Copy link to clipboard
Copied
FMS 4.5 on centOS 6.2
Issue - error in stream either encoder issue or packet loss (rtmp sent to server to livepkgr HTTP streaming out, using f4m for flash machines and m3u8 for iOS) causes FMS to ramp up memory until 100%usage causing the server to crash. why is this happening????? need help pronto! Thanks!
Copy link to clipboard
Copied
FMS 4.5 is not officialy supported on CentOS 6.2 - so we've not tested if from our end - but i don't see whether there should issue from FMS end - may be something is changed in 6.2 from 5.5 which is causing this.
Do you see anything in you master/core/edge logs which can be little indicative of what might be happening?
Copy link to clipboard
Copied
We are having a similar problem, although not on CentOS 6.2 We are running on RedHat:
Red Hat Enterprise Linux Server release 5.8 (Tikanga)
Kernel:
Linux stream1 2.6.18-308.4.1.el5 #1 SMP Wed Mar 28 01:54:56 EDT 2012 x86_64 x86_64 x86_64 GNU/Linux
FMS:
FMS_4_5_0_r297
The FMS httpd server is consuming all the memory in the system (here is the output from top):
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
26118 fms 19 0 12.6g 7.3g 1672 S 99.6 93.6 9:46.69 httpd
Our system has 8Gb of physical memory and 10Gb of swap. It consumes all the memory then all the sway until the system is trashed to death.
After some time the system will kill off the process:
May 9 10:03:57 stream1 kernel: Out of memory: Killed process 26118, UID 505, (httpd).
There is nothing obvious in the master.log:
#Fields: date | time | x-pid | x-status | x-ctx | x-comment |
2012-05-09 | 08:45:40 | 31937 | (i)2581173 | FMS detected IPv6 protocol stack! | - |
2012-05-09 | 08:45:40 | 31937 | (i)2581173 | FMS config <NetworkingIPv6 enable=false> | - |
2012-05-09 | 08:45:40 | 31937 | (i)2581173 | FMS running in IPv4 protocol stack mode! | - |
2012-05-09 | 08:45:40 | 31937 | (i)2581173 | Host: stream1 IPv4: 127.0.0.1 | - |
2012-05-09 | 08:45:40 | 31937 | (i)2571011 | Server starting... | - |
2012-05-09 | 08:45:41 | 31937 | (i)2581413 | /opt/adobe/fms/Apache2.2/bin/httpd -f ./conf/httpd.conf -d "/opt/adobe/fms/Apache2.2" -k start returned 0: [Wed May 09 08:45:40 2012] [info] mod_jithttp - FMS installation path: "/opt/adobe/fms/" | |
httpd: Could not reliably determine the server's fully qualified domain name, using 127.0.0.1 for ServerName | - | ||||
2012-05-09 | 08:45:41 | 31937 | (i)2581224 | Edge (31960) started, arguments : -edgeports ":1935,80" -coreports "localhost:19350" -conf "./conf/Server.xml" -adaptor "_defaultRoot_" -name "_defaultRoot__edge1" -edgename "edge1". | - |
2012-05-09 | 08:45:41 | 31937 | (i)2571111 | Server started (./conf/Server.xml). | - |
2012-05-09 | 08:48:08 | 31937 | (i)2581221 | Core (32387) started, arguments : -adaptor "_defaultRoot_" -vhost -app -inst -tag -conf "./conf/Server.xml" -name "_defaultRoot_::::". | - |
2012-05-09 | 09:34:58 | 31937 | (w)2581171 | System memory load (91) is high. | - |
2012-05-09 | 10:03:59 | 31937 | (i)2581172 | System memory load (58) is now below the maximum threshold. | - |
2012-05-09 | 10:05:58 | 31937 | (w)2581171 | System memory load (91) is high. | - |
Or the edge.log:
#Date: 2012-05-09
#Fields: date time x-pid x-status x-ctx x-comment
2012-05-09 08:45:41 31960 (i)2581173 FMS detected IPv6 protocol stack! -
2012-05-09 08:45:41 31960 (i)2581173 FMS config <NetworkingIPv6 enable=false> -
2012-05-09 08:45:41 31960 (i)2581173 FMS running in IPv4 protocol stack mode! -
2012-05-09 08:45:41 31960 (i)2581173 Host: stream1 IPv4: 127.0.0.1 -
2012-05-09 08:45:41 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : localhost:19350/v4 -
2012-05-09 08:45:42 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 80/v4 -
2012-05-09 08:45:42 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 1935/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 127.0.0.1:19350 (rtmfp-core)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 128.252.28.71:19350 (rtmfp-core)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 128.252.28.70:19350 (rtmfp-core)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 10.0.0.1:19350 (rtmfp-core)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 192.168.122.1:19350 (rtmfp-core)/v4 -
2012-05-09 08:45:43 31960 (i)2631509 Public rtmfp-core addresses for listener _defaultRoot__edge1 are: 127.0.0.1:19350;128.252.28.71:19350;128.252.28.70:19350;10.0.0.1:19350;192.168.122.1:19350 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 127.0.0.1:1935 (rtmfp)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 128.252.28.71:1935 (rtmfp)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 128.252.28.70:1935 (rtmfp)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 10.0.0.1:1935 (rtmfp)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 192.168.122.1:1935 (rtmfp)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 127.0.0.1:80 (rtmfp)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 128.252.28.71:80 (rtmfp)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 128.252.28.70:80 (rtmfp)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 10.0.0.1:80 (rtmfp)/v4 -
2012-05-09 08:45:43 31960 (i)2631174 Listener started ( _defaultRoot__edge1 ) : 192.168.122.1:80 (rtmfp)/v4 -
2012-05-09 08:48:08 31960 (i)2581252 Registering core (32387). -
2012-05-09 09:34:58 31960 (w)2581171 System memory load (91) is high. -
2012-05-09 10:04:00 31960 (i)2581172 System memory load (58) is now below the maximum threshold. -
2012-05-09 10:05:58 31960 (w)2581171 System memory load (91) is high. -
Our exact stream setup is like this:
We are using livepkgr. Three streams 200,500,1000Kb streams. We have written a f4m and m3u8 manifest for the 3 streams. We modified the event.xml based on Adobes recommendations. We modified the manifiest.xml making sure it matches the f4m and m3u8 bit rates.
Anything you can help us with would be appricated.
Copy link to clipboard
Copied
Having the same issue CentOS release 5.7 (Final) 12G ram.
httpd eats up all resources and kills the server.
Using livepkr for HDS and HLS.
Any help on this issue would be greatly appreciated.
Copy link to clipboard
Copied
Hello,
it seems to be a common issue.
same problem for me on Linux Machine with RAM 16 Gbytes:
Linux 2.6.18-274.3.1.el5 #1 SMP Tue Sep 6 20:13:52 EDT 2011 x86_64 x86_64 x86_64 GNU/Linux
Into Messages logs i have a lot of: Server[17763]: System memory load (91) is high. and then the system crashes... no more services up.
Please any advice?
Thanks
Copy link to clipboard
Copied
Please use latest FMS 4.5.2 as there has been lots of memory and performance improvements from 4.5.0.297 build. Let me know if you still see the issue.
Copy link to clipboard
Copied
WIll try and let you guys know how it works.
Copy link to clipboard
Copied
Still happening on lastest FMS.
Processes on this host with the greatest increase in usage of Memory during the last 5 minute(s) (from June 15 23:36 UTC to June 15 23:40 UTC):
httpd | 18620 | 10.32 GB |
Server crash.
Copy link to clipboard
Copied
I've added RLimitMEM 1000000000 1000000000 on the apache conf to see if that would prevent the entire machine from crashing.
Will see...
Copy link to clipboard
Copied
That didn't work either...
Never had a problem with FMS using the live application, only when i started using livepkr and HLS this started happening every day.
Unfortunately we need HLS/HTML5 streaming, if this cannot be fixed soon we will have to dump the fms platform in favor of a competitor.
Copy link to clipboard
Copied
Guys,
I am looking to reproduce this at my end. I am working with FMS 4.5.2 on CentOS 6.2.
Few questions:
1. What kind of load does the server have?
2. After how long does this occur?
3. For how long are the streams published?
Thanks,
Shiven
Copy link to clipboard
Copied
I put FMS in it's own CentOS 5.5 Xen VM and it's behaving now.
Can't really experiment with it since it's on a production machine.
Copy link to clipboard
Copied
I`m using Adobe Media Server 5 on a EC2 running CentOs 5.5 and I have the exact same issue.
on my logs, there are several messages of high memory usage, then below threshold, even when I have less than 50 clients.
I use two interactive applications.
My daily load is around 5.000 recorded streams.
Copy link to clipboard
Copied
To anyone struggling with similar issue make sure you have the licence serial loaded AND the server fully restarted. At least on Linux restarting the server through Flash Media Administrator Console leave some zombie processes running which blocks loading the licence correctly. Restart whole OS instead.
ShivenK, you can easily reproduce this by:
1. Install FMS 4.5.7 on Linux. Don't forget to specify Interactive serial licence.
2. Test hls-vod and hls-live. It should work without issue.
3. Restart the server through Flash Media Administrator Console.
4. Test hls-vod and hls-live again. This time VOD will be limited to 30 minutes, and Live will crash the server after 30 minutes by using all RAM. Even though serial is visible in Administrator Console.
Reproduced multiple times here.