0

I installed SAP HANA Vora on a 3 node MapR cluster. While trying to bring up Vora service via Vora Manager UI, I get the following error:

Error occurred while starting all services: vora-dlog refused to start. Cannot continue Start All Jobs. Error: There are no health checks registered for service vora-dlog.

The vora-manager log file displays the following error:

vora.vora-dlog: [c.xxxxxxx] : Error while creating dlog store. 
nomad[xxxxx]: client: failed to query for node allocations: no known servers
nomad[xxxxx]: client:rpcproxy: No servers available.

All 3 nodes in the cluster have 2 IPs in different subnets. Can anyone suggest how to configure a health check for consul? And what else can be wrong here?

Single Entity
  • 2,925
  • 3
  • 37
  • 66
AnshulN
  • 1

1 Answers1

1

The messages from the VoraMgr log file are not sufficient to understand the actual problem. Are there other messages from dlog before 'Error while creating dlog store.'? I have seen that message e.g. if the disk was full and the dlog could not create its local persistency.

Also, the 2 different networks could cause an issue like you described. You can configure the use of different network interface names on different nodes. However, on each node all Vora services as well as the Vora Manager must use the same network interface name. If using 2 different subnets the configuration must allow network traffic between them. Could you give some additional info on your topology + network configuration?

Frank Legler
  • 716
  • 1
  • 4
  • 10
  • Few more VoraMgr log messages: agent: failed to sync remote state: No known Consul servers agent: http request failed 'http://192.168.61.192:5646/v1/agent/servers': Get http://192.168.61.192:5646/v1/agent/servers: dial tcp 192.168.61.192:5646: getsockopt: connection refused memberlist: Failed to resolve ERROR: lookup ERROR on 17.0.0.25:53: server misbehaving agent: Join failed: 1 error(s) occurred:#012#012* Failed to resolve ERROR: lookup ERROR on 17.0.0.25:53: server misbehaving, retrying in 5s serf: Failed to re-join any previously known node manager: No servers available – AnshulN Apr 28 '17 at 09:27
  • The MapR cluster is on 3 virtual machine nodes. eth0 has public IPs 192.xx range and eth1 has private IPs 17.xx range. Vora by default takes eth0 as the connecting interface and I changed it to eth1 in the service configuration. Still a lot of services seem to be taking 192.xx IP for connection. – AnshulN Apr 28 '17 at 09:27
  • Is it possible that VoraMgr is not running on eth1? Did you change it to use eth1? (see https://help.sap.com/viewer/p/SAP_VORA -> Install&Admin Guide -> 3.4.3 Configure Network Interfaces for the Vora Manager for MapR. – Frank Legler May 01 '17 at 17:22
  • Is this a secure MapR cluster? We have seen a similar issue on a secure MapR cluster where the user running the Vora Manager (root) did not have a MapR ticket. Giving the ticket to root or export MAPR_TICKETFILE_LOCATION in /opt/mapr/conf/conf.d/vora_default_settings.sh resolved this issue. – Frank Legler May 18 '17 at 16:32