Hi @scullxbones , could you please help me figure out the issue with tags I'm facing - The tags aren't written even though i've wrapped the event inside 'akka.persistence.journal.Tagged'. Is there any specific configuration for tags in the journal for this which i'm missing?
I'm using akka 2.6.10 and java serialization for now.
ReadJournal works absolutely fine - i'm able to get events by tag if i add tags manually to the above document.
Hi @alokshukla121:matrix.org -
There's no configuration to turn on. But it looks like there may be a configuration setting to turn off.
Looking at: https://github.com/scullxbones/akka-persistence-mongo/blob/master/common/src/main/scala/akka/contrib/persistence/mongodb/MongoDataModel.scala I see your type is considered to be Legacy. Do you have legacy serialization enabled? That may play havoc with tagging.
Legacy serialization is really only for projects that have used this library for many years, I think it's from Akka 2.3 series :)
Hi @scullxbones , I switched off java serialization and used protobuf this time but still i see the type as 'repr' in the database. Moreover, i'm not able to wrap my event to 'akka.persistence.journal.Tagged' anymore as akka attempts to use java serialization for 'Tagged' which is turned off.
@scullxbones: Sorry i missed your point and after checking for a flag to toggle legacy serialization, i could find this in reference.conf
this was set to true earlier and hence tags were not getting added. After i switched it to false, it works as expected. Thanks a lot for your help!
@scullxbones: How do we debug the journal as to what's going on with replayed events? How do i enable logging to see what's going wrong? I've migrated the existing events to from Cashbah schema to the schema of your journal but i think in the process i'm missing something and only 'RecoveryCompleted' event is being replayed for all the migrated messages/events. Is it possible to change default 'JournlingFieldNames'?
No it's not possible to change the Journaling field names without a fork of the project.
Regarding any events, make sure that your message handler in the persistent actor has a catch all section, or that you are catching all possible infrastructure messages, like
RecoveryFailed. Logging is done via slf4j, but it is fairly limited, preferring instead to communicate failures through akka's plugin interface, e.g. Try[Unit]
1 reply
@scullxbones: I'm facing a strange problem with eventsByTag query - sometimes it works and sometimes it doesn't. I've compared the offset and figured out that the offset stored at projector was of earlier time than the recently added events. It doesn't even show any error.
final EventsByTagQuery readJournal = PersistenceQuery.get(ACTOR_SYSTEM)
.getReadJournalFor(EventsByTagQuery.class, MongoReadJournal.Identifier());
final Source<EventEnvelope, NotUsed> source = readJournal.eventsByTag(getTagName(),
getCurrentOffset());
source.runForeach(this::processEvent, ACTOR_SYSTEM);
.getReadJournalFor(EventsByTagQuery.class, MongoReadJournal.Identifier());
final Source<EventEnvelope, NotUsed> source = readJournal.eventsByTag(getTagName(),
getCurrentOffset());
source.runForeach(this::processEvent, ACTOR_SYSTEM);
@scullxbones: Thanks for the reference! I did check #370 and other related issues but all those possibly relate to missing events due to time difference within a second. In my case, events aren't appearing even when the difference is around 20 days in offset - the read side processor has offset pertaining to 6th July, 2021 while the events present in messages(the collection i use to store all events) are as latest as 26nd July, 2021. In my local with embedded tomcat in spring boot it works fine but when i deploy the same application in Azure App Service, the events aren't received. Neither i see any error logs. Is it to do with the fact all events including missing 20 days are in messages collection not realtime? Should i copy the unprocessed events to realtime collection? I had migrated the events from previous schema to the one with official mongo driver supports and in the process i copied all the events to same collection.
@alokshukla121:matrix.org do you have any live events at all? The realtime collection is a capped collection that is written to when events are written to the journal. It's a kind of event bus to allow for distributed listeners (read processors) without requiring a cluster. It can be disabled by configuration, do you have this enabled?
realtime-enable-persistence - is enabled by default ... https://github.com/scullxbones/akka-persistence-mongo/blob/master/common/src/main/resources/reference.conf#L34
Hi @jsarrelli - the plugin respects the calls that are made from the library. If multiple messages are passed via asyncWriteMessages then they are stored as a batch in a single document in mongodb. This is because the unit of atomicity for mongodb is the document
This should be triggered by this documented usage or if you prefer classic
Hi @jsarrelli - i have not looked into that yet. If it uses existing plugins as is, then yes. In general, the team has been good about reuse and backward compatibility with new persistence features. But I'll have to research it more to know. Don't feel like you need to wait on me though. A proof of concept could answer fairly quickly i think.
Also look at scullxbones/akka-persistence-mongo#214