jrconlin/pushgo

server dies after a while with 'too many open files' message

Closed this issue · 6 comments

edwindotcom commented

not sure what this is but my server dies at some point with this. The only STR is run the server and run the API test 20 times.

2013/06/14 15:02:51 [1] worker: ERROR: RegisterAppID failed dial tcp 127.0.0.1:11211: too many open files
2013/06/14 15:02:51 [3] worker: Reg result:: dial tcp 127.0.0.1:11211: too many open files
2013/06/14 15:02:51 [3] worker: Sending error dial tcp 127.0.0.1:11211: too many open files
2013/06/14 15:02:51 [1] worker: Websocket Error EOF
2013/06/14 15:02:51 [3] server: Server Handling command {%!s(int=6) }
2013/06/14 15:02:51 [3] server: Server cleanup...
2013/06/14 15:02:51 [3] server: Cleaning up socket reg_noshake_uaid_1
2013/06/14 15:02:51 [3] timer: Socket connection terminated
2013/06/14 15:02:51 [4] main: Returning Result %!s(int=0)
2013/06/14 15:02:51 open config.ini: too many open files
exit status 1

jrconlin commented

I patched in a force close of the config file. That should solve the "too many open files" issue in most cases. I've not been able to replicate it, but then, I'm guessing that I'm not hitting the server as hard.

Please close the bug if you no longer see the same issue.

edwindotcom commented

I'm still reproducing this bug. I just did a git pull, about an hour ago and restarted server.

edwindotcom commented

I was on the wrong branch - I've switched to dev

edwindotcom commented

I'm still repro'ing this on master - If you run the mod_test/test_push_api.py about 10 times, you'll hit it.

edwindotcom commented

I can repro this when i use Loads to send hello and register 200 times. Test is here:
https://github.com/edmoz/loads/blob/issue-21-test-case/loads/examples/test_pushgo.py

run this: loads-runner loads.examples.test_pushgo.TestPushgo.test_reg -c 200 -u 1

Started: 2013-07-10 19:46:17.666200
Duration: 0.67 seconds
Approximate Average RPS: 0
Average request time: 0.00s
Opened web sockets: 0
Bytes received via web sockets : 26500

Success: 125
Errors: 75
Failures: 0

server:

2013/07/10 13:01:07 [3] server: Generated Endpoint map[uaid:1449NZ5ATEAF5VVW channelID:abc token:1449NZ5ATEAF5VVW.abc endpoint:http://localhost:8080/update/1449NZ5ATEAF5VVW.abc]
2013/07/10 13:01:07 [4]   main: Returning Result %!s(int=200)
2013/07/10 13:01:07 [4] worker: Server returned {%!s(int=200) map[messageType:register channelID:abc uaid:1449NZ5ATEAF5VVW status:%!s(int=200) pushEndpoint:http://localhost:8080/update/1449NZ5ATEAF5VVW.abc]}
2013/07/10 13:01:07 [4] worker: sending response map[messageType:register uaid:1449NZ5ATEAF5VVW status:%!s(int=200) channelID:abc pushEndpoint:http://localhost:8080/update/1449NZ5ATEAF5VVW.abc]
2013/07/10 13:01:07 open config.ini: too many open files
exit status 1
edwindotcom commented

might be mac limits - closing for now as this doesn't happen with AWS server.