DNS alias in station names does not work

[JonnoGT]

I've had an issue open with CA for weeks and weeks about this problem and I'm getting nowhere fast. Wondered if anyone else had an idea.

Got TNG 2.4 workload manager on NT and a TNG 2.4 Workload agent on solaris 2.8

It all works fine. CCI works, workload works etc.

However, if I change the station record for my workload agent to be the DNS alias rather than the actual nodename workload stops working and all jobs go into a status of WRSRC. And I get the "station unavailable..." message.

We've done the same for our NT agents - using the DNS alias in the station record - and it works fine.
But for some reason it doesn't work for NT to Unix.

My feeling is that is has to be a problem / feature of CCI remote. If it works for NT to NT then this means that CCI Transport is able to resolve the DNS alias OK. So it must be that CCI remote can't do it.

Has anyone else seen this before and even better, did anyone manage to fix it?
Thanks

 

[UberGeek27]

I'm not familiar with the Workload agents in particular, but for CCI & name resolution I've chased my tail a few times...

Before you switched the name in the station record, did you create a CCI connection to the alias name in the EM Connection Manager app (or manually edited the cci files)?

It could be as simple as registering the "new" name with CCI...

 

[JonnoGT]

No, I didn't switch the CCI connection.
It's a long and boring story why....

Basically, "other" people are desperate to use DNS aliases so that disaster recovery is easier. They don't seem to believe me when I say that we only have to switch the station record anyway.
So to be able to do this I need the following:

Active CCI connections to 2 servers, Prod and DR:
So that rmtcntrl says something like:
Sysid State
----------|---------------------|
SERVER-A ACTIVE
SERVER-B ACTIVE

We've got this - all's fine. At the moment we have 2 station records. One for server-a, one for server-b. Both are hunky-dory, no problems.

What "they" want is one station record which specifies the DNS alias in the node name field (so instead of saying SERVER-A it says PROD_APP_SERVER for example.
Then if server-a goes down, the dns alias is switched to server-n and our station record will automatically start to schedule to server-b.

If I put the dns-alias in the cci remote file it works, but then I cannot switch over without manually changing the file and restarting cci on the servers. This defeats the point of having a DNS alias really.

It's a pointless exercise I know, but one that I have to do.

 

[UberGeek27]

It's not gonna happen... CCI maintains a constant connection, and only performs a DNS lookup when the connection is first formed (ie, at agent/manager startup).

The only way to handle failover here is with a cautil script that alter's the station name field in a search/replace manner... unfortunate, but understandable given the nature of the protocol.

If I remember right, there was some decent documentation on CCI in the CA FAQ pages that explained the one-time DNS lookup "feature". There's plenty of documentation on the cautil alter command as well...

 

[JonnoGT]

Yeah, I thought as much. CAn't get CA to admit it though.

The only thing that confuses me slightly is that it works fine for our NT clients.
CCI Transport must be better at handling DNS, and I guess this proves that the problem / feature lies with CCI remote.

Thanks anyway