AMF integration with IMM

This page contains development information related to AMF Integration with IMM; which is OpenSAF Release 4 work item covered by these tickets: #161, #617, #612, #700 & #701.

Developer Repositories

• Hans Feldt: -
• Nagendra Kumar: http://devel.opensaf.org/hg/opensaf-avsv-oi-nagu/

Work Items

Item	Owner	Beta	Status
Convert all logging/tracing in AMF to use Logtrace API	All	5	Ongoing
Relax order of object creation in CCB		5
Support for SG admin operations		5
Support for CSI dependencies	Nags	5	Done
Align impl. of SI dependency class (impl)	Nags	5
Align impl. of SI Ranked SU		5	Floated for Review
Open Issue: Any impact on AMF model because of modularity	Mario	5
Enable "External proxied component" support	Nags	5
Modify "External proxied component" configuration reading to local IMM-db rather than AVNd-AVNd checkpointing	Nags	5
Support change of types		5
Support for Application admin operations		5
Dynamic changes to AMF Model (for MW Entities)		5
Correct deviation of type for saClmNodeID (config->runtime)		5
Support for dynamic deletion of objects		5
Tool/script to support generation of initial config (imm.xml) for cluster sizes for OpenSAF AMF Entities (1-X nodes) (#700)
Tool/script to support dynamic changes to OpenASF cluster size (for MW enities) (#701)
Splitting imm.xml in per-service imm.xml		5
Divide scap into two processes: opensaf_avd & opensaf_avnd: ticket #755.	Hans	5	Started

Additional Ideas (Hans)

• AvD-AvND protocol could be more logical, what is SET-LED? Why not have an admin-state-set message and let avnd handle lock-instantiation etc
• Old bugs needs a permanent solution:
  • Ticket: #210: I suggest that we add a node_type attribute to the CLM-NODE-UP message, then AVD can determine if it should start heart beats or not. How does the AMF-CLM split affect this problem, will we still have heart beats?
  • Ticket #689: Probably timer memory has been freed.
  • Ticket: #579
• I would like traces always written to an internal circular buffer. That makes the last execution visible in a core dump. Kind of a black box recorder. Should not be hard to extend logtrace to handle that. Will need enter/leave tracing in most functions.
• We need an AMF monitor that can handle that pcap dies. It need to reboot the node. This is needed in a PC cluster (non HPI managed). ticket #763