Hi Dan,
Let me re-explain the problem you met in case I misunderstood it.
The problem you met is that the server load is heavier (about 2
times) because of the keep-alive job in RTMPConnection. Is that what
you mean?
On 12/9/06, *Daniel Daley* <[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>> wrote:
Hi Steven,
Sorry to just jump in on this thread but I thought it was
probably too similar to start a new one. I've been working all
day on tracking down a performance problem on my server. We have
recently started receiving more traffic to our Red5 box and it
has caused the load to skyrocket. After some investigation I
think it is multiple problems.
As we approach around 70 concurrent VOD (full audio and video)
the load on the server spikes from around 2.00 to 4-5. I
disabled the addScheduledJob and removeScheduledJob inside
RTMPConnection that was just added and the load stays down at
least until the connections build up. I saw a condition like
this once when I was trying to create my own ghost connection
remover. After a while of running when the close() method was
called on certain ghost connections the call would never return.
As the job was a repeating one it would continue hanging each
time it called it until the job scheduler was no longer
executing jobs. I'm not sure that's what would be going on in
this case but they seem very similar.
Even with the ghost remover disabled a load of 2.00 is still
pretty high. I'm having some luck implementing caching of the
keyframe data but the results are marginal. It's times like this
that I wish the memory mapping system in java wasn't messed up.
Thanks,
--Dan--
On Dec 8, 2006, at 7:16 PM, Steven Gong wrote:
Adam,
Thanks for your figures. I'll set these as default value if we
don't find any problems in it.
On 12/8/06, *Adam* < [EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>> wrote:
Hi Julian,
ived played around with the values rtmptConnection,
rtmpMinaConnection at conf/red5-core.xml.
When i took 1000 for both; red5 was all the time running,
but at some point in time the streaming stopped. (no
motioned picture)
The last test i did with rtmpMinaConnection = 10000,
rtmptConnection = 5000 with more than 80 Clients.
It worked fine for more than 12h till my wife closed the
notebook covers today in the morning, arghh!
Ive to rerun this test to be sure, but it looks like that
adjusting these values can fix the problems for this
particular test.
I guess its this (ghost) connection problem. With
rtmptConnection and rtmpMinaConnection you can set the keep
alive time. its pinging the clients. they answer with a
pong including a timestamp. is this timestamp older than in
rtmpMinaConnection, rtmptConnection the connection get closed.
If this keep alive is to big you probably have to much
ghost connection (before they get closed) what is causing
the halt.
ill keep on updating you with the results.
greets
-Adam-
>I have the same problem.
>Steven Gong escribi?:
>>>
>>>
>>> On 12/7/06, *Adam* < [EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>
>>> <mailto: [EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>>>> wrote:
>>>
>>> Hi Chris,
>>
>> i've rerun the test with revision 1583.
>> With less than 10 test clients it worked fine for more
than 12 hours.
>> Than i increased to 20 test clients and after a few
minutes red5
>> freezed as before!!
>>
>> I've checked the connections with netstat -an and while
the server
>> was running or freezed. the connection count was fine!
>> even when red5 was freezed the connections where
established and
>> closed.
>>
>> Also new: The broadcaster receives a unpublish.success .
at least i
>> havent noticed it before.
>>
>> Thats a vital issue and a show stopper. what can i do to
help out
>> finding the problem.
>>
>>
>> Please report it to JIRA first and describe the
reproducing steps as
>> detailed as possible. Thanks.
>>
>> greets
>> -Adam-
>>
>> >Thanks for the quick response.
>> >I will do so and reply the results to this list!!
>> >>
>> >greets
>> >-Adam-
>> >>
>> >>On 12/6/06, Chris <[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>
>> <mailto: [EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>>>> wrote:
>> >>>
>> >>Hi Adam,
>> >>Steven recently fixed a problem with ghost connections
not being
>> >removed automatically. Perhaps this problem that you are
>> experiencing is related. See this thread for more info on
it: ><_
>>
http://osflash.org/pipermail/red5devs_osflash.org/2006-November/002566_
<http://osflash.org/pipermail/red5devs_osflash.org/2006-November/002566_>.
>> >html>>
>> >
http://osflash.org/pipermail/red5devs_osflash.org/2006-November/002566.h
<http://osflash.org/pipermail/red5devs_osflash.org/2006-November/002566.h>
>> >tml
>> >>Could you try your test with r1582 or later? You will
more than
>> likely
>> >have to build it from the source code in the trunk.
>> >>Thanks very much for reporting the bug.
>> >>-Chris
>> >>On 12/6/06, Adam < [EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>
>> <mailto: [EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>>>> wrote:
>> >>>
>> >>>
>> >>> Ups,
>> >>> forgot again!
>> >>> iam using 0.6rc1 Revision 1579 on Windows XP
>> >>>
>> >>>
>> >>>
>> >>> >On 12/6/06, Steven <[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>
>> <mailto: [EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>>>> wrote:
>> >>> >Which version?
>> >>> >On 12/6/06, Adam < [EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>
>> <mailto: [EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>>>> wrote:
>> >>> >>>
>> >>> >>> Hi list,
>> >>> >>>
>> >>> >>> i tried to check my application with respect to it's
>> stability. So
>> >>> >>> i ran a few times the following code(every 10 sec
connecting or
>> >>> >>> disconnecting) on each of 3 machines in my LAN.
>> >>> >>> setInterval(reconnect,10000);
>> >>> >>> function reconnect() {
>> >>> >>> connToggle=!connToggle;
>> >>> >>> if (connToggle) {
>> >>> >>> trace ("disconnect");
>> >>> >>> my_ns.close();
>> >>> >>> my_nc.close();
>> >>> >>> } else {
>> >>> >>> trace ("connect");
my_nc.connect("rtmp://localhost/fitcDemo",
>> >>> >>> roomName,
>> >>> checkId,
>> >>> >>> loginName,
>> >>> >>> broadcaster);
>> >>> >>> remote_so.connect(my_nc);
>> >>> >>> my_ns = new NetStream(my_nc);
>> >>> >>> vid_stream.attachVideo(my_ns);
>> >>> >>> my_ns.play(roomName);
>> >>> >>> }
>> >>> >>> }
>> >>> >>> After a while (sometimes after tens of minutes,
another time
>> a few
>> >>> >>> hours) the server gets halted. Even my timer
that's frequently
>> >>> >>> executed to kill this ghost connection isn't
executed anymore.
>> >>> >>> Its a complete halt. No logs, nothing.
>> >>> >>> The only thing that's working is that the client
receives the
>> >>> >>> netconnection close event on disconnecting. Apart
from this
>> Nothing!!
>> >>> >>> I noticed this with my and the sample applications.
>> >>> >>> After restarting the server it works again. But
after every halt
>>> and
>> >>> >>> restart of the system this halt happens earlier.
At the end of my
>> tests
>> >>> (2
>> >>> >>> days) this halt happened a few minutes after
restarting and
>> >>> >>> connecting to red5.
>> >>> >>>
>> >>> >>> Is this known. Is there a solution, workaround?
>> >>> >>>
>> >>> >>> please reply
>> >>> >>>
>> >>> >>> greets
>> >>> >>> -Adam-
_______________________________________________
Red5 mailing list
[email protected] <mailto:[email protected]>
http://osflash.org/mailman/listinfo/red5_osflash.org
--
I cannot tell why this heart languishes in silence. It is for
small needs it never asks, or knows or remembers. -- Tagore
Best Regards
Steven Gong
_______________________________________________
Red5 mailing list
[email protected] <mailto:[email protected]>
http://osflash.org/mailman/listinfo/red5_osflash.org
<http://osflash.org/mailman/listinfo/red5_osflash.org>
_______________________________________________
Red5 mailing list
[email protected] <mailto:[email protected]>
http://osflash.org/mailman/listinfo/red5_osflash.org
--
I cannot tell why this heart languishes in silence. It is for small
needs it never asks, or knows or remembers. -- Tagore
Best Regards
Steven Gong
_______________________________________________
Red5 mailing list
[email protected] <mailto:[email protected]>
http://osflash.org/mailman/listinfo/red5_osflash.org
