Excessive memory growth by one request

[leonskim]

I deployed parse-server on Heroku + mLab and the latency between them is fine.

However, when the server receives a request from the mobile client, it's very slow so I checked the memory usage, it wouldn't stop growing. Th response time slows down and eventually timeout.

I tried Garbage Collect manually by global.gc() and --gc_interval=100. It seemed to help for a while but it's not the solution. If I set the client to point to Parse.com backend, it works perfectly without timeout.

2016-03-28T22:09:19.644662+00:00 app[web.1]: { rss: 63250432, heapTotal: 45734944, heapUsed: 43474792 }
2016-03-28T22:09:25.316197+00:00 heroku[router]: at=info method=POST path="/parse/classes/Like" host=XXX.herokuapp.com request_id=4881af9a-976d-4f96-ad93-7e4532df1692 fwd="xxx.xxx.xxx.xxx" dyno=web.1 connect=1ms service=343ms status=200 bytes=532
2016-03-28T22:09:26.540261+00:00 heroku[router]: at=info method=PUT path="/parse/classes/Profile/YB23FPPzdN" host=XXX.herokuapp.com request_id=ee21080c-f9ba-488d-9c37-3e6929cf2a35 fwd="xxx.xxx.xxx.xxx" dyno=web.1 connect=0ms service=1171ms status=200 bytes=559
2016-03-28T22:09:26.532459+00:00 heroku[router]: at=info method=POST path="/parse/classes/Trainer" host=XXX.herokuapp.com request_id=bba43f05-7886-4130-b761-3965f41bba33 fwd="xxx.xxx.xxx.xxx" dyno=web.1 connect=0ms service=1473ms status=200 bytes=134054
2016-03-28T22:09:26.701823+00:00 heroku[router]: at=info method=POST path="/parse/classes/Like" host=XXX.herokuapp.com request_id=fcc5a5f4-2253-4652-8df9-67af3a8ed9d3 fwd="xxx.xxx.xxx.xxx" dyno=web.1 connect=0ms service=1120ms status=200 bytes=532
2016-03-28T22:09:30.499274+00:00 app[web.1]: { rss: 73048064, heapTotal: 52941856, heapUsed: 50060024 }
2016-03-28T22:09:41.265809+00:00 app[web.1]: { rss: 94691328, heapTotal: 67388960, heapUsed: 64001176 }
2016-03-28T22:09:55.292990+00:00 app[web.1]: { rss: 106524672, heapTotal: 81836064, heapUsed: 77893432 }
2016-03-28T22:09:55.580280+00:00 heroku[router]: at=error code=H12 desc="Request timeout" method=POST path="/parse/classes/Video" host=XXX.herokuapp.com request_id=afa3d66b-867d-442c-be72-4b04529f4e49 fwd="xxx.xxx.xxx.xxx" dyno=web.1 connect=0ms service=30004ms status=503 bytes=0

parse-server handles queries differently than Parse.com? or the other reasons?

I'd appreciate any suggestion or comment. Thanks.

[gfosco]

Do you have cloud code involved? Have not had any reports of this.

[leonskim]

@gfosco Yes but if I get rid of the cloud code and run, the result is the same.

[gfosco]

Interesting.... but there's not enough information here to go on. What kind of query, what level of usage, how are you monitoring, what else is changed, etc?

[leonskim]

@gfosco I just added the log in the post (cloud code isn't running) and I haven't changed anything but the cloud code so far. I run both parse-server-example and my own parse-server but it's all the same.

[gfosco]

I'd love to help, but there is no actionable information here and it does not look like others are experiencing this problem.

[woodardj]

#2482, #2434