#1310 AMF: amfd heart beat timeout is observed for improperly configured nway active appl

[Srikanth R]

Changeset : 6377
Version : 4.6FC

Issue : AMF director heart beat is observed for improperly configured Nway Active application.

Below is the snippet of syslog for which heart beat timeout is observed.

Apr 4 09:39:36 CONTROLLER-2 osafamfd[2789]: ER avd_new_assgn_susi: Either component type or component is missing for CSI 'safCsi=Norm1,safSi=Swan_nwayact_2Norm_3,safApp=NwActApp'
Apr 4 09:39:36 CONTROLLER-2 osafamfd[2789]: ER sg_nwayact_fsm.cc:210: safSi=Swan_nwayact_2Norm_3,safApp=NwActApp (42)
Apr 4 09:39:36 CONTROLLER-2 osafimmnd[2733]: NO Implementer locally disconnected. Marking it as doomed 341 <10, 2020f> (safAmfService)
Apr 4 09:39:36 CONTROLLER-2 osafimmnd[2733]: NO Implementer disconnected 341 <10, 2020f> (safAmfService)
Apr 4 09:39:37 CONTROLLER-2 osafamfnd[2799]: Rebooting OpenSAF NodeId = 131599 EE Name = , Reason: AMF director heart beat timeout, OwnNodeId = 131599, SupervisionTime = 60
Apr 4 09:39:39 CONTROLLER-2 opensaf_reboot: Rebooting local node; timeout=60
Apr 4 09:40:11 CONTROLLER-2 syslog-ng[1173]: syslog-ng starting up; version='2.0.9'

Attached is the script, which shall bring up N way active application, with the following comman
"setNWay_Active.sh Swan 2 create"

[Srikanth R]

Amfd seems to be in while loop while assigning, after admin unlock operation is issued on the first instantiated SU. Attached is the amfd traces on active controller.

[Praveen]

Analysis:
Configuration has some missing association class objects as indicated in syslog warning.
When a SU gets unlocked, AMFD runs red model specific assignment logic for assigning SIs protected by SG of SU to this particular SU. First it creates new SUSI relationship then it tries to create COMPCSI relationship. In this particular case, while creating COMPCSI AMF did not find compType/comp for any CSI of first SI in this SU. Because of this SUSI relationship is deleted and error was logged. In the error part in NwayAct FSM innermost loop, which is for equal ranked SU case, did not break. Since same SU will come as the preferred SU AMF will go into loop. Issue would not have hit if atleast one of the CSIs in SI gets assigned to any component in the SU.

[Nagendra Kumar]

changeset: 6691:23497bc00231
branch: opensaf-4.5.x
parent: 6688:b7d1c951081f
user: Nagendra Kumarnagendra.k@oracle.com
date: Fri Jul 24 15:29:26 2015 +0530
summary: amfd: take next si for assignment in error case [#1310]

changeset: 6692:667850510161
branch: opensaf-4.6.x
parent: 6689:9b9fa4c0fe57
user: Nagendra Kumarnagendra.k@oracle.com
date: Fri Jul 24 15:30:07 2015 +0530
summary: amfd: take next si for assignment in error case [#1310]

changeset: 6693:f900a1faafd7
tag: tip
parent: 6690:d5c0e3088161
user: Nagendra Kumarnagendra.k@oracle.com
date: Fri Jul 24 15:30:17 2015 +0530
summary: amfd: take next si for assignment in error case [#1310]

[staging:23497b]
[staging:667850]
[staging:f900a1]