Diagnosing issues with Deployed Meteor App


#1

I am hosting a Meteor Application on Azure and over time the publications/methods/etc… seem to get slower and slower to the point of freezing. I don’t really have any idea where to even approach how to deal with this. Any suggestions on how I can begin to diagnose what could be the issue?

Thanks!


#2

The only way to clear these issues is going for a redeployment of the application. Restarting the app does not seem to help much at all.


#3

I’m facing a similar battle. My game is www.StarCommanderOnline.com where players have been hit hard with lag. its been a lot to figure out.

Currently I’m trying to better manage users and manually resubscribe them on poor connections. I pop a little load screen.

Are you running an OpLog enabled MongoDB server?


#4

Yes, I have a paid account on MLab and am connecting with the OpLog URL. That being said, I am kind of leaning towards this being a memory issue?

I run the Heap Memory Screenshot and it grows from about 27mb on load to about 34 after starting to insert into the database, but it doesn’t seem to go much higher than that (got up to about 38mb), so I don’t know :frowning:


#5

And you’ve got your settings.json entered correctly with the like, 3 entries for OpLog?

How many entries are you putting in?

SCO as I linked before is pretty hardcore on CPU.

My OpLog server is now 2 cores, 16 gb ram. My Node server is only 1 CPU, 4 gb ram.

The amount of optimization I’ve had to do is staggering. I’m returning 30 documents with 20 fields each. Managing their subscription and publication is something that’s JUST as important as oplog.

Where do YOU think the weak point is?


#6

I only have 1 entry for an OpLog url. What do you mean? I’ll check some of the subscriptions, but personally I think it is a memory leak (trying to figure out how to diagnose it).


#7

MONGO_URL and MONGO_OPLOG_URL should point to all Replica Set members, see this article


#8

I recommend Meteor shelling in and then running v8-profiler to get a CPU dump of the server