Hey guys,

Got a bit of oddball here. I have the need to sync computersystem CI’s to the Atrium CMDB, this works wonderful from within ADDM. However ADDM syncs it with the FQDN name, i only want the hostname in the CI name field.

If I look at the mapping:

If i change this line:

to

Would that solve my issue?

Thanks!

Marco

(added code tags to make more readable – Charles)

Hi Marco,

What you suggest will work, but it is best practice to avoid changing the default mapping definitions if you possibly can, so you have an easier time during upgrades. You can change the BMC_ComputerSystem Name attribute in an extension mapping, like this:

Hope that helps.

Duncan.

That would help a lot ;-)))

Thanks!!

Marco.

MMmmzzz.. it does work for some hosts…. but not all of them.

Look at this one taken from the file: tw_svc_cmdbsync_transformer

How could i solve this with ADDM? I could create a remedy filter for this, but i rather not…

Kind regards,

Marco.

What about it doesn’t work? The log you’ve posted doesn’t say anything about it either way.

Is is that the HostName attribute is a fully qualified name? Sometimes hosts really do have a fully qualified name as their hostname. If you want to strip that off, you can use a regular expression. Something like this

Thanks, that was it…

Our unix administrator did have some comments on the standard AIX platform script (since it is a AIX environment).

He states if you want the hostname command to fetch the hostname, one should always use the -s option to make sure you dont get a FQDN.

Maybe that would be a good idea to take into account with one of the next TKU’s? Updates

Thank again, your solution worked for me.

Marco

Thanks for the feedback Marco I’ll make sure it’s passed on. Platform scripts can’t currently be updated during a TKU, they are a core release only today.

Can I ask a favour, could you ask your AIX guru if that option is available across all version of AIX? I’d hate to end up getting an error and no hostname at all.

Hey Charles,

Here is the feedback from our AIX guru:

Hi,

I have checked the following AIX versions (which span the last 10-15 years) :-
6.1
5.3
4.3

and all support the hostname -s command, so I expect that it will work everywhere, I don’t have access to AIX7, but don’t expect it will have changed

Thanks!

Marco

Unfortunately, some of our customers have machines where the unqualified part of the name overlaps with other machines, and the full name from hostname is required. There’s no easy way out that copes with those situations, and gives a simple name where it would be preferred. That’s why our discovery scripts do not use hostname -s or equivalent.

What we have contemplated doing is stripping the trailing parts of the hostname if it exactly matches the domain name. Would that work for you?

Hi Duncan,

Yes, that would work us as well.

Thanks for giving it some thought.

Marco.

Duncan,

The Host_Name_Override example was very helpful to me to. Just for me to be clear in understanding,

Where, you use

computersystem.Name := computersystem.HostName;

I am correct in thinking that a similar effect could be acheived by writing this (with a possible perfomance hit)?

computersystem.Name := host.hostname;

The difference appears to be that it would be reading the value from the original Host node, rather than reading the mapped value previously set in the Host_ComputerSystem module. So this way I could just use a completely different variable like host.acme_custom_host_identity, and that would override the mapping created by the original module.

Rob I suspect this is related to your thread here, I’m working on example to help that thread if you can give me a bit of time. I’d rather check that it worked before I post it and let you find out ;)

ADDM 8.1 platform scripts ensured that the hostname is short (on most platforms by cutting away everything after the first dot, ie: ihn=`hostname 2>/dev/null | cut -f1 -d.`)

ADDM 8.2 platform scripts just take what is configured to be the hostname. On Debian, it is clearly stated that this must be the short hosname. On Redhat, it may be the short hostname or the local fqdn. Other OSs have other guidelines.

As a result, ADDM 8.2 displays (and even worse: syncs to cmdb) inconsistant host names.

We fixed this by stripping the domain part in the platform script, not the sync-mapping. This ensures that not only in Atrium CMDB, but also in ADDM, the names are consistantly displayed as short names. If they should happen to not be unique, you can just combine the hostname with dns_domain or domain or take local_fqdn.

@Duncan: why are overlapping hostnames a problem for some of your customers? I think the new platform scripts are not fixing whatever problem they might have because the scripts do not ensure the hostname to be either long or short. However, they definitely broke our recon rules (we use {hostname, domain} as key for ComputerSystems from various sources).

They are only inconsistent in how you want to use ADDM and CMDB. If a client was trying to use the same to pick up RedHat machines that had been configured with the hostname as local fqdn against their policies then your approach with platform scripts (and the one we did briefly in one of the 8.1 release) prevents them from doing this and hides a misconfiguration. That’s an undesirable thing in a discovery tool.

Many clients also expect the name in the CMDB to be wholly unique and don’t have your approach.

Rather than edit the platform scripts more stable customisation for you would be to update the syncmappings via an override to change the name we sync to be the “name – domain name” as Duncan suggests. That will mean you don’t have to customise the platform scripts on every upgrade or risk missing an important change to them.

.

You should be able to do just as the example but use text.upper() just as you would in any other pattern.