Tyk startup and die

Imported Google Group message. Original thread at: Redirecting to Google Groups Import Date: 2016-01-19 21:02:00 +0000.
Sender:Vladimir Ustinov.
Date:Monday, 6 October 2014 14:48:04 UTC+1.

Hello
We give tyk a try. And have no success with it. When started it make something like stress test with redis and consume memory. When memory running out it killed with OOM killer.
Thats all our experience for now :slight_smile:

This is tyk.conf:
{
“listen_port”:5000,
“secret”:“very long secret”,
“template_path”:“/etc/tyk/templates”,
“use_db_app_configs”:true,
“app_path”:“/etc/tyk/apps/”,
“storage”:{
“type”:“redis”,
“host”:“localhost”,
“port”:6379,
“username”:“user”,
“password”:“very short password”
},
“enable_analytics”:true,
“analytics_config”:{
“type”:“mongo”,
“csv_dir”:“/tmp”,
“mongo_url”:“mongodb://localhost/tyk_analytics”,
“mongo_db_name”:“tyk_analytics”,
“mongo_collection”:“tyk_analytics”,
“purge_delay”:0
}
}

We tried this in Amazon Linux AMI 2014.09. Maybe we miss something?
Advance thank you

Imported Google Group message.
Sender:Martin Buhr.
Date:Monday, 6 October 2014 15:09:39 UTC+1.

Hi there,

The first thing I would suggest is changing purge_delay to 30 (or higher, maybe 60), this will cause Tyk to purge Redis every X seconds instead of 0 which is causing it to start many goroutines over and over again faster than the GC can keep up and may lead to OOM issue (fixed in master branch). If you are not collecting analytics, I’d set enable_analytics to ‘false’, then you don’t need the purge and you’ll lower the overhead on your Redis instance.

If that doesn’t help, could you let me know the following to help diagnose:
Are you using Tyk master branch or the 1.1 release?
Are you getting any output from Tyk? If so, could you post the log output when it dies?
Thanks,
Martin

Imported Google Group message.
Sender:Martin Buhr.
Date:Tuesday, 7 October 2014 09:41:18 UTC+1.

Just to close this issue - the unlicensed dashboard will not show data gathered for the day, an will only show it at a today-1 rate, this is a bug and has been raised in our issue tracker.