Opened 11 years ago

Closed 11 years ago

#3428 closed defect (fixed)

Cache-EX mode 3 CCcam Cache Reloop

Reported by: sat4ber Owned by:
Priority: critical Component: Cache-EX
Severity: medium Keywords: Cache Reloop
Cc: Sensitive: yes

Description

Revision

Tested with OSCam 8906 and older version of OScam with or without the CW cycle patch.
Protocol: Cache-EX mode 3 with CCcam

Issue Description

Cache Reloop. Sending the same cache back from who you have received.

Server > Friend (receiving)
Friend (sending the same cache back) > Server

Friend has for now no other cache shares * See example *

I have test this with a other guy and he have the same problems.

When the issue occurs

Everytime or you must put every client in a unique group.

How the issue is reproducable

Put one cache share in a not existing group for example group 50 (user and reader). Give the reader one hop more than the user and you will see that everything you receive is sending back. You don't see this when you have more cache shares.

This problem cause a higher upload traffic.

Example from Friend:

Client Server X X REVERSE CACHE PUSH 0 9425 0 0 0

Reader Server X X REVERSE CACHE PUSH 9425 0 0 0 0

Change History (13)

comment:1 by bunbun, 11 years ago

I can confirm; reader is sending back all cache it receives.

Workaround: put all users in a unique group. Give all readers all groups except the one of the corresponding user. Feasible with a small number of readers/users but it becomes complex and unmanageable with larger numbers.

comment:2 by dropsman, 11 years ago

CacheEx-Mode3 is running without any problems here.

Please discuss this first in forum, coz this seems to be a configuration problem.

comment:3 by dropsman, 11 years ago

Resolution: worksforme
Status: newclosed

comment:4 by Dimmie, 11 years ago

Resolution: worksforme
Status: closedreopened

This is NOT a configuration problem. Its an old loop problem that was solved before. Very critical since it also increases cpu usage over 4 times !!

comment:5 by dropsman, 11 years ago

Resolution: worksforme
Status: reopenedclosed

Please discuss this first in the forum with config files.

You can easily create a loop by setting the false groups......

comment:6 by Dimmie, 11 years ago

If you have 450 clients with cacheex you cannot solve it with separate groups. So please do not close this ticket but give a solution. Since this problem was solved before by node-checking it should not be difficult to put back the old node-checking solution. It only works the old way when you have less than 90 clients with cacheex.

comment:7 by Dimmie, 11 years ago

Resolution: worksforme
Status: closedreopened

comment:8 by MrWho, 11 years ago

In networking field we call it "Split Horizon"

:)

comment:9 by blueven, 11 years ago

Until which revision this worked correct?

comment:10 by Dimmie, 11 years ago

Until 8894 it works fine.

comment:11 by blueven, 11 years ago

Are you using cwc on? please try turn it off and report.

comment:12 by CapNCooK, 11 years ago

Remove any cccstealth = 1 options from global cc-cam and from users..

comment:13 by CapNCooK, 11 years ago

Resolution: fixed
Status: reopenedclosed

Should be fixed in 8923

Note: See TracTickets for help on using tickets.