FerretDB: default to polling-only on all platforms; OpLog is opt-in (#6503).

FerretDB v1 can tail an OpLog, but on the SQLite backend Meteor's tailable+awaitData OpLog tail keeps FerretDB CPU pinned (reporters saw ~190-390% even idle; a 2-core box maxed out) and a struggling tail also surfaces as 'oplog catching up took too long', stalling board and login-page loading. Despite the FerretDB-side mitigations (16 MiB capped oplog, 500ms awaitData poll, SQLite pragmas, CPU self-regulation) the peg persisted on real deployments, and the polling-only path reliably drops CPU to ~10%. So make POLLING the default on every launcher (snap wekan-control, bundle start-wekan.sh, docker wekan-entrypoint.sh + docker-compose.yml, windows start-wekan.bat, sandstorm start.js): WEKAN_FERRETDB_OPLOG now defaults false, docker-compose reactivity is 'polling' with MONGO_OPLOG_URL commented out. The OpLog machinery still exists and is opt-in via WEKAN_FERRETDB_OPLOG=true. The snap exposes it as a setting: 'snap set wekan wekan-ferretdb-oplog=true|false' (registered in snap-src/bin/config, default false). The calmer polling throttle (#6467/#6468) and SQLite pragma tuning keep poll-and-diff cheap.

Tests: tests/ferretdbOplog.test.cjs updated to assert the polling-only default everywhere, the commented-out docker-compose MONGO_OPLOG_URL, the opt-in OpLog machinery still present, and the new wekan-ferretdb-oplog snap setting (default false).

Thanks to xet7 !

docker-compose.yml

@@ -142,13 +142,15 @@ services:
       # https://github.com/meteor/meteor/blob/release-3.5/v3-docs/docs/performance/change-streams-observer-driver.md#choosing-the-reactivity-driver-order
       # Keep oplog/polling fallback available for deployments that cannot use Change Streams:
       # https://github.com/wekan/wekan/issues/6307#issuecomment-4299349231
-      # wekan/wekan#6480/#6481: FerretDB v1 now ships an OpLog (the ferretdb service
-      # above runs with --repl-set-name=rs0), so prefer OpLog tailing and fall back
-      # to polling — this is the main fix for FerretDB's high CPU on busy boards.
-      # FerretDB has no MongoDB change streams. To force polling only, set both to
-      # `polling` and comment out MONGO_OPLOG_URL below.
-      - DEFAULT_METEOR_REACTIVITY_ORDER=oplog,polling
-      - METEOR_REACTIVITY_ORDER=oplog,polling
+      # wekan/wekan#6503/#6480/#6481: FerretDB v1 CAN tail an OpLog (the ferretdb
+      # service above runs with --repl-set-name=rs0), but on the SQLite backend the
+      # tailable+awaitData OpLog tail keeps FerretDB CPU pinned (~190–390% even idle)
+      # and a struggling tail shows as "oplog catching up took too long" and stalls
+      # loading, so the DEFAULT here is now POLLING ONLY. FerretDB has no MongoDB
+      # change streams. To opt INTO OpLog tailing, set both to `oplog,polling` and
+      # uncomment MONGO_OPLOG_URL in the wekan service below.
+      - DEFAULT_METEOR_REACTIVITY_ORDER=polling
+      - METEOR_REACTIVITY_ORDER=polling
       # uws does not work at s390x. sockjs works.
       #- DDP_TRANSPORT=uws
       - DDP_TRANSPORT=sockjs
@@ -215,11 +217,14 @@ services:
       #   - Latency from 50ms to 2000ms
       # Must configure MongoDB replica set first
       # See: https://blog.meteor.com/tuning-meteor-mongo-livedata-for-scalability-13fe9deb8908
-      # FerretDB v1 emulates the OpLog via --repl-set-name=rs0 (set on the ferretdb
-      # service above). This enables real-time DDP updates and stops the poll-and-diff
-      # CPU burn (wekan/wekan#6480/#6481). To force polling only, comment this out and
-      # set METEOR_REACTIVITY_ORDER=polling above.
-      - MONGO_OPLOG_URL=mongodb://ferretdb:27017/local?replicaSet=rs0
+      # FerretDB v1 CAN emulate the OpLog via --repl-set-name=rs0 (set on the
+      # ferretdb service above), but on the SQLite backend the tail keeps FerretDB
+      # CPU pinned and stalls loading (wekan/wekan#6503), so it is DISABLED by
+      # default here. Merely SETTING MONGO_OPLOG_URL makes Meteor start an OpLog
+      # tail regardless of the reactivity order (#6498), so it is commented out.
+      # To opt INTO OpLog tailing, uncomment the next line AND set
+      # METEOR_REACTIVITY_ORDER=oplog,polling above.
+      #- MONGO_OPLOG_URL=mongodb://ferretdb:27017/local?replicaSet=rs0
       #---------------------------------------------------------------
       # ==== OPTIONAL: KADIRA PERFORMANCE MONITORING FOR METEOR ====
       # https://github.com/edemaine/kadira-compose

releases/ferretdb/start-wekan.bat

@@ -23,11 +23,11 @@ set "FERRETDB_SQLITE_DIR=%FILES%\db"
 if not defined PORT set "PORT=8080"
 if not defined ROOT_URL set "ROOT_URL=http://localhost:%PORT%"
 if not defined MONGO_URL set "MONGO_URL=mongodb://127.0.0.1:27017/wekan"
-REM  #6480/#6481: FerretDB v1 now ships an OpLog (started below with
-REM  --repl-set-name), so by default WeKan's Meteor TAILS the OpLog instead of
-REM  poll-and-diff (fixes high FerretDB CPU on busy boards). Set
-REM  WEKAN_FERRETDB_OPLOG=false to force the old polling-only behaviour.
-if not defined WEKAN_FERRETDB_OPLOG set "WEKAN_FERRETDB_OPLOG=true"
+REM  #6503/#6480/#6481: FerretDB v1 CAN tail an OpLog (started below with
+REM  --repl-set-name), but on the SQLite backend the tail keeps FerretDB CPU
+REM  pinned (~190-390% even idle) and stalls loading, so the DEFAULT is now
+REM  POLLING ONLY. Opt into OpLog tailing with WEKAN_FERRETDB_OPLOG=true.
+if not defined WEKAN_FERRETDB_OPLOG set "WEKAN_FERRETDB_OPLOG=false"
 if not defined WEKAN_FERRETDB_REPL_SET set "WEKAN_FERRETDB_REPL_SET=rs0"
 set "FERRET_REPL_ARG="
 if /I "%WEKAN_FERRETDB_OPLOG%"=="true" (

releases/ferretdb/start-wekan.sh

@@ -69,16 +69,16 @@ FERRET_PID=""
 stop_ferret() { [ -n "$FERRET_PID" ] && kill "$FERRET_PID" 2>/dev/null || true; }
 trap 'stop_ferret; exit 0' INT TERM
 
-# #6480/#6481: FerretDB v1 now ships an OpLog (auto-created capped local.oplog.rs
-# + replica-set hello handshake — the ferretdb below is launched with
-# --repl-set-name). By default WeKan's Meteor TAILS the OpLog instead of
-# poll-and-diff, the main fix for FerretDB's high CPU on busy boards. Set
-# WEKAN_FERRETDB_OPLOG=false to force the old polling-only behaviour.
-# The polling settings remain the FALLBACK for queries Meteor cannot observe via
-# the OpLog. #6467/#6468: Meteor's defaults (re-poll 50 ms after any write, at
-# least every 10 s) hammer the database; calmer defaults re-poll at most every
-# 2 s / 30 s. Overridable by exporting different values before running this script.
-WEKAN_FERRETDB_OPLOG="${WEKAN_FERRETDB_OPLOG:-true}"
+# #6503/#6480/#6481: FerretDB v1 CAN tail an OpLog (auto-created capped
+# local.oplog.rs + replica-set hello handshake), but on the SQLite backend the
+# tailable+awaitData OpLog tail keeps FerretDB CPU pinned (reporters saw ~190–390%
+# even idle) and a struggling tail shows as "oplog catching up took too long" and
+# stalls loading, so the DEFAULT is now POLLING ONLY. Opt into OpLog tailing with
+# WEKAN_FERRETDB_OPLOG=true if you specifically want it.
+# The polling settings apply either way. #6467/#6468: Meteor's defaults (re-poll
+# 50 ms after any write, at least every 10 s) hammer the database; calmer defaults
+# re-poll at most every 2 s / 30 s. Overridable by exporting values before running.
+WEKAN_FERRETDB_OPLOG="${WEKAN_FERRETDB_OPLOG:-false}"
 WEKAN_FERRETDB_REPL_SET="${WEKAN_FERRETDB_REPL_SET:-rs0}"
 FERRET_REPL_ARG=""
 if [ "$want_ferret" = true ]; then

releases/ferretdb/wekan-entrypoint.sh

@@ -47,12 +47,13 @@ if [ "$want_ferret" = true ]; then
   fi
   export MONGO_URL="${MONGO_URL:-mongodb://$FERRETDB_LISTEN_ADDR/wekan}"
   mkdir -p "$FERRETDB_SQLITE_DIR"
-  # #6480/#6481: FerretDB v1 now ships an OpLog (auto-created capped
-  # local.oplog.rs + replica-set hello handshake), launched below with
-  # --repl-set-name. By default WeKan's Meteor TAILS the OpLog instead of
-  # poll-and-diff — the main fix for FerretDB's high CPU on busy boards. Set
-  # WEKAN_FERRETDB_OPLOG=false to force the old polling-only behaviour.
-  WEKAN_FERRETDB_OPLOG="${WEKAN_FERRETDB_OPLOG:-true}"
+  # #6503/#6480/#6481: FerretDB v1 CAN tail an OpLog (auto-created capped
+  # local.oplog.rs + replica-set hello handshake), but on the SQLite backend the
+  # tailable+awaitData tail keeps FerretDB CPU pinned (~190–390% even idle) and a
+  # struggling tail shows as "oplog catching up took too long" and stalls loading,
+  # so the DEFAULT is now POLLING ONLY. Opt into OpLog tailing with
+  # WEKAN_FERRETDB_OPLOG=true if you specifically want it.
+  WEKAN_FERRETDB_OPLOG="${WEKAN_FERRETDB_OPLOG:-false}"
   REPL_SET_NAME="${WEKAN_FERRETDB_REPL_SET:-rs0}"
   if [ "$WEKAN_FERRETDB_OPLOG" = "true" ]; then
     export MONGO_OPLOG_URL="${MONGO_OPLOG_URL:-mongodb://$FERRETDB_LISTEN_ADDR/local?replicaSet=$REPL_SET_NAME}"

sandstorm-src/start.js

@@ -54,13 +54,14 @@ const DB_PORT   = '4001';                                      // FerretDB stead
 const SRC_PORT  = '4003';                                      // mongod 3.0 (migration source)
 const NISCU_PORT = '4004';                                     // niscud (niscu→3.0 source)
 
-// #6480/#6481: FerretDB v1 ships an OpLog (auto-created capped local.oplog.rs +
-// replica-set hello handshake). Run the STEADY-STATE FerretDB with a replica-set
-// name and point WeKan at it so Meteor tails the OpLog instead of poll-and-diff.
-// Polling stays the fallback (METEOR_REACTIVITY_ORDER=oplog,polling → OpLog only
-// when tailing works). WEKAN_FERRETDB_OPLOG=false forces polling only. The
-// transient migration-target FerretDB above stays OpLog-free (no --repl-set-name).
-const FERRET_OPLOG  = process.env.WEKAN_FERRETDB_OPLOG !== 'false';
+// #6503/#6480/#6481: FerretDB v1 CAN tail an OpLog (auto-created capped
+// local.oplog.rs + replica-set hello handshake), but on the SQLite backend the
+// tailable+awaitData tail keeps FerretDB CPU pinned (~190–390% even idle) and a
+// struggling tail shows as "oplog catching up took too long" and stalls loading,
+// so the DEFAULT is now POLLING ONLY. Opt into OpLog tailing with
+// WEKAN_FERRETDB_OPLOG=true. The transient migration-target FerretDB above stays
+// OpLog-free (no --repl-set-name) regardless.
+const FERRET_OPLOG  = process.env.WEKAN_FERRETDB_OPLOG === 'true';
 const REPL_SET_NAME = process.env.WEKAN_FERRETDB_REPL_SET || 'rs0';
 
 function ensureDirs() {

snap-src/bin/config

@@ -3,7 +3,7 @@
 # All supported keys are defined here together with descriptions and default values
 
 # list of supported keys
-keys="DEBUG DATABASE DDP_TRANSPORT METEOR_REACTIVITY_ORDER MONGO_LOG_DESTINATION MONGO_URL MONGO_OPLOG_URL MONGODB_BIND_UNIX_SOCKET MONGODB_BIND_IP MONGODB_PORT MAIL_URL MAIL_FROM MAIL_SERVICE MAIL_SERVICE_USER MAIL_SERVICE_PASSWORD ROOT_URL PORT DISABLE_MONGODB CADDY_ENABLED CADDY_BIND_PORT WITH_API RICHER_CARD_COMMENT_EDITOR CARDS_LOADING CARD_OPENED_WEBHOOK_ENABLED ACCOUNTS_LOCKOUT_KNOWN_USERS_FAILURES_BEFORE ACCOUNTS_LOCKOUT_KNOWN_USERS_PERIOD ACCOUNTS_LOCKOUT_KNOWN_USERS_FAILURE_WINDOW ACCOUNTS_LOCKOUT_UNKNOWN_USERS_FAILURES_BERORE ACCOUNTS_LOCKOUT_UNKNOWN_USERS_LOCKOUT_PERIOD ACCOUNTS_LOCKOUT_UNKNOWN_USERS_FAILURE_WINDOW ACCOUNTS_COMMON_LOGIN_EXPIRATION_IN_DAYS ATTACHMENTS_UPLOAD_EXTERNAL_PROGRAM ATTACHMENTS_UPLOAD_MIME_TYPES ATTACHMENTS_UPLOAD_MAX_SIZE AVATARS_UPLOAD_EXTERNAL_PROGRAM AVATARS_UPLOAD_MIME_TYPES AVATARS_UPLOAD_MAX_SIZE MAX_IMAGE_PIXEL IMAGE_COMPRESS_RATIO BIGEVENTS_PATTERN NOTIFICATION_TRAY_AFTER_READ_DAYS_BEFORE_REMOVE NOTIFY_DUE_DAYS_BEFORE_AND_AFTER NOTIFY_DUE_AT_HOUR_OF_DAY DEFAULT_BOARD_ID EMAIL_NOTIFICATION_TIMEOUT CORS CORS_ALLOW_HEADERS CORS_EXPOSE_HEADERS MATOMO_ADDRESS MATOMO_SITE_ID MATOMO_DO_NOT_TRACK MATOMO_WITH_USERNAME METRICS_ALLOWED_IP_ADDRESSES BROWSER_POLICY_ENABLED TRUSTED_URL WEBHOOKS_ATTRIBUTES OAUTH2_ENABLED OIDC_REDIRECTION_ENABLED OAUTH2_CA_CERT OAUTH2_LOGIN_STYLE OAUTH2_CLIENT_ID OAUTH2_SECRET OAUTH2_SERVER_URL OAUTH2_AUTH_ENDPOINT OAUTH2_USERINFO_ENDPOINT OAUTH2_TOKEN_ENDPOINT OAUTH2_LOGOUT_ENDPOINT OAUTH2_ID_MAP OAUTH2_USERNAME_MAP OAUTH2_FULLNAME_MAP OAUTH2_ID_TOKEN_WHITELIST_FIELDS OAUTH2_EMAIL_MAP OAUTH2_REQUEST_PERMISSIONS OAUTH2_ADFS_ENABLED OAUTH2_B2C_ENABLED OAUTH2_MERGE_EXISTING_USERS LDAP_ENABLE LDAP_PORT LDAP_HOST LDAP_AD_SIMPLE_AUTH LDAP_BASEDN LDAP_LOGIN_FALLBACK LDAP_RECONNECT LDAP_TIMEOUT LDAP_IDLE_TIMEOUT LDAP_CONNECT_TIMEOUT LDAP_AUTHENTIFICATION LDAP_AUTHENTIFICATION_USERDN LDAP_AUTHENTIFICATION_PASSWORD LDAP_LOG_ENABLED LDAP_BACKGROUND_SYNC LDAP_BACKGROUND_SYNC_INTERVAL LDAP_BACKGROUND_SYNC_KEEP_EXISTANT_USERS_UPDATED LDAP_BACKGROUND_SYNC_IMPORT_NEW_USERS LDAP_ENCRYPTION LDAP_CA_CERT LDAP_REJECT_UNAUTHORIZED LDAP_USER_AUTHENTICATION LDAP_USER_AUTHENTICATION_FIELD LDAP_USER_SEARCH_FILTER LDAP_USER_SEARCH_SCOPE LDAP_USER_SEARCH_FIELD LDAP_SEARCH_PAGE_SIZE LDAP_SEARCH_SIZE_LIMIT LDAP_GROUP_FILTER_ENABLE LDAP_GROUP_FILTER_OBJECTCLASS LDAP_GROUP_FILTER_GROUP_ID_ATTRIBUTE LDAP_GROUP_FILTER_GROUP_MEMBER_ATTRIBUTE LDAP_GROUP_FILTER_GROUP_MEMBER_FORMAT LDAP_GROUP_FILTER_GROUP_NAME LDAP_UNIQUE_IDENTIFIER_FIELD LDAP_UTF8_NAMES_SLUGIFY LDAP_USERNAME_FIELD LDAP_FULLNAME_FIELD LDAP_MERGE_EXISTING_USERS LDAP_SYNC_USER_DATA LDAP_SYNC_USER_DATA_FIELDMAP LDAP_SYNC_GROUP_ROLES LDAP_DEFAULT_DOMAIN LDAP_EMAIL_MATCH_ENABLE LDAP_EMAIL_MATCH_REQUIRE LDAP_EMAIL_MATCH_VERIFIED LDAP_EMAIL_FIELD LDAP_SYNC_ADMIN_STATUS LDAP_SYNC_ADMIN_GROUPS HEADER_LOGIN_ID HEADER_LOGIN_FIRSTNAME HEADER_LOGIN_LASTNAME HEADER_LOGIN_EMAIL HEADER_LOGIN_TRUSTED_IPS HEADER_LOGIN_TRUSTED_PROXIES LOGOUT_WITH_TIMER LOGOUT_IN LOGOUT_ON_HOURS LOGOUT_ON_MINUTES DEFAULT_AUTHENTICATION_METHOD PASSWORD_LOGIN_ENABLED CAS_ENABLED CAS_BASE_URL CAS_LOGIN_URL CAS_VALIDATE_URL SAML_ENABLED SAML_PROVIDER SAML_ENTRYPOINT SAML_ISSUER SAML_CERT SAML_IDPSLO_REDIRECTURL SAML_PRIVATE_KEYFILE SAML_PUBLIC_CERTFILE SAML_IDENTIFIER_FORMAT SAML_LOCAL_PROFILE_MATCH_ATTRIBUTE SAML_ATTRIBUTES ORACLE_OIM_ENABLED RESULTS_PER_PAGE WAIT_SPINNER NODE_OPTIONS LDAP_AUTHENTIFICATION_PASSWORD_FILE OAUTH2_SECRET_FILE MAIL_SERVICE_PASSWORD_FILE MONGO_PASSWORD_FILE S3 S3_SECRET_FILE OAUTH2_AUTO_REGISTRATION OAUTH2_ADMIN_GROUPS NOTIFY_ON_ASSIGN LDAP_BACKGROUND_SYNC_DISABLE_NONEXISTANT_USERS LDAP_SYNC_ORGANIZATIONS LDAP_SYNC_ORGANIZATIONS_GROUPS LDAP_SYNC_TEAMS LDAP_SYNC_TEAMS_GROUPS"
+keys="DEBUG DATABASE WEKAN_FERRETDB_OPLOG DDP_TRANSPORT METEOR_REACTIVITY_ORDER MONGO_LOG_DESTINATION MONGO_URL MONGO_OPLOG_URL MONGODB_BIND_UNIX_SOCKET MONGODB_BIND_IP MONGODB_PORT MAIL_URL MAIL_FROM MAIL_SERVICE MAIL_SERVICE_USER MAIL_SERVICE_PASSWORD ROOT_URL PORT DISABLE_MONGODB CADDY_ENABLED CADDY_BIND_PORT WITH_API RICHER_CARD_COMMENT_EDITOR CARDS_LOADING CARD_OPENED_WEBHOOK_ENABLED ACCOUNTS_LOCKOUT_KNOWN_USERS_FAILURES_BEFORE ACCOUNTS_LOCKOUT_KNOWN_USERS_PERIOD ACCOUNTS_LOCKOUT_KNOWN_USERS_FAILURE_WINDOW ACCOUNTS_LOCKOUT_UNKNOWN_USERS_FAILURES_BERORE ACCOUNTS_LOCKOUT_UNKNOWN_USERS_LOCKOUT_PERIOD ACCOUNTS_LOCKOUT_UNKNOWN_USERS_FAILURE_WINDOW ACCOUNTS_COMMON_LOGIN_EXPIRATION_IN_DAYS ATTACHMENTS_UPLOAD_EXTERNAL_PROGRAM ATTACHMENTS_UPLOAD_MIME_TYPES ATTACHMENTS_UPLOAD_MAX_SIZE AVATARS_UPLOAD_EXTERNAL_PROGRAM AVATARS_UPLOAD_MIME_TYPES AVATARS_UPLOAD_MAX_SIZE MAX_IMAGE_PIXEL IMAGE_COMPRESS_RATIO BIGEVENTS_PATTERN NOTIFICATION_TRAY_AFTER_READ_DAYS_BEFORE_REMOVE NOTIFY_DUE_DAYS_BEFORE_AND_AFTER NOTIFY_DUE_AT_HOUR_OF_DAY DEFAULT_BOARD_ID EMAIL_NOTIFICATION_TIMEOUT CORS CORS_ALLOW_HEADERS CORS_EXPOSE_HEADERS MATOMO_ADDRESS MATOMO_SITE_ID MATOMO_DO_NOT_TRACK MATOMO_WITH_USERNAME METRICS_ALLOWED_IP_ADDRESSES BROWSER_POLICY_ENABLED TRUSTED_URL WEBHOOKS_ATTRIBUTES OAUTH2_ENABLED OIDC_REDIRECTION_ENABLED OAUTH2_CA_CERT OAUTH2_LOGIN_STYLE OAUTH2_CLIENT_ID OAUTH2_SECRET OAUTH2_SERVER_URL OAUTH2_AUTH_ENDPOINT OAUTH2_USERINFO_ENDPOINT OAUTH2_TOKEN_ENDPOINT OAUTH2_LOGOUT_ENDPOINT OAUTH2_ID_MAP OAUTH2_USERNAME_MAP OAUTH2_FULLNAME_MAP OAUTH2_ID_TOKEN_WHITELIST_FIELDS OAUTH2_EMAIL_MAP OAUTH2_REQUEST_PERMISSIONS OAUTH2_ADFS_ENABLED OAUTH2_B2C_ENABLED OAUTH2_MERGE_EXISTING_USERS LDAP_ENABLE LDAP_PORT LDAP_HOST LDAP_AD_SIMPLE_AUTH LDAP_BASEDN LDAP_LOGIN_FALLBACK LDAP_RECONNECT LDAP_TIMEOUT LDAP_IDLE_TIMEOUT LDAP_CONNECT_TIMEOUT LDAP_AUTHENTIFICATION LDAP_AUTHENTIFICATION_USERDN LDAP_AUTHENTIFICATION_PASSWORD LDAP_LOG_ENABLED LDAP_BACKGROUND_SYNC LDAP_BACKGROUND_SYNC_INTERVAL LDAP_BACKGROUND_SYNC_KEEP_EXISTANT_USERS_UPDATED LDAP_BACKGROUND_SYNC_IMPORT_NEW_USERS LDAP_ENCRYPTION LDAP_CA_CERT LDAP_REJECT_UNAUTHORIZED LDAP_USER_AUTHENTICATION LDAP_USER_AUTHENTICATION_FIELD LDAP_USER_SEARCH_FILTER LDAP_USER_SEARCH_SCOPE LDAP_USER_SEARCH_FIELD LDAP_SEARCH_PAGE_SIZE LDAP_SEARCH_SIZE_LIMIT LDAP_GROUP_FILTER_ENABLE LDAP_GROUP_FILTER_OBJECTCLASS LDAP_GROUP_FILTER_GROUP_ID_ATTRIBUTE LDAP_GROUP_FILTER_GROUP_MEMBER_ATTRIBUTE LDAP_GROUP_FILTER_GROUP_MEMBER_FORMAT LDAP_GROUP_FILTER_GROUP_NAME LDAP_UNIQUE_IDENTIFIER_FIELD LDAP_UTF8_NAMES_SLUGIFY LDAP_USERNAME_FIELD LDAP_FULLNAME_FIELD LDAP_MERGE_EXISTING_USERS LDAP_SYNC_USER_DATA LDAP_SYNC_USER_DATA_FIELDMAP LDAP_SYNC_GROUP_ROLES LDAP_DEFAULT_DOMAIN LDAP_EMAIL_MATCH_ENABLE LDAP_EMAIL_MATCH_REQUIRE LDAP_EMAIL_MATCH_VERIFIED LDAP_EMAIL_FIELD LDAP_SYNC_ADMIN_STATUS LDAP_SYNC_ADMIN_GROUPS HEADER_LOGIN_ID HEADER_LOGIN_FIRSTNAME HEADER_LOGIN_LASTNAME HEADER_LOGIN_EMAIL HEADER_LOGIN_TRUSTED_IPS HEADER_LOGIN_TRUSTED_PROXIES LOGOUT_WITH_TIMER LOGOUT_IN LOGOUT_ON_HOURS LOGOUT_ON_MINUTES DEFAULT_AUTHENTICATION_METHOD PASSWORD_LOGIN_ENABLED CAS_ENABLED CAS_BASE_URL CAS_LOGIN_URL CAS_VALIDATE_URL SAML_ENABLED SAML_PROVIDER SAML_ENTRYPOINT SAML_ISSUER SAML_CERT SAML_IDPSLO_REDIRECTURL SAML_PRIVATE_KEYFILE SAML_PUBLIC_CERTFILE SAML_IDENTIFIER_FORMAT SAML_LOCAL_PROFILE_MATCH_ATTRIBUTE SAML_ATTRIBUTES ORACLE_OIM_ENABLED RESULTS_PER_PAGE WAIT_SPINNER NODE_OPTIONS LDAP_AUTHENTIFICATION_PASSWORD_FILE OAUTH2_SECRET_FILE MAIL_SERVICE_PASSWORD_FILE MONGO_PASSWORD_FILE S3 S3_SECRET_FILE OAUTH2_AUTO_REGISTRATION OAUTH2_ADMIN_GROUPS NOTIFY_ON_ASSIGN LDAP_BACKGROUND_SYNC_DISABLE_NONEXISTANT_USERS LDAP_SYNC_ORGANIZATIONS LDAP_SYNC_ORGANIZATIONS_GROUPS LDAP_SYNC_TEAMS LDAP_SYNC_TEAMS_GROUPS"
 
 #-------------------- CHANGE STREAMS --------------------
 # https://forums.meteor.com/t/meteor-3-5-beta-change-streams-performance-improvements/64461#change-streams-setup-3
@@ -114,6 +114,10 @@ DESCRIPTION_DATABASE="Database backend: 'mongodb' (default) or 'ferretdb' (Ferre
 DEFAULT_DATABASE="mongodb"
 KEY_DATABASE="database"
 
+DESCRIPTION_WEKAN_FERRETDB_OPLOG="FerretDB reactivity mode. 'false' (default) = polling only, which keeps FerretDB CPU low; 'true' = tail the FerretDB OpLog (real-time), but on the SQLite backend the tail can pin FerretDB CPU and stall loading (wekan/wekan#6503). Only applies when database=ferretdb. Force polling: snap set ${SNAP_NAME} wekan-ferretdb-oplog=false ; opt into OpLog: snap set ${SNAP_NAME} wekan-ferretdb-oplog=true"
+DEFAULT_WEKAN_FERRETDB_OPLOG="false"
+KEY_WEKAN_FERRETDB_OPLOG="wekan-ferretdb-oplog"
+
 DESCRIPTION_CADDY_ENABLED="Enable caddy service (caddy - Every Site on HTTPS) personal license for non-commercial use only, see https://caddyserver.com/products/licenses . Set to 'true' to enable caddy\n\t\tcaddy settings are handled through $SNAP_COMMON/Caddyfile"
 DEFAULT_CADDY_ENABLED="false"
 KEY_CADDY_ENABLED="caddy-enabled"

snap-src/bin/wekan-control

@@ -86,20 +86,22 @@ else
     fi
 fi
 
-# #6480/#6481: FerretDB v1 now ships an OpLog (auto-created capped local.oplog.rs
-# + replica-set hello handshake — see bin/ferretdb-control, which launches with
-# --repl-set-name). So by default WeKan's Meteor TAILS the OpLog instead of
-# poll-and-diff, which is the main fix for FerretDB's high CPU on busy boards.
-# Enabled by default; set WEKAN_FERRETDB_OPLOG=false (snap set) to force the old
-# polling-only behaviour if OpLog tailing ever misbehaves.
+# #6503/#6480/#6481: FerretDB v1 CAN tail an OpLog (auto-created capped
+# local.oplog.rs + replica-set hello handshake — see bin/ferretdb-control), but
+# on the SQLite backend Meteor's tailable+awaitData OpLog tail keeps FerretDB CPU
+# pinned (reporters saw ~190–390% even idle; deleting local.sqlite* dropped it to
+# ~10%), and a struggling tail also shows up as "oplog catching up took too long"
+# and stalls board/login loading. So the DEFAULT is now POLLING ONLY. With the
+# calmer polling defaults (#6467/#6468) and the SQLite pragma tuning, poll-and-diff
+# is cheap. Opt into OpLog tailing with WEKAN_FERRETDB_OPLOG=true (snap set) if you
+# specifically want it and have verified it stays calm on your data.
 #
-# The polling settings below still apply as the FALLBACK for the queries Meteor
-# cannot observe through the OpLog (and when OpLog is disabled). #6467/#6468: the
-# Meteor defaults (re-poll 50 ms after any write, at least every 10 s) hammered
-# FerretDB; calmer defaults re-poll at most every 2 s / 30 s. Own writes still
-# show instantly (local cache); only OTHER users' changes wait up to ~2 s.
+# The polling settings below apply either way (they are also the OpLog fallback).
+# #6467/#6468: the Meteor defaults (re-poll 50 ms after any write, at least every
+# 10 s) hammered FerretDB; calmer defaults re-poll at most every 2 s / 30 s. Own
+# writes still show instantly (local cache); only OTHER users' changes wait ~2 s.
 if [ "ferretdb" = "${DATABASE}" ]; then
-    WEKAN_FERRETDB_OPLOG="${WEKAN_FERRETDB_OPLOG:-true}"
+    WEKAN_FERRETDB_OPLOG="${WEKAN_FERRETDB_OPLOG:-false}"
     if [ "true" = "${WEKAN_FERRETDB_OPLOG}" ]; then
         if [ -z "${MONGO_OPLOG_URL}" ]; then
             REPL_SET_NAME="${WEKAN_FERRETDB_REPL_SET:-rs0}"

tests/ferretdbOplog.test.cjs

@@ -1,15 +1,17 @@
 'use strict';
 
-// Guards for the FerretDB OpLog + data-reduction changes (#6480/#6481).
+// Guards for the FerretDB OpLog + data-reduction changes (#6503/#6480/#6481).
 //
-// With FerretDB there is no MongoDB oplog, so Meteor observed every live query by
-// poll-and-diff, pinning FerretDB CPU at 100-390% on busy boards. FerretDB v1 now
-// ships an OpLog (auto-created capped local.oplog.rs + replica-set hello
-// handshake), so every FerretDB launch path must (a) start ferretdb with
-// --repl-set-name so the OpLog exists and the server advertises the replica set,
-// and (b) point WeKan at it via MONGO_OPLOG_URL so Meteor TAILS the OpLog instead
-// of polling — with a WEKAN_FERRETDB_OPLOG kill-switch back to polling. Also
-// asserts the oversized activity page was trimmed.
+// FerretDB v1 CAN tail an OpLog (auto-created capped local.oplog.rs + replica-set
+// hello handshake), but on the SQLite backend the tailable+awaitData tail pins
+// FerretDB CPU (~190-390% even idle) and a struggling tail stalls loading
+// ("oplog catching up took too long"), so #6503 makes POLLING the DEFAULT on every
+// platform. The OpLog machinery still EXISTS and is opt-in via
+// WEKAN_FERRETDB_OPLOG=true. These guards assert: (a) the default is polling only
+// (WEKAN_FERRETDB_OPLOG defaults false; docker-compose reactivity is `polling` and
+// MONGO_OPLOG_URL is commented out); (b) when opted in, ferretdb still launches
+// with --repl-set-name and WeKan is pointed at MONGO_OPLOG_URL with polling as the
+// fallback; and (c) the oversized activity page was trimmed.
 //
 // Run: node tests/ferretdbOplog.test.cjs
 
@@ -58,33 +60,42 @@ test('windows start-wekan.bat passes a repl-set arg to ferretdb.exe', () => {
   assert.ok(/ferretdb\.exe.*%FERRET_REPL_ARG%/.test(src), 'passes it on the ferretdb.exe line');
 });
 
-// ── WeKan points at the OpLog by default (with a kill-switch) ────────────────
-test('snap wekan-control sets MONGO_OPLOG_URL for FerretDB, gated by WEKAN_FERRETDB_OPLOG', () => {
+// ── #6503: FerretDB is POLLING ONLY by default; OpLog is opt-in ──────────────
+test('snap wekan-control: polling-only default for FerretDB, OpLog opt-in', () => {
   const src = read('snap-src/bin/wekan-control');
-  assert.ok(/WEKAN_FERRETDB_OPLOG:-true/.test(src), 'OpLog on by default');
-  assert.ok(/export MONGO_OPLOG_URL="mongodb:\/\/.*\/local\?replicaSet=/.test(src), 'sets oplog url to local?replicaSet');
-  assert.ok(/"true" = "\$\{WEKAN_FERRETDB_OPLOG\}"/.test(src), 'kill-switch check present');
+  assert.ok(/WEKAN_FERRETDB_OPLOG:-false/.test(src), 'OpLog OFF by default (polling only)');
+  assert.ok(/export MONGO_OPLOG_URL="mongodb:\/\/.*\/local\?replicaSet=/.test(src), 'still sets oplog url when opted in');
+  assert.ok(/"true" = "\$\{WEKAN_FERRETDB_OPLOG\}"/.test(src), 'opt-in check present');
 });
-test('bundle start-wekan.sh sets MONGO_OPLOG_URL by default, kill-switch WEKAN_FERRETDB_OPLOG', () => {
+test('snap config registers wekan-ferretdb-oplog (default false, snap set toggle)', () => {
+  const src = read('snap-src/bin/config');
+  assert.ok(/\bWEKAN_FERRETDB_OPLOG\b/.test(src.match(/keys="[^"]*"/)[0]), 'listed in the settings keys');
+  assert.ok(/DEFAULT_WEKAN_FERRETDB_OPLOG="false"/.test(src), 'defaults to false (polling)');
+  assert.ok(/KEY_WEKAN_FERRETDB_OPLOG="wekan-ferretdb-oplog"/.test(src), 'snap set key is wekan-ferretdb-oplog');
+});
+test('bundle start-wekan.sh: polling-only default, OpLog opt-in via WEKAN_FERRETDB_OPLOG=true', () => {
   const src = read('releases/ferretdb/start-wekan.sh');
-  assert.ok(/WEKAN_FERRETDB_OPLOG="\$\{WEKAN_FERRETDB_OPLOG:-true\}"/.test(src), 'default true');
+  assert.ok(/WEKAN_FERRETDB_OPLOG="\$\{WEKAN_FERRETDB_OPLOG:-false\}"/.test(src), 'default false');
   assert.ok(/export MONGO_OPLOG_URL="\$\{MONGO_OPLOG_URL:-mongodb:\/\/\$FERRETDB_LISTEN_ADDR\/local\?replicaSet=/.test(src));
 });
-test('docker wekan-entrypoint.sh sets MONGO_OPLOG_URL, kill-switch present', () => {
+test('docker wekan-entrypoint.sh: polling-only default, OpLog opt-in', () => {
   const src = read('releases/ferretdb/wekan-entrypoint.sh');
-  assert.ok(/WEKAN_FERRETDB_OPLOG="\$\{WEKAN_FERRETDB_OPLOG:-true\}"/.test(src), 'default true');
+  assert.ok(/WEKAN_FERRETDB_OPLOG="\$\{WEKAN_FERRETDB_OPLOG:-false\}"/.test(src), 'default false');
   assert.ok(/export MONGO_OPLOG_URL="\$\{MONGO_OPLOG_URL:-mongodb:\/\/\$FERRETDB_LISTEN_ADDR\/local\?replicaSet=/.test(src));
 });
-test('docker-compose prefers oplog reactivity + sets MONGO_OPLOG_URL (uncommented)', () => {
+test('docker-compose: polling-only reactivity by default + MONGO_OPLOG_URL commented out', () => {
   const src = read('docker-compose.yml');
-  assert.ok(/METEOR_REACTIVITY_ORDER=oplog,polling/.test(src), 'reactivity prefers oplog then polling');
-  // the MONGO_OPLOG_URL line must be active (not commented out)
-  assert.ok(/\n\s*- MONGO_OPLOG_URL=mongodb:\/\/ferretdb:27017\/local\?replicaSet=rs0/.test(src),
-    'MONGO_OPLOG_URL must be an active env line, not commented');
+  assert.ok(/METEOR_REACTIVITY_ORDER=polling/.test(src), 'reactivity is polling by default');
+  // the MONGO_OPLOG_URL line must NOT be active (commented out for polling-only)
+  assert.ok(!/\n\s*- MONGO_OPLOG_URL=mongodb:\/\/ferretdb:27017/.test(src),
+    'MONGO_OPLOG_URL must NOT be an active env line');
+  assert.ok(/#\s*-\s*MONGO_OPLOG_URL=mongodb:\/\/ferretdb:27017\/local\?replicaSet=rs0/.test(src),
+    'the opt-in MONGO_OPLOG_URL line is present but commented');
 });
-test('windows start-wekan.bat sets MONGO_OPLOG_URL under the kill-switch', () => {
+test('windows start-wekan.bat: polling-only default, OpLog opt-in', () => {
   const src = read('releases/ferretdb/start-wekan.bat');
-  assert.ok(/WEKAN_FERRETDB_OPLOG%"=="true"/.test(src), 'gated on the kill-switch');
+  assert.ok(/set "WEKAN_FERRETDB_OPLOG=false"/.test(src), 'default false');
+  assert.ok(/WEKAN_FERRETDB_OPLOG%"=="true"/.test(src), 'opt-in branch gated on =true');
   assert.ok(/MONGO_OPLOG_URL=mongodb:\/\/127\.0\.0\.1:27017\/local\?replicaSet=/.test(src));
 });
 
@@ -107,7 +118,7 @@ test('#6498: polling mode UNSETS MONGO_OPLOG_URL so Meteor never tails the OpLog
 // OpLog only when tailing actually works, else polling — so a broken/absent
 // OpLog never stops WeKan starting. Every FerretDB launch path must keep polling
 // last (like the MongoDB "changeStreams,oplog,polling" default).
-test('every FerretDB launch path keeps polling as the reactivity fallback (oplog,polling)', () => {
+test('every FerretDB launch path always has polling (as the default or the OpLog fallback)', () => {
   for (const f of [
     'snap-src/bin/wekan-control',
     'releases/ferretdb/start-wekan.sh',
@@ -117,8 +128,10 @@ test('every FerretDB launch path keeps polling as the reactivity fallback (oplog
     'sandstorm-src/start.js',
   ]) {
     const src = read(f);
-    assert.ok(/METEOR_REACTIVITY_ORDER[^\n]*oplog,polling/.test(src),
-      `${f} must set METEOR_REACTIVITY_ORDER to oplog,polling (polling fallback)`);
+    // Default is now polling-only (METEOR_REACTIVITY_ORDER=polling); the opt-in
+    // OpLog branch keeps polling last (oplog,polling). Both contain "polling".
+    assert.ok(/METEOR_REACTIVITY_ORDER[^\n]*polling/.test(src),
+      `${f} must keep polling available in METEOR_REACTIVITY_ORDER`);
   }
 });