NBU has to use "hosts" file?

[manatee]

NBU 7.6.0.3

so i was configuring all my Solaris machines to use my Active Directory DNS server because of ntp client configuration requirements. ntp work done and working fine. name lookups via Active Directory DNS also working fine without me having to always manually update each server's "/etc/hosts" file.

however, today i found that NBU master can't seem to resolve names when done this way. i have to use the old way by entering the client's hostname and ip address in the "/etc/hosts" file. then and only then did the master server able to resolve and find the client.

why is this so? how can we move NBU to the 21st century and make use of a DNS server?

[Marianne]

NBU works 100% with DNS. Just ensure that DNS reverse lookup is also enabled. If you can use nslookup to resolve hostname to IP and IP to hostname, then NBU will work with DNS. Just remember to clear host cache after OS-level changes are made.

[manatee]

yes i can resolve both ways. still if i remove the entry in "/etc/hosts" the master/media server gives "(48) client hostname could not be found".

[root@manager bin]# ./bpclntcmd -clear_host_cache
[root@manager bin]# ./bpclntcmd -clear_host_cache
[root@manager bin]# ./bpclntcmd -clear_host_cache
[root@manager bin]# ./bpclntcmd -clear_host_cache
[root@manager bin]# ./bpclntcmd -clear_host_cache
[root@manager bin]# nslookup 192.168.1.10
Server:         192.168.1.50
Address:       192.168.1.50#53

10.1.168.192.in-addr.arpa        name = clientA.acme.com.

[root@manager bin]# nslookup clientA
Server:         192.168.1.50
Address:       192.168.1.50#53

Name:   clientA.acme.com
Address: 192.168.1.10

[manatee]

ok i found this command parameter is failing:

[root@manager bin]# ./bpclntcmd -hn clientA
client hostname could not be found
alter client: clientA : not found. (48)

but if i do the resolving thru the OS it resolves fine as noted in my previous post.

so how does one troubleshoot this "bpclntcmd" command?

[manatee]

found the problem! at least from the master/media server side:

#hosts:     db files nisplus nis dns
hosts:      files dns
ipnodes:     files dns

in "/etc/nsswitch.conf" there was a missing colon after the "ipnodes" keyword. after i inserted the colon, "./bpclntcmd -hn clientA" works as expected.

still from NBU console it gives the same error 48. doing "bpclntcmd -pn" from the client side produces no output.

[sdo]

As Marianne suggested... ...try this on the master and all media servers: bpclntcmd -clear_host_cache

[manatee]

did that many times over. i only have one master server which is also a media server.

it's resolving fine now from master/media server but from the NBU../Host Properties/Clients it still gives the same error message.

without restarting the services, anything else i should check?

[sdo]

Are all the names the same case?

[manatee]

yes. both master/media server and client are unix machines so i use consistent case.

[sdo]

Pure guess... ...try on master: bprdreq -rereadconfig

[Marianne]

It is difficult to try and assist you if you cannot show us actual names (including Host Properties screenshot).

Another look at your nslookup commands seems that reverse lookup only resolves FQDN -  clientA.acme.com but you have only clientA in NBU policy.

Please check DNS to ensure that shortname and FQDN is resolved in nslookup commands.

If you change NBU policy now to clientA.acme.com, you are going to experience issues with restores, as the 2 hostnames will be stored in different images folders on the master.

[Genericus]

100% confident that this is network/OS DNS issue. Agree with Marianne - you need to confirm that you can communicate at the OS level forward and backward, with the same host name, using DNS. 

[manatee]

i have already confirmed that everything resolves fine from the OS:

1. the ntp client is working fine; it will not work if it can't resolve properly any hostname (be it FQDN or not)
2. my post above shows "nslookup" from the OS is able to resolve by name and by IP
3. i did change the NBU policy from using just the hostname to using FQDN and NBU still can't resolve
4. i have also cleared the host cache severals times

[Marianne]

Your example shows that DNS resolves by FQDN: clientA.acme.com

You have configured NBU for shortname: clientA

clientA.acme.com != clientA

[Tape_Archived]

Can you please check if you have any specific setting for these clients??

Go to : Hosts Properties => Clients => clientA.acme.com

[manatee]

but that's how name resolving works from the OS. if the OS can resolve why not the application? unless the application has its own concept of networking then i should not be concentrating at the OS level.

i'm just confuse why NBU uses "/etc/hosts" only and not a DNS server.

in Hosts Properties => Clients => clientA.acme.com the client is showing. as for the specific settings you are referring to i'm not sure what you want. i created a test policy, with clientA (or clientA.acme.com) as the client and that's it.

[Marianne]

The moment you can show us screenshots of real life config and tests, we will be able to see what is going wrong in your environment. NBU works fine with DNS. We have many customers using DNS only. Good luck.

[Genericus]

One thing you can also check is the client properties: There are THREE places that MUST match (or at least NOT be different)

From java admin console, expand NetBackup Management, then Host Properties, then select Clients, in the right window, select your client - it should polulate as it connects.

Once it is connected, right click and select Properties, or menu Actions, Properties.

A window should pop up - check these tabs:

Main Properties in left window - what is the Host name in the right side?

Universal Setting Tab - is there anything in the Use specified network interface: field? IT CAN BE BLANK, as long as the client name matches the NIC you use.

Client Name - what is the Client Name field?

Some of my server admins cannot seem to match these, short name vs long name, case mis matches, client vs client-backup - all these cause issues.

You generally will not go wrong making them all the same - as long as they match the client in the policy...

[manatee]

as much as i would like to share the actual screenshots, due to security reasons i have to mask the actual hostnames and ip addresses. i hope you understand. however, all that i posts are exactly what i see here except the change in hostnames and ip addresses.

i checked the client properties as suggested by Genericus and everything checks out: long name everywhere:

1. in client's "bp.conf" it uses FQDN
2. in client's "/etc/hosts" it uses FQDN with alias
3. in my DNS server it uses FQDN with alias
4. in NetBackup Management, then Host Properties, then select Clients it uses FQDN in client name

again, master have connection to client only IF i use "/etc/hosts" and not when i use my DNS server.

[Genericus · Accepted solution]

OK, so now check the system files and make sure it is set to use DNS before local files.

Personally - I have had this same problem before, and since it DOES connect using /etc/hosts, it should not be a networking issue. I just added that one host to /etc/hosts and did not waste any more time on it.

Just a thought - 

Mostly this happens with multi-NIC systems, where traffic comes in one NIC and for some reason, returns via another path. Master definitely sees a difference sending to client-backup and getting a response via client instead. I have had a problem before on a linux box, I had to do this:

Here is the official link describing how to change RHEL6 and RHEL7 to support loose asymmetrical routing:

https://access.redhat.com/solutions/53031