Huge memory consumption with CACHE-EX

[turbopower]

This settings depend from max_cache_time = "maximum time CWs resist in cache after 1st client request, the time must be 2 seconds higher than the parameter clienttimeout."

So 6 sec max life on cache ecm is ok and you must set clienttimeout = 4000 in this case.

[mikie8]

I agree on lowering the timeout value, although, its a "max responsetime" setting, so if his server finds a valid ECM any earlier, it will respond offcourse.

if it reponds , its not a problem . it when it times out , or more to point , its how long it takes to timeout compared to when it receives the next request .

next suggestion :- .

double check ring transmissions . at its simplest its a loop . dont have reader and user settings from the same person in the same group . same goes for cache ex reader / users .

ring transmissions also work after hops , so if 1 connects to 2 and 2 connects to 3 and 3 connects to 1 , 1 has max hops of 3 , that could mean 1 to 2 to 3 back to 1 .

since 7110 , there is code to stop ring transmissions , same name/ip blocks .

i now seem to be collecting memory , which i never did before , maybe as my peers upgrade their oscam , it effects me more.

[CapNCooK]

This settings depend from max_cache_time = "maximum time CWs resist in cache after 1st client request, the time must be 2 seconds higher than the parameter clienttimeout."

So 6 sec max life on cache ecm is ok and you must set clienttimeout = 4000 in this case.

Well, personally i don't think you want cached ECM's any longer than 3 / 4 seconds in your memory.
The validity of the average NDS ECM for instance, will expire in about 5 seconds... Keeping ECM's longer than that will generate freezing, because Oscam can't check the validity for the client.

Also, i expect an answer from my peers within 1500ms max. Waiting any longer is useless in my opinion, it only creates delay for the client to check the next source for the ECM.
If a peer is not able to generate a valid ECM in 1500ms, it should be skipped.

Therefore i still advise to put;
clienttimeout to 1500
and max_cache_time to 3 or 4

[mikie8]

This settings depend from max_cache_time = "maximum time CWs resist in cache after 1st client request, the time must be 2 seconds higher than the parameter clienttimeout."

So 6 sec max life on cache ecm is ok and you must set clienttimeout = 4000 in this case.

the wiki is only a guideline . i use cache time =8 , client time=1750 , these timings are more dependent on the main caid you share and your network ping / response time your server has .

im now 68% mem usage

[mikie8]

Well, personally i don't think you want cached ECM's any longer than 3 / 4 seconds in your memory.
The validity of the average NDS ECM for instance, will expire in about 5 seconds... Keeping ECM's longer than that will generate freezing, because Oscam can't check the validity for the client.

on my free to view white card . i get requests between every 7-9 seconds , thats why i use 8 seconds ,but im willing to test futher on timings once this mem fill goes away .

now 79% mem usage , im going off line and install anolder version .7111 me thinks

[CapNCooK]

double check ring transmissions . at its simplest its a loop . dont have reader and user settings from the same person in the same group . same goes for cache ex reader / users .
..
since 7110 , there is code to stop ring transmissions , same name/ip blocks .

I checked the changes from 7100 to 7125 in the oscam timeline, but nothing is altered on the suggested "ring loops".
Also, by protocol for CC-Cam peers, the configured "nodeid" will prevent loops in requests..

Do you have a link to these modifications ?

now 79% mem usage , im going off line and install anolder version .7111 me thinks

I still believe camd35 is the main cause. Do you use camd35 ?
I disabled camd35 support and recompiled.. Now running steady for 90 minutes @ 200Mb ram..

[turbopower]

on my free to view white card . i get requests between every 7-9 seconds , thats why i use 8 seconds ,but im willing to test futher on timings once this mem fill goes away .

now 79% mem usage , im going off line and install anolder version .7111 me thinks

One of my local has 10 sec between norm requests but with time set to 6 sec there is no freeze with cacheex=3.
When i set time to higher seconds my cpu and ram consumption is increased.

[mikie8]

sorry my mem is bad , it was version 7210 which brought in block ip/name to stop loops .

i use camd35 udp for cache ex 3

i have only 512 in my server and oscam uses but a few percent .

now compiled older version

Unix starttime: 1343910038
Starttime:      02.08.2012 13:20:38
Version:        1.20-unstable_svn  Rev. 7111
Max PID:        unlimited


Active modules:
Web interface support:      yes
SSL support:                no
DVB API support:            no
Anti-cascading support:     no
Irdeto guessing:            no
Debug mode:                 yes
Monitor:                    yes
Loadbalancing support:      yes
LCD support:                no
IPv6 support:               no
Cache exchange support:     yes
camd 3.3x:                  no
camd 3.5 UDP:               yes
camd 3.5 TCP:               no
newcamd:                    yes
CCcam:                      yes
Pandora:                    no
gbox:                       no
radegast:                   no
serial:                     no
constant CW:                no
Cardreader:                 yes
Nagra:                      no
Irdeto:                     no
Conax:                      no
Cryptoworks:                no
Seca:                       no
Viaccess:                   no
NDS Videoguard:             yes
DRE Crypt:                  no
TONGFANG:                   no
Bulcrypt:                   no

[CapNCooK]

sorry my mem is bad , it was version 7210 which brought in block ip/name to stop loops .

OK, yes i have all these options enabled.

i use camd35 udp for cache ex 3
i have only 512 in my server and oscam uses but a few percent .
now compiled older version

If you keep getting high mem, try to rebuild latest version (i run 7228) without camd35 support.
I'll bet your problem is solved..

Also, when having 512MB.. Check the cacheexenablestats parameter.
It should be disabled, as it consumes massive memory, and its for webgui display only.

[mikie8]

One of my local has 10 sec between norm requests but with time set to 6 sec there is no freeze with cacheex=3.
When i set time to higher seconds my cpu and ram consumption is increased.

i slowly increased mine untill i got a freeze using cache3 ecms only ,ill change but it will make no big difference as my server is dedicated to oscam only .

ill use 6 for now ,thanks.

running 7111 , 9% cpu 2.9% mem