Mark Goodwin wrote:
> >Nathan Scott wrote:
> > src/pmns/ReplacePmnsSubtree (its after my time too, but IIRC
> > Jonathan wrote it to resolve some PMNS locking issues).
>
> yes he did. IIRC, the pmns manipulation scripts had no mutex.
> I think we tripped up during rc when one PMDA dynamically
> reconfigured it's pmns whilst another was auto-installing
> via .NEEDINSTALL during rc. Something like that.
The PMNS update scripts were mostly OK. I fixed a couple of minor
niggles with signal handling releasing unheld locks and moved some code
back inside a lock-protected region.
If I recall correctly, ReplacePmnsSubtree is for C programs to call,
so they don't have to have signal handling, etc, to clean up locks. If
pmdaavg was open sourced, it contains example usage. The MMV PMDA had
the same problem as pmdaavg, but since it wasn't being used actively,
it went on the low-priority fix list and I didn't get to it.
I think the script itself contains a comment describing the problem.
Maybe at the bottom? Sorry don't have tim to lokk right now.
Regards,
Jonathan
|