-
Notifications
You must be signed in to change notification settings - Fork 388
Scenario Troubleshooting
ajayhn edited this page Mar 25, 2015
·
7 revisions
1. check on hypervisor 8085/Snh_ItfReq to see if config is missing (i.e. state in ERROR, vn-null etc.).
if missing, config has not come from controller
a. check api-server 8082/virtual-machine-interface/<uuid> if routing_instance_refs are present
if missing,
a. check if contrail-schema is running on atleast one controller
if running
a. check if VMI present in ifmap on all controllers via
ifmap-view localhost 8443 visual visual | grep <vmi-name>
if missing in ifmap,
a. check if rabbitmq is clustered fine and all api servers are connected to it.
b. check contrail-api introspect port (8084) for rest/db/messagebus/ifmap traces
else
a. need to check why contrail-schema didnt establish link between VMI to RI maybe check /var/log/contrail/schema.err
b. check contrail-logs --object-types config --object-values followed by --object-id
else
a. check last log from contrail to stdout + log file
b. check connection status http://<config-node-where-schema-active>:8084/Snh_SandeshUVECacheReq?x=NodeStatus
- Initiate continuous ping from VM to subnet gateway
- Find tap interface of VM (TODO a. from contrail UI b. agent introspect)
- tcpdump -ni
<tap-intf>
- see ICMP requests but no replies received - tcpdump -ni pkt0 -X - don't see the packet going to Vrouter-Agent with metadata
- watch -n1 "dropstats | egrep '[1-9]+'" - observed flow unsable count increases
- find VRF id for net from introspect (TODO)
- rt --dump 34 | less observed NH for VJX0 ip is 0 which is discard.
- Source code check to determine bug in state machine sending from Agent to kernel module.