Platform: Hp - DL380p Gen8 (12 LFF), 2x Intel Hexa-Core Xeon E5-2620, RAM: 32 GB, Hdd 2x1 TB
Network: 1 Gbit/sec
OS: Ubuntu 14.04 64bit
Java: As shipped with Wowza
HLS streams
We are running Wowza 4.6.02 with nDVR (3 hours window). The platform includes Origin and Edge applications with 50 streams at about 700 kbits bitrate. When connections on the Edge reach around 300-400, Wowza starts dropping connections and merely becomes unresponding.
This kind of messaged appeared in the logs:
HTTPDvrByteReader.sendRequest[<a href="http://91.109.16.7:12346/dlorigin/_definst_/kinonova.stream.0/ChunkTypes%289%29/ChunkIndexes%28934%29.dvr">http://91.109.16.7:12346/dlorigin/_definst_/kinonova.stream.0/ChunkTypes(9)/ChunkIndexes(934).dvr</a>] : <a href="http://java.net/">java.net</a>.SocketTimeoutException: Read timed out|at java.net.SocketInputStream.socketRead0(Native Method)|at java.net.SocketInputStream.socketRead(SocketInputStream.java:116)|at java.net.SocketInputStream.read(SocketInputStream.java:170)|at java.net.SocketInputStream.read(SocketInputStream.java:141)|at com.wowza.wms.dvr.impl.HTTPDvrByteReader.sendRequest(HTTPDvrByteReader.java:154)|
We were advised by support, that may be the problem is with disk speed. We have tried:
2 HDD’s in RAID0 - no improvement
1 SSD - - no improvement
5 SSD’s in RAID0 - no improvement
So, if anybody has similar problem, please advise us what to do. My opinion is that first Wowza becomes unresponsive, and after this timeout messages appears. When the problem happens, even the wowza manager becomes unresponsive.