Author: Mufeed Usman
Original Publication Date: April 29 2015
Inability to collect support dump data when all the CLDB services are not available.
To be able to collect diagnostic data dump from the cluster nodes even when the CLDB is fully broken.
The file /opt/mapr/server/data/nodelist.txt came into existence with the enhancement bug 4362 filed to address the issue of having to collect support dump (clusterwide) in a scenario where CLDB was completely unavailable. This will be the file the dump collection tool will be made to reach out to be made aware of the MFS nodes in the cluster.
The file gets created the very first time an MFS registers with the CLDB. Subsequently, the following scenarios can arise:
- MFS node gets deleted.
In this case, CLDB will not clean the entry in /opt/mapr/server/data/nodelist.txt. But support dump script can handle this condition gracefully by just reporting the ssh error to old IPs.
- MFS node IP gets changed.
In this case also, CLDB will not modify the entry. Support dump script will not collect information from that missing IP by default.
- MFS node name is changed.
In this case, CLDB will add a duplicate entry with the same IP. Support dump script can handle this gracefully because it has de-duplicate logic built-in.
Bug References (Internal):