Mihail Egorov
2005-Mar-05 16:01 UTC
[Icecast] how can I identify disconnect due to low <queue-size>
1. How can I identify disconnect due to low <queue-size>. Suppose, I have enabled loglevel=4 (debug). Suppose, I have network jam. What shall I see at error.log? 2. Is it possible to mark reason of disconnect at standard access/error log at level 2 (warn)? That is, when disconnect is FORCED by server, this will be marked at level 2 (warn) or may be even 1(error).
Karl Heyes
2005-Mar-05 18:34 UTC
[Icecast] how can I identify disconnect due to low <queue-size>
On Sat, 2005-03-05 at 23:52, Mihail Egorov wrote:> 1. How can I identify disconnect due to low <queue-size>. Suppose, I have > enabled loglevel=4 (debug). Suppose, I have network jam. What shall I see at > error.log?There is a log message that signifies the removal of a listener for being too slow and that is "Client has fallen too far behind, removing" It currently appears at debug level (4)> 2. Is it possible to mark reason of disconnect at standard access/error log > at level 2 (warn)?not in the access.log. If you don't get the above message then it will probably be because the listener closed the connection. The only other ways are an explicit kill, via the admin interface, or a shutdown of the source or server.> That is, when disconnect is FORCED by server, this will be marked at level 2 > (warn) or may be even 1(error).I could see a reason for logging at the info level, but you generally don't want such reporting at warn or error as it is unlikely that you can do much about it, the usual cause being a saturated link. This sort of thing is best left for the stats. karl.
Mihail Egorov
2005-Mar-05 19:00 UTC
[Icecast] how can I identify disconnect due to low <queue-size>
> > 2. Is it possible to mark reason of disconnect at standard access/errorlog> > at level 2 (warn)? > > not in the access.log. If you don't get the above message then it will > probably be because the listener closed the connection. The only other > ways are an explicit kill, via the admin interface, or a shutdown of the > source or server. > > > That is, when disconnect is FORCED by server, this will be marked atlevel 2> > (warn) or may be even 1(error). > > I could see a reason for logging at the info level, but you generally > don't want such reporting at warn or error as it is unlikely that you > can do much about it, the usual cause being a saturated link. This sortI REALLY CAN DO MUCH ABOUT IT! I CAN INCREASE <queue-size>!> of thing is best left for the stats.Queue-size was default 100k. I have a problem of disconnecting clients after 2...20 second. One client drops connection even when he is the single client. Server link was far far far from saturation, and another client has not difficulties. I connected to unhappy client using video conferencing software -- no jams, good quality. Error level was 3, so I can not determine reason. Now I increased queue-size to 1M -- still no problems. I still think to change log level of FORCED disconnection to warn, but even changing to info will be appreciated.
Mihail Egorov
2005-Mar-05 19:09 UTC
[Icecast] how can I identify disconnect due to low <queue-size>
> On Sat, 2005-03-05 at 23:52, Mihail Egorov wrote: > > 1. How can I identify disconnect due to low <queue-size>. Suppose, Ihave> > enabled loglevel=4 (debug). Suppose, I have network jam. What shall Isee at> > error.log? > > There is a log message that signifies the removal of a listener for > being too slow and that is > "Client has fallen too far behind, removing" > > It currently appears at debug level (4)It is not clear, what client has fallen too far behind, when reading error log. Could you add client IP and may be id also?
Apparently Analagous Threads
- Could anybody kindly add statically linked icecast 2.2.0 download to www.icecast.org?
- how can I identify disconnect due to low <queue-size>
- how can I identify disconnect due to low <queue-size>
- how can I identify disconnect due to low <queue-size>
- rect.hclust fails when k is specified (PR#6740)