Best way to update mgmt_uri of a shcluster member

[mpettitt1]

Hello all,

I have a search head cluster where everything is working well. We have one member of the cluster where the mgmt_uri needs to be changed. When i change the mgmt_uri to the new address in server.conf and restart that search head, it will come back up but the cluster will show the old uri in use and that member as being down.

I can switch it back to the old uri and it will show as being back up.

What is the best way to change the mgmt_uri of this search head.

thanks!

[richgalloway]

Are you changing the right server.conf file?  Run this command to see which file is setting that value.

splunk btool --debug server list mgmt_uri

[mpettitt1]

Running that returns nothing. Tried playing around with the syntax but couldn't get the btool to return what you may be asking.

[richgalloway]

Are you running the command on the search head in question?

Try this command.

splunk btool --debug server list | grep mgmt_uri

If you still get no results then perhaps the mgmt_uri setting is in a file Splunk can't find and that's why it's not taking effect. 

[mpettitt1]

Thank you that worked.   It returned

/system/local/server.conf                                 mgmt_uri=old ip (current)

When i change the old ip in the server.conf file and restart, it takes on that search head but when i run splunk show shcluster-status it will show the old mgmt_uri and a status of Down.

So i assume the actual cluster still looks for the old mgmt_uri ip until i tell the cluster somehow to look for the new mgmt_uri that i changed on the search head.

thanks

[richgalloway]

Are you using the servers_list setting in servers.conf anywhere in your SHC?  If so, that may clash with mgmt_uri so use one or the other throughout the cluster.

[mpettitt1]

I am not using the server list.  Is there a way to make the shcluster recheck the mgmt_uri of its members, or is this something that should update automatically.

[richgalloway]

AFAIK, it's supposed to update automatically.

[mpettitt1]

Tried it again,  it did not take.  I would hate to have to kick it out of the cluster and rejoin it back in but it looks like that may be what i ultimately have to do.