On 02/11/09 13:04, Michael Schuster wrote: > On 02/11/09 11:52, Sangeeta Misra wrote: >> >> >> Thus far we have agreed that the persistent config files of ILB would >> only be editable by the ilbd daemon and while the user can view the >> file, he/she cannot edit it directly. The next question to address is >> what would be the layout of the config file? Configuration consists of >> load balancer rules >> health check details >> server groups details ( identifying server IP addresses/server ids) >> >> Should this info be all contained in one file, or be put in seperate >> files? One good thing about putting it in one file is that this is all >> that needs to be scped to the secondary loadbalancer when configureing >> HA capability. >> >> Any preferences? > > not preferences, just thoughts: > > for import, a single file is simpler to handle - the routine doing the > actual parsing doesn't care where the strings comes from, and handles > everything (ie, servergroups and rules so far, and I guess also HCs once > Jan is done), and the glue around it is obviously simpler if it has to > handle one file only. > Note that (currently anyway) order does matter, ie creating a rule using > a servergroup that's defined later in the input stream will fail. > > for export, if we use only one file, we must make sure that the order I > allude to above is preserved/created. If we used seperate files for > servergroups, HC and rules, it would be the import-side's task to get > that right .. see above :-) > > I'll follow up if I develop a preference.
this question came up again today in a meeting with Sharon ... we need to decide this issue, either way. If you have an opinion, please speak up ASAP, otherwise I'll make an arbitrary decision (unless Sangeeta does) and implement that. thx Michael -- Michael Schuster http://blogs.sun.com/recursion Recursion, n.: see 'Recursion'
