Error: Removed non-existent document about meteor-reactive-publish HOT 12 CLOSED

Hm, can you make a small reproduction using directcollection which is a direct connection to Mongo.

I am unsure why would this happen though with this package. Really strange. I do not know that it would rely on changes coming from inside.

from meteor-reactive-publish.

Thanks @mitar, so you want me to (a) replace the mongo collection in the reactive publish with the directcollection or (b) have meteor make the change with directcollection and see how the reactive publish behaves?

As mentioned above, I faced the same issue with redis-oplog (issue raised but I volunteered to get to the bottom of it).

from meteor-reactive-publish.

(b). So direct collection directly changes the Mongo. I am curious if you can replicate with that. Then we know that there is nothing fancy with redis-oplog integration (or hooking machinsm) or something like that.

from meteor-reactive-publish.

Ok, but this is similar to what I did already. I altered the DB externally to my app using mongoclient [Edit: fixed app name] (running as an external app, so for all practical purposes, independent from the app we are testing), and the reactive publish triggered this error. Not using redis-oplog at the moment, just wanted to highlight that we face the same issue there.

from meteor-reactive-publish.

Ok, so we found what was going on (at least I believe we do). We had in our reactive publish a reactive find which contained extra fields, that seemed to trigger reruns of the publish -- looks like a race condition.

What is odd is that line 36 in server.coffee is within a try, yet triggers an error.

I'll go ahead and close this so not to litter, but it's something that could cause issues with redis-oplog. I'll investigate and revert if need be.

Thanks @mitar for your continued courtesy.

from meteor-reactive-publish.

Yes, I wanted to have a reproduction without redis-oplog, and a command from inside which changes the meteor so that I could investigate. I understand that this is probably the same as changing it manually. But it is good to have it in a code because it then possible to create a test to test in the future.

from meteor-reactive-publish.

What is odd is that line 36 in server.coffee is within a try, yet triggers an error.

Sometimes I realized that numbers do not match always. All this preprocesing of CoffeeScript/JavaScript can move things around and make lines wrong.

But still, why would extra fields throw such an error?

from meteor-reactive-publish.

But still, why would extra fields throw such an error?

The only unscientific reason I can think of is a race condition somewhere, that causes the reactive publishing to be triggered twice (or more), with the second call failing.

I wanted to have a reproduction without redis-oplog, and a command from inside which changes the meteor so that I could investigate.

Understood. It's high on our to-do's as are supporting @theodorDiaconu in getting redis-oplog to work with reactive publishing. I'll work on getting you a test repo.

from meteor-reactive-publish.

Has this ever be resolved? @theodorDiaconu?

from meteor-reactive-publish.

@mitar not yet :(

from meteor-reactive-publish.

So is this error still happening?

from meteor-reactive-publish.

Closing because of no activity.

from meteor-reactive-publish.