[Opensim-dev] Stability & Memory issues

[Melanie]

Hi,

as far as I see this, it is a locking issue.

When cutting hollows from prims, OpenSim pauses for a while, and 
runs heavy database accesses. Database accesses are done under 
locks, and those locks prevent the response to the master UUID query 
to be received in time.
For my sims (4000+prims in 4 regions) the only way I found to get 
them started was to modify the UUID request timeout in the source, 
so that the reply received and queued would still be considered 
valid after all prim cutting has been done.

To really fix this, the quick way is to either wait for the master 
UUID request to be complete before entering the prim hollow cutting 
portion, or to not request that UUID until later, after all hollows 
are cut.

The right way to fix it would be to eliminate the excessive locking 
and break long-held locks occasionally.

Melanie


Justin Clark-Casey wrote:
> Charles Krinke wrote:
>>
>> 1. There are many times when regions must be started more then once. 
>> That is, they throw and exception and refuse to start. This is usually 
>> a master user UUID exception, but sometimes a timeout. It seems that 
>> finding those things which throw fatal exceptions is to our advantage.
>>
>>
> Ooop, any chance of a stack trace?
> 
> I've also heard that some people have resolved this by fiddling with the 
> database (I know no details about this, so this is just to provoke a 
> revelation if there is one out there).
>