[Freeipa-users] Confused/lost at promoting a replica into a master

[David Copperfield]

Hi Deon and all,

>> Hi follks,
>>
>>  I'm completely lost at reading the IPA document on how to promote a IPA replica into master IPA. When I'm try to follow the steps listed in the chapter '16.8.1 Promoting a Replica with a Dogtag Certificate System CA' at the link http://docs.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/6/html/Identity_Management_Guide/promoting-replica.html#promoting-pki, the last steps 'g' said:
>>
>>    g. Disable the redirect settings for CRL generation requests:
>>         master.ca.agent.host=hostname
>>         master.ca.agent.port=port number
>>
>> The above instructions don't give any hints of 'hostname', or 'port number'. users don't have any clues about them, should them be this replica's name, or the original master's name? and what is the por
>> t number? it is a TCP port, or a UDP port?
>
>The replica is configured to check for information from the master CA -- in this case, asking the master CA to generate a CRL. Those parameters tell the replica where to look. Part of promoting the replica is telling it *not* to look for a master CA. So, those parameters should be blanked or removed.
>
>I can definitely make that more clear.

Sure, please elabroate -- I'll still half undertstand only :) This part is pretty confusing by itself.

First, when a IPA replica is first installed, the dogtag certification system is not installed at all, so the directory /var/lib/pki-ca/conf doesn't exist on IPA slave at all. The directory shows after 
only after 'ipa-ca-install' command is run on the replica.

After running the command 'ipa-ca-install', in the configuration file '/var/lib/pki-ca/conf/CS.conf', there are no 'ca.crl.*' statements on IPA replica at all; there are no master.ca.agent.{host/port} s
tatement either. 

What we really need to clarify here, from users' respective, are elaborated below(may not be completed):

1, how to promote a IPA replica into a IPA master?

2, What's the effect on other sibling IPA replicas? -- do we need to break original replication agreement with old IPA master? and create new aggrement with new server? If so, how to do it?

3, How to check/verify that new IPA replica is really promoted into new IPA master?

4, how to check/verify that old IPA Master is stopped its orignal master function? disowning the master CA in the PKI hierarchy as claimed? 

4, what's the operations on the original IPA master?
  4.1 case #1, what is the 'official' steps to remove/decommission original IPA master? -- what's the steps besides final 'ipa-master-install --uninstall'?
  4.2 case #2, if the original IPA server is broken completely and all IPA replica could not reach it? -- Then what's are the steps to promote a IPA replica? Do we need the orignal /root/cacert.p12?
  4.3 case #2, if the original IPA server is only temporarily unreachable? -- then after an IPA replica is promoted into new IPA master, how to depromote the orignal IPA master to replica after it is up?

>>
>> As a serious evaluator of IPA, I have to think more above just for fun. So it is a natural thought to think about disaster recovery and smooth/continuous operations(simulation and real case): how to back up data,
>
>Currently, there is no disaster recovery or backup information. There are a couple of RFEs open to develop this information. My understanding (and this is something that Dmitri or one of the engineers can explain better) is that the best thing to do is to back up the DS instances using db2ldif and then spin up a new server/replica instance and import the backed up data using ldif2db.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/freeipa-users/attachments/20120430/535bfd55/attachment.htm>