powersync failed host lookup

[DunklerPhoenix]

Priority/Impact

High (critical issue, blocks workflow)

Description

The powersync server, wger, etc are online.
The flutter app can connect to wger (connection activity in logs), but powersync always says name not resolved. I can try any dns server it's always the same. Its like a timeout error.

Even my browser, termux and co can connect to my domain without problems.

server: self-hosted
connected: true, connecting: false, downloading: true, uploading: true
last successful sync: 2026-08-27T23:02:03.877Z
pending uploads: 2
checkpoint request: target 9223372036854775807, seen -, applied -
local buckets: last op 326969, ops since checkpoint 0, oplog rows 8186
download progress: 0 / 0
error (Connection error): ClientException with SocketException: Failed host lookup: 'wger.domain.cloud' (OS Error: No address associated with hostname, errno = 7), uri=https://wger.domain
cloud/ps/sync/stream

Server version

2.7.0a2+git6050749

Mobile app version

2.0.5

Edit: If I open the url in the browser I get this error, but I think it should be normal, right?
{"error":{"code":"PSYNC_S2002","status":404,"description":"The path does not exist on this server","details":"The path GET \"/sync/stream\" does not exist on this server","severity":"info","name":"RouteNotFound"}}

[DunklerPhoenix]

I think it is caused by a failing powersync server. The server throws the following error over and over again.
Why the data are so big in the requests? I'm the only user on wger at the moment

powersync-1  | {"level":"info","message":"[powersync_1_8ed1] Initial replication already done","timestamp":"2026-08-29T00:59:10.494Z"}
powersync-1  | {"level":"error","message":"[powersync_1_8ed1] Replication error invalid memory alloc request size 1233909616","stack":"PgError.XX000: invalid memory alloc request size 1233909616\n    at PgConnection.stream (file:///app/node_modules/.pnpm/pgwire@0.8.1/node_modules/pgwire/mod.js:331:15)\n    at process.processTicksAndRejections (node:internal/process/task_queues:104:5)\n    at async ReplicationStream._start (file:///app/node_modules/.pnpm/pgwire@0.8.1/node_modules/pgwire/mod.js:1486:40)\n    at async PgoutputReader.decodeStream (file:///app/node_modules/.pnpm/pgwire@0.8.1/node_modules/pgwire/mod.js:1577:22)\n    at async file:///app/modules/module-postgres/dist/replication/WalStream.js:910:30\n    at async PostgresSyncRulesStorage.startBatch (file:///app/modules/module-postgres-storage/dist/storage/PostgresSyncRulesStorage.js:181:9)\n    at async WalStream.streamChangesInternal (file:///app/modules/module-postgres/dist/replication/WalStream.js:897:9)\n    at async WalStream.streamChanges (file:///app/modules/module-postgres/dist/replication/WalStream.js:835:13)\n    at async WalStream.replicate (file:///app/modules/module-postgres/dist/replication/WalStream.js:808:13)\n    at async WalStreamReplicationJob.replicateOnce (file:///app/modules/module-postgres/dist/replication/WalStreamReplicationJob.js:117:13)","timestamp":"2026-08-29T00:59:25.198Z"}
powersync-1  | {"code":"XX000","file":"mcxt.c","level":"error","line":1194,"message":"[powersync_1_8ed1] cause invalid memory alloc request size 1233909616","routine":"repalloc","severity":"ERROR","severityEn":"ERROR","timestamp":"2026-08-29T00:59:25.198Z"}
powersync-1  | {"level":"warn","message":"Query error, retrying postgres query late","stack":"Error: postgres query late\n    at PgPool.stream (file:///app/node_modules/.pnpm/pgwire@0.8.1/node_modules/pgwire/mod.js:107:13)\n    at stream.next (<anonymous>)\n    at PgResult.fromStream (file:///app/node_modules/.pnpm/pgwire@0.8.1/node_modules/pgwire/mod.js:1063:22)\n    at PgPool.query (file:///app/node_modules/.pnpm/pgwire@0.8.1/node_modules/pgwire/mod.js:103:21)\n    at Module.retriedQuery (file:///app/libs/lib-postgres/dist/utils/pgwire_utils.js:37:29)\n    at sendKeepAlive (file:///app/modules/module-postgres/dist/replication/WalStream.js:34:24)\n    at WalStream.abort_signal.addEventListener.once (file:///app/modules/module-postgres/dist/replication/WalStream.js:84:33)\n    at [nodejs.internal.kHybridDispatch] (node:internal/event_target:843:20)\n    at AbortSignal.dispatchEvent (node:internal/event_target:776:26)\n    at runAbort (node:internal/abort_controller:547:10)","timestamp":"2026-08-29T00:59:25.199Z"}
powersync-1  | {"level":"warn","message":"[powersync_1_8ed1] Failed to ping connection postgres query late","stack":"Error: postgres query late\n    at PgPool.stream (file:///app/node_modules/.pnpm/pgwire@0.8.1/node_modules/pgwire/mod.js:107:13)\n    at stream.next (<anonymous>)\n    at PgResult.fromStream (file:///app/node_modules/.pnpm/pgwire@0.8.1/node_modules/pgwire/mod.js:1063:22)\n    at PgPool.query (file:///app/node_modules/.pnpm/pgwire@0.8.1/node_modules/pgwire/mod.js:103:21)\n    at Module.retriedQuery (file:///app/libs/lib-postgres/dist/utils/pgwire_utils.js:37:29)\n    at process.processTicksAndRejections (node:internal/process/task_queues:104:5)\n    at async sendKeepAlive (file:///app/modules/module-postgres/dist/replication/WalStream.js:34:5)","timestamp":"2026-08-29T00:59:25.199Z"}
powersync-1  | {"level":"info","message":"[powersync_1_8ed1] Initial replication already done","timestamp":"2026-08-29T00:59:55.241Z"}

[rolandgeider]

ugh, that's a new one...

This error seems to come from postgres (PgError.XX000: invalid memory alloc request size 1233909616), not directly from the sync service. Can you run docker compose exec db psql -U wger -c "SELECT version();" and docker compose logs db --tail 200 | grep -B2 -A5 "invalid memory alloc" and see if there's something interesting there?

Something that will possibly work is simply nuking powersync's replication slot and letting it recreate it from scratch

docker compose stop powersync
docker compose exec db psql -U wger -c "SELECT pg_drop_replication_slot('powersync_1_8ed1');"
docker compose start powersync

[DunklerPhoenix]

docker compose exec db psql -U wger -c "SELECT version();"
PostgreSQL 15.13 (Debian 15.13-1.pgdg120+1) on x86_64-pc-linux-gnu, compiled by gcc (Debian 12.2.0-14) 12.2.0, 64-bit

docker compose logs db --tail 200 | grep -B2 -A5 "invalid memory alloc"

postgres-1  | 2026-08-29 13:35:15.568 CEST [2310938] LOG:  starting logical decoding for slot "powersync_1_8ed1"
postgres-1  | 2026-08-29 13:35:15.568 CEST [2310938] DETAIL:  Streaming transactions committing after 16/EDB67608, reading WAL from 16/EDB00A30.
postgres-1  | 2026-08-29 13:35:15.568 CEST [2310938] STATEMENT:  START_REPLICATION SLOT "powersync_1_8ed1" LOGICAL 0/0 ("proto_version" '1',"publication_names" 'powersync',"messages" 'true')
postgres-1  | 2026-08-29 13:35:15.574 CEST [2310938] LOG:  logical decoding found consistent point at 16/EDB00A30
postgres-1  | 2026-08-29 13:35:15.574 CEST [2310938] DETAIL:  Logical decoding will begin using saved snapshot.
postgres-1  | 2026-08-29 13:35:15.574 CEST [2310938] STATEMENT:  START_REPLICATION SLOT "powersync_1_8ed1" LOGICAL 0/0 ("proto_version" '1',"publication_names" 'powersync',"messages" 'true')
postgres-1  | 2026-08-29 13:35:26.149 CEST [2310938] ERROR:  invalid memory alloc request size 1233909616
postgres-1  | 2026-08-29 13:35:26.149 CEST [2310938] STATEMENT:  START_REPLICATION SLOT "powersync_1_8ed1" LOGICAL 0/0 ("proto_version" '1',"publication_names" 'powersync',"messages" 'true')
postgres-1  | 2026-08-29 13:35:57.161 CEST [2311009] LOG:  starting logical decoding for slot "powersync_1_8ed1"
postgres-1  | 2026-08-29 13:35:57.161 CEST [2311009] DETAIL:  Streaming transactions committing after 16/EDB67608, reading WAL from 16/EDB00A30.
postgres-1  | 2026-08-29 13:35:57.161 CEST [2311009] STATEMENT:  START_REPLICATION SLOT "powersync_1_8ed1" LOGICAL 0/0 ("proto_version" '1',"publication_names" 'powersync',"messages" 'true')
postgres-1  | 2026-08-29 13:35:57.183 CEST [2311009] LOG:  logical decoding found consistent point at 16/EDB00A30
postgres-1  | 2026-08-29 13:35:57.183 CEST [2311009] DETAIL:  Logical decoding will begin using saved snapshot.
postgres-1  | 2026-08-29 13:35:57.183 CEST [2311009] STATEMENT:  START_REPLICATION SLOT "powersync_1_8ed1" LOGICAL 0/0 ("proto_version" '1',"publication_names" 'powersync',"messages" 'true')
postgres-1  | 2026-08-29 13:36:16.718 CEST [2311009] ERROR:  invalid memory alloc request size 1233909616
postgres-1  | 2026-08-29 13:36:16.718 CEST [2311009] STATEMENT:  START_REPLICATION SLOT "powersync_1_8ed1" LOGICAL 0/0 ("proto_version" '1',"publication_names" 'powersync',"messages" 'true')

(I will not nuke the data yet. If you want anything from the database (even an complete backup), just ask.) There are not really sensitive data in it)

[simolus3]

(I'm on the PowerSync team and stumbled upon this issue) I'm not sure why exactly Postgres wants to allocate this much memory here. Could you try running this query to check how far your PowerSync container is lagging behind Postgres here?

A less destructive option is to just drop the replication slot (the PowerSync container needs to be stopped for this): SELECT pg_drop_replication_slot('powersync_1_8ed1');. If you then restart the PowerSync service, it should complain about the replication slot not being there and restart the initial replication from a fresh state.

[DunklerPhoenix]

select slot_name, confirmed_flush_lsn, active, pg_size_pretty(pg_wal_lsn_diff(pg_current_wal_lsn(), restart_lsn)) as lag from pg_replication_slots;

60GB is the answer

[rolandgeider]

@DunklerPhoenix 60GB!!!

it might be that you've run into this pg issue, which was fixed in 15.14, can you update the postgres image and see if it fixes it? I would also drop the replication slow, just in case (note that dropping it doesn't delete any data, the single source of truth are still the wger tables, it will just start the replication again)

[DunklerPhoenix]

Ah yes. Updating fixed the problem
Thank you very much.