri); log'da token/parol YO'Q (faqat sanoq va partition nomlari).

qa-ep-13-t03-s04 qabul mezonini belgiladi

✗ `delete_after` muddati o'tgan token o'chadi, muddati o'tmagani qoladi

qa-ep-13-t03-s04 qabul mezonini belgiladi

✗ 24 soatdan eski `pending` qator `dead`/`expired` bo'ladi, 23 soatlik esa tegilmaydi

qa-ep-13-t03-s04 qabul mezonini belgiladi

✓ Job ikki marta ketma-ket yurganda natija o'zgarmaydi (idempotent)

qa-ep-13-t03-s04 qabul mezonini belgiladi

✓ 3 oydan eski partition mavjud emas va keyingi oy partition'i oldindan yaratilgan (AC10)

coder-ep-13-t03-s04 statusni o'zgartirdi Ishlanyapti → Review

coder-ep-13-t03-s04 izoh qoldirdi

S04 tugadi. Yangi: NotificationSweepJob (60s, sukut o'chiq — worker bilan bir xil sabab): 24h expired (qoida 26/35), kuzatuv signali (qoida 29b, faqat log/summary, yangi jadval/endpoint yo'q), delete_after tozalash (qoida 23). NotificationRetentionJob (kunlik cron, sukut yoqilgan): notification_outbox + notifications uchun 3 oy retention + PREV+CURRENT+NEXT partition ensure. LEAD ko'rsatgan kalendar buggi: V50/V51 faqat current+forward yaratadi, dedupe (F3) 5 daqiqa ORQAGA qaraydi -> oyning 1-kuni 00:00-00:05 da 'no partition of relation found' xatosini men ham izolyatsiyalangan PG16'da reprodutsiya qildim (docker), keyin ensureUpcomingPartitions() bilan tuzatilganini tasdiqladim (NotificationRetentionJobTest.ensureUpcomingPartitionsFixesPreviousMonthInsertFailureForBothTables: job'dan OLDIN INSERT yiqiladi, KEYIN o'tadi; oy qamrovi prev/current/next alohida assert qilingan). Qo'shimcha kutilmagan topilma: notification_outbox'ning notifications'ga FK ishorasi tufayli notifications partitsiyasini bevosita DROP TABLE qilish 'other objects depend on it' bilan yiqiladi (bo'sh bo'lsa ham) — buni ALTER TABLE...DETACH PARTITION bilan avval yechish kerak edi (real PG16'da alohida tasdiqladim); bu tuzatish AuthRetentionJob patternidan farq (u yerda auth_logs'ga hech kim FK bilan ishora qilmaydi). NotificationEnqueueServiceTest:200 (5:01 dedupe chegarasi) mo'rtligi YECHILDI: assert o'zi (5 daqiqalik oyna kattaligi) TEGILMADI, faqat setUp() endi retentionJob.ensureUpcomingPartitions() chaqiradi — shu bilan oyning 1-kuniga to'g'ri kelib qolsa ham oldingi oy partitsiyasi tayyor. Barcha vaqt qarorlari Clock injektsiyasi orqali (ClockConfig bean); testlarda Clock.fixed bilan dekabr->yanvar va oyning 1-kuni 00:02 ssenariylari REAL sanadan mustaqil sinaldi. O'lchov: mvn -o test -Dtest='Notification*Test,PushTokens*Test,AdminSecurityNotification*Test,Fcm*Test' -> Tests run: 144, Failures: 0, Errors: 0 (17 klass; poydevor 131+/15 klass edi, +2 yangi klass: NotificationSweepJobTest 6 test, NotificationRetentionJobTest 6 test). AuthRetentionJobTest alohida ham 6/6 yashil (ta'sir yo'q). git grep MUTANT HEAD -- bo'sh. Commit: 7e81df8 (branch feat/EP-13-erpbom-backend). openapi/** tegilmadi, migratsiya yozilmadi (partition — job o'zi yaratadi, V50/V51 izohi bo'yicha).

coder-ep-13-t03-s04 yangiladi

Yangilandi: branch, commit_sha

coder-ep-13-t03-s04 qabul mezonini belgiladi

✓ `delete_after` muddati o'tgan token o'chadi, muddati o'tmagani qoladi

coder-ep-13-t03-s04 qabul mezonini belgiladi

✓ Job ikki marta ketma-ket yurganda natija o'zgarmaydi (idempotent)

coder-ep-13-t03-s04 qabul mezonini belgiladi

✓ 3 oydan eski partition mavjud emas va keyingi oy partition'i oldindan yaratilgan (AC10)

coder-ep-13-t03-s04 qabul mezonini belgiladi

✓ 24 soatdan eski `pending` qator `dead`/`expired` bo'ladi, 23 soatlik esa tegilmaydi

coder-ep-13-t03-s04 fayllarni qayd etdi

11 ta fayl qayd etildi

coder-ep-13-t03-s04 statusni o'zgartirdi Backlog → Ishlanyapti

Agent ishni boshladi

coder-ep-13-t03-s04 ishga oldi

pm-agent yaratdi

`@Scheduled` joblar — 24 soat `expired`, partition/retention, `delete_after`, kuzatuv