[samba-tng] spoolss conversion and others

[Luke Kenneth Casson Leighton]

On Sat, 29 Jan 2000, Nicolas Williams wrote:

> On Sat Jan 29 2000, Luke Kenneth Casson Leighton wrote:
> > On Fri, 28 Jan 2000, Peter Svensson wrote: 
> > 
> > > By switching to a real database (I assume that is what is being done, I 
> > > was too quick with the "D"elete at the beginning of the thread) queuing up 
> > > the incoming changes should be doable if we don't want to stall them 
> > > until the unthrottle. 
> > 
> > weelll..... that's a bit of a pain. by "locking" you will stop EVERYONE 
> > from being able to log in, access new shares etc. 
> > 
> > it's a bit like shutting off the pdc! 
> 
> Why? Use separate TDBs for SAM, WINs, shares. You've already stated that
> the DCE/RPC daemons are single-threaded and will continue to be so for a
> while. So if each DCE/RPC daemon has its own TDB locking those during
> transaction commits cannot impact any services. Or did you mean that a
> sysadmin locking a TDB could cause services to hang for the duration of
> the sysadmin's transaction. Hmmm.

yes.

> 
> Just say "caveat emptor"! :^)
> 
> It shan't be your problem!

correct!
 
> But if you really want to avoid that make TDB locking not lock out TDB
> queries, use a transaction logfile to store all write operations and
> truly commit them to the TDB at TDB unlock time. Unlocking would briefly
> lock out TDB queries, sure, but the length of that lockout would be very
> short, whereas a sysadmin might inadvertently keep writes locked out for
> a long time.

urr.  that's fairly major, and it's no longer a "trivial data base".

hmm... maybe someone should consider writing a "ttdb" transaction trivial
data base.

any volunteers?