[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference azur::mcc

Title:	DECmcc user notes file. Does not replace IPMT.
Notice:	Use IPMT for problems. Newsletter location in note 6187
Moderator:	TAEC::BEROUD

Created:	Mon Aug 21 1989
Last Modified:	Wed Jun 04 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	6497
Total number of notes:	27359

6443.0. "DECmcc and DNS timing?" by GADWAL::W_MCGAW () Mon Apr 22 1996 22:08

    Hi,
    
    Is anyone aware of timeouts for DECmcc that affect processing of
    entities within DECdns?  For example:
    
    I have a customer that has a command procedure that deletes a domain,
    recreates the domain and then populates the domain with entities that
    are already registered in DNS.  If he executes this procedure on his
    development lab system it works fine.  the development system is
    running VMS V5.5-2, DECnet/OSI V5.6Beco12 and DECdns server is running
    local as the only server in the namespace.
    
    If he tries this same procedure on his test lab system, the results are
    not good.  The test lab system is the same except for DNS server is
    NOT running on the DECmcc node.  DNS server is running on two other
    nodes (one mastering the MCC directories and the other with read-only
    copies.  
    
    The delete domain returns successfully but when he tries to create the
    domain again, it comes back with a entity already exists.  If he waits
    for a longer period of time, he can then successfully create the
    domain.  When he tries to populate the entites back, some work fine and
    others return a message saying that it could not find the entity.  The
    populate command is using the full entity name for DNS.  After the
    procedure has executed, he can go back and manually populate the
    entities that could not be found and they work fine.
    
    Is there a way to increase a timeout value that DECmcc may be using to
    allow it to query DNS longer and hopefully be able to find the entity
    to populate the domain?  Is this problem a result of asking DNS to do
    too many things in too short of a time period?
    
    Thanks,
    Walt McGaw
    Digital Network Services Unit
    CSC/CS

T.R	Title	User	Date	Lines
6443.1	it is a timing issue	SPANKY::WALSH	`Mon Apr 22 1996 22:24`	13
	It is definitely an issue with DECdns updating its databases via skulking. A potential way around this could be to place wait statements in the command procedure. Another potential way around this would be to incorporate a DECdns skulk update before doing the creates. I've been down that road before and it can get quite ugly fast if you try to write to much stuff to the namespace in a short period of time. Hope this helps, Mike Walsh Network Management Services DTN 227-3894
6443.2	Try to increase confidence !	DSVB03::TABONE	`Tue Apr 23 1996 10:58`	14
	Hi Walt, What you describe is a normal behaviour (I think). To avoid traffic, replicas are updated each period of time following a parameter named dns$confidence. The values of this parameter could be : low, medium (default) or high. I think that if you set the parameter to high, mcc will read/write directly the master. But performance could be affected. But a DNS guru could give you more precise informations. Hope this helps you. Best regards, Philippe Tabone - DECmcc Maintenance Team -
6443.3	Thank you and a couple more questions.	GADWAL::W_MCGAW	`Tue Apr 23 1996 14:14`	21
	Thank you very much for the replies! This confirms what I thought was happening to this customer. RE: .2 The dns$convergence directory characteristic only varies the skulk time if you set it to high. In the default setting of medium, a directory update tries to complete immediately but if not successful, it will wait until the next scheduled skulk time to try to update the directory again. If it is set to high, the skulk is scheduled for no more than one hour into the future. Is there a setting in DECmcc that allows you to force DECmcc to read/write the master directory only? What about logiclas etc. that allow you to alter the amount of time that DECmcc will wait to hear from DNS? Thanks again! Walt
6443.4	More explanation	DSVB03::TABONE	`Wed Apr 24 1996 16:23`	32
	Hi Walt, There are 2 different logicals : dns$convergence is a DNS logical which defines the time to skulk the replicas : Low is every 24h, medium (default) every 12h and retry 12h after, and high every 12h and retry every 1h if it fails (it is what you explained in .3). mcc_conf_dns is a mcc logical which can take the same values : - low : local cache - medium : any server (replica or master) - high : master server For read action : it is following the definition of mcc_dns_conf. For write action : it is high level. So, when your customer deletes the domain, the level is high and it is deleted in the master...but the replica will be updated following the dns$convergence level. When he tries to create it again, there is before a read action to verify if the domain does not already exist. If the skulk time is not passed, it sees the domain and answers "domain already exists..". If the skulk time is passed, the replica is updated and no problem. So, the solution is to set mcc_dns_conf to high. Hope this helps you and near comprehensible :-) Best regards, Philippe Tabone - DECmcc Maintenance Team -
6443.5	We will try the mcc logical.	GADWAL::W_MCGAW	`Wed Apr 24 1996 22:11`	7
	Thank you Philippe! We will try that. Best regards, Walt
6443.6	Custoemr is working with waits.	GADWAL::W_MCGAW	`Fri Apr 26 1996 22:11`	9
	The customer is working with waits in between the populate commands and I also gave him the MCC_DNS_CONF logical. He will define the logical just for the delete and create domain commands and then remove it since normally you will want to read from which ever DNS server can respond (or respond quickest). Thanks! Walt