 |
 |
On Fri, 13 Oct 2006 11:26:18 +0100 Kostas Georgiou <[log in to unmask]> wrote: > On Fri, Oct 13, 2006 at 11:58:04AM +0200, [log in to unmask] wrote: > > > On Fri, 13 Oct 2006 10:49:09 +0100 > > Kostas Georgiou <[log in to unmask]> wrote: > > > > > On Fri, Oct 13, 2006 at 10:39:29AM +0100, Olivier van der Aa > > > wrote: > > > > > > > I restared dcache-core on the head node and srm works again. > > > > Don't know why. I have attached a fraction of the todays logs > > > > when it was not working. > > > > > > This is scary, last night I tried all possible ways of restarting > > > dcache with no effect (restart just the head node, take down > > > everything and start again, clean up the srm databases and start > > > again, restart each of the cells in the head node individually > > > etc..). > > > > > > > > > If the thing fails to start correctly with no reason we really > > > have a problem, next time it dies it might not come back at all. > > > We really need to find out what caused the problem and what a real > > > fix requires. > > > > > > Kostas > > > > Well the good news is that the SRM code is completely rewrittern and > > ment to be coming out today, although the upgrade scripts are still > > being worked upon. > > If it's completely rewritten it means that we start again from zero > and while this bug might not exist there, new ones will surely show up > :( Was the code/design that bad that it needed a complete rewrite? > This is also scary. > > Yes I am good at finding things to complain about ;P So am I ;P I don't know if it was that bad, I am making time by automating things, but have not yet scripted away as much of my job as I have done in the past and maybe in the next few months I shall get this done but I certainly don't have time to look into Timur's work at the moment. The reason is SRM v2 which to me obviously requires a complete rewrite from SRM v1. I wrote a Castor SRM v1 and it would not be a good place to start to write Castor SRM v2 because the spec got overloaded with functions and features that are so complicated you may have to have been at the design meetings to understand the API. Regards Owen