Initial Sync loop

[redsky17]

[2019-02-11 10:43:16.041] [db] [error] failed to save state after initial sync: mdb_dbi_open: MDB_DBS_FULL: Environment maxdbs limit reached
[2019-02-11 10:43:16.041] [net] [info] trying initial sync

[AndrewJDR]

I was able to successfully log in with this patch in place! I think you can safely merge this to master.

[redsky17]

I'm glad it worked for you and I'll merge the PR, but I want to leave this issue open. I have a feeling that there's some greater design issue that needs to be addressed.

[rnhmjoj]

I made the mistake of deleting the nheko cache and I can't seem to log in anymore.
I have a different error, though:

[2019-03-19 20:33:47.052] [crypto] [info] creating new olm account
[2019-03-19 20:33:47.105] [crypto] [info] ed25519   : <deleted>
[2019-03-19 20:33:47.105] [crypto] [info] curve25519: <deleted>
[2019-03-19 20:33:47.105] [crypto] [info] generating one time keys
[2019-03-19 20:33:48.324] [net] [info] uploaded 50 signed_curve25519 one-time keys
[2019-03-19 20:33:48.324] [net] [info] trying initial sync
[2019-03-19 20:33:59.237] [net] [error] initial sync error: 2067873536 
[2019-03-19 20:33:59.237] [net] [info] trying initial sync
[2019-03-19 20:34:10.264] [net] [error] initial sync error: 2067873536 M_UNRECOGNIZED
[2019-03-19 20:34:10.264] [net] [info] trying initial sync
[2019-03-19 20:34:21.630] [net] [error] initial sync error: 2067873536 
[2019-03-19 20:34:21.630] [net] [info] trying initial sync
[2019-03-19 20:34:32.990] [net] [error] initial sync error: 2067873536 

[deepbluev7]

This issue could also be related to matrix-org/synapse#4898 . Matrix.org (and probably other servers using federation workers) sends wrong m.read events over federation, which leads to wrong events being sent to clients. Basically you get a json::type_error here /~https://github.com/Nheko-Reborn/mtxclient/blob/5422d281bd8d1ae5db14b5a8e0c4093decbedeb9/lib/structs/responses/sync.cpp#L72 while parsing the timestamp field.

Because there is no handling of parse errors in initialSync, Nheko just loops forever trying to sync.

[rnhmjoj]

I also noted I can't receive any new messages until I restart nheko, so it's not limited to the initial sync.

[deepbluev7]

Well, nheko can't parse some ephemeral events (reading receipts) properly, so I wouldn't be surprised, if that bugs out all of the sync code, i.e. also receiving normal messages,. I get quite a few messages of wrong timestamp fields when I'm already logged in with my workaround (see Nheko-Reborn/mtxclient#10 for my changes).

[rnhmjoj]

I confirm it solved the issue completely. Thank you!

[redsky17]

I think this has been sufficiently resolved at this point.

[bjesus]

I'm getting the same (?) error with the most recent version from git. Can't login at all using Nheko. It used to work fine for me with previous builds.

[2020-09-24 12:42:32.394] [net] [info] initial sync completed
[2020-09-24 12:42:32.867] [db] [info] mark room !XXX:matrix.org as encrypted
[2020-09-24 12:42:33.591] [db] [error] failed to save state after initial sync: mdb_dbi_open: MDB_DBS_FULL: Environment maxdbs limit reached
[2020-09-24 12:42:33.591] [net] [info] trying initial sync
[json.exception.out_of_range.403] key 'device_id' not found
{
  "content": {
    "algorithm": "m.megolm.v1.aes-sha2",
    "ciphertext": "XXX",
    "sender_key": "XXX",
    "session_id": "XXX"
  },
  "event_id": "$XXX",
  "origin_server_ts": 1598530409679,
  "sender": "@chat:XXX.com",
  "type": "m.room.encrypted",
  "unsigned": {
    "age": 2413780228
  }
}

[deepbluev7]

@bjesus : Are you in a lot of rooms like 1000 or so? Or ist that message about device_id not found repeating itself a lot, while you are initial sync looping. In the former case, bumping MAX_DBS should help (we need to increase that anyway, I think), in the latter case that should be easily fixable, but I need to dig into, why this is actually still failing and not caught immediately.

[bjesus]

I have about 250 rooms and 750 direct people, yes. How do I bump MAX_DBS? Is it an environment variable? and what should I set there? Thank you @deepbluev7 !

[deepbluev7]

It's currently a hardcoded variable in here: /~https://github.com/Nheko-Reborn/nheko/blob/master/src/Cache.cpp#L51

Nheko uses multiple dbs per room (direct chats are also a room). Arguably this should be much higher, since we recently added a few addional dbs per room and Nheko should (at least at some point) scale to 10k rooms or more. I'll investigate, what's the reasonable maximum and if this should be configurable by the user.

[bjesus]

Thanks! I've changed it to 38092UL (just guessing) and now Nheko works perfectly.

[deepbluev7]

@bjesus : I'll try to fix that soon. I'm pretty sure you have one of the largest accounts, that ever used Nheko. I'll definitely need to figure out a proper solution to the max db thingy.