Bug #2337
closed
NameLsa should be segmented when "Data size exceeds maximum limit"
Added by Junxiao Shi over 9 years ago.
Updated over 7 years ago.
Description
Steps to reproduce:
- write a configuration file where 4200 distinct prefixes are advertised; on average, each prefix has 60 bytes
- start NLSR with this configuration file
Expected: NLSR starts normally
Actual: NLSR terminates with ERROR: Data size exceeds maximum limit written to stderr
Reason:
Currently, NameLSAs are sent in one packet containing all of the names advertised by the router.
Instead, NameLSAs should be segmented when sent.
Files
- Subject changed from Error in configuration file processing: Data size exceeds maximum limit to NameLsa should be segmented when "Data size exceeds maximum limit"
- Description updated (diff)
- Assignee set to Muktadir Chowdhury
- Target version set to v0.2.1
when the packet size will be too large, NameLSAs should be segmented and sent.
Why should segmentation happen only when the packet size is too large?
Having both segmented and unsegmented NameLsa increases complexity.
Why not always do segmentation?
- Description updated (diff)
Junxiao Shi wrote:
when the packet size will be too large, NameLSAs should be segmented and sent.
Why should segmentation happen only when the packet size is too large?
Having both segmented and unsegmented NameLsa increases complexity.
Why not always do segmentation?
Segmentation should always be done; I've updated the task description.
- Status changed from New to In Progress
- Assignee changed from Muktadir Chowdhury to Vince Lehman
- Target version changed from v0.2.1 to v0.3.0
This task's target version has been moved to v0.3.0 in favor of refactoring LSA publishing for each LSA type. Instead of only segmenting Name LSAs, all LSA publishing should follow the same segmentation logic (Task #2965).
- Related to Task #2965: LSAs should be segmented before they are published added
Hi everyone,
I'm currently trying to use NLSR on a "near-tree" topology (ie trees whose roots are connected in a cycle). Attached is an example of a 3-2-4 topology, on which NLSR worked.
In these kinds of topology, each node (so 3 + 3*2 + 3*2*4 = 33 here) announce one prefix and those prefixes are expected to be propagated to every node.
Unfortunately, when I tried using NSLR on a 10-10-10 topology (so 10 + 10*10+ 10*10*10= 1110 nodes) I got "ERROR: Data size exceeds maximum limit" and nlsr failed to correctly establish the FIB/RIB. Attached is the log of lsr on one node (n5), and a yaml description of the whole setup which should be self explanatory.
Is there any chance my error is linked to the one reported here ? Can you think of a way to prevent this error while still deploying huge topologies ?
Thank in advance,
Arthur Carcano
(Intern at LORIA / Inria Nancy)
@Arthur, Thanks for notifying us about the problem. It looks like the same problem as this one, from the error message. We are working on this issue, and will let you know once we resolve it. Right now there are no other work around. We are working on segmenting all the LSA messages, not only Name LSA. Here is the link of it: http://redmine.named-data.net/issues/2965
- Assignee deleted (
Vince Lehman)
- Status changed from In Progress to Closed
Related issue #2965 was finished, which superseded this issue.
Also available in: Atom
PDF
Updated by 
Updated by 
Updated by 
Updated by 
Updated by 
Updated by 