Ken Hornstein wrote: >> At 02:08 PM 9/19/2006, Ken Hornstein wrote: >>> Stupid question #2: is it just a matter of tidying things up that you want >>> those prefs removed? >> In our case yes, and to prevent time out failover lag. > > Okay ... but I thought that you said that all of the volumes had been > removed from these servers. If that's the case, the client shouldn't > try talking to these nonexistant fileservers at all (the vlserver would > tell the client that they're not there anymore). If they're still > listed as replicas, why haven't you vos remsite'd those replica > locations? > > The reason I'm asking this is that I don't think removing the server prefs > will help with failover lag. Stopping the client from pinging the > non-existant server, yes, but that doesn't sound like that's your issue.
Now I'm getting confused. Rodney, what exactly are you talking about. Are you asking about "file servers or vldb servers"? If you are changing the list of vldb servers you had better be replacing your CellServDB files. In that case you will want to issue "fs newcell" after the new CellServDB file is in place in order to flush the server list. If you are concerned about file servers, then when you remove the volumes from the file servers before shutting them down, the vldb servers will stop telling the clients to access them and there will be no issues with failover since the clients will never contact them. If you are manually "fs setserverprefs" then you are creating a manual entry for the server and that will not be flushed by a "fs newcell" since you are expected to know what you are doing. You must restart the client to remove the "fs serverprefs". Server preferences that are automatically configured by the client will be flushed by "fs newcell". I'm not sure whether the failure of "fs newcell" to clear the manually configured server preferences is a bug or not. Jeffrey Altman
smime.p7s
Description: S/MIME Cryptographic Signature
