Hi Matt (cc storage),
I'm just having a play aroung with the new dCache version now. I've
upgraded from a fresh YAIM installed 1.6.6-5 dCache to 1.7.0-16 from the
DESY repository. Here is a list of things that that I've noticed so far:
1. Make sure (I mean really make sure) that you're running postgres pnfs
and not gdbm pnfs, otherwise the upgrade will break things and it will be
a bit of a hassle to fix them. I think all GridPP dCache sites are running
postgres pnfs already.
2. You should be aware that tomcat (for the SRM server) is running as
root. I don't know enough about tomcat security at the moment, so can't
tell you if there are any exploits, but it's something to look out for.
3. Related to this, there are java processes listening on ports that
weren't previously being used. I'm looking to find out which ones should
be firewalled off.
4. Still not tested the SRM v2.2 interface. In fact, I'm having trouble
even determining if it's running. In dCacheSetup there is a line:
# srmVersion=version1
(yes it's commented out, presumably meaning it's the default). What about
srmVersion=version2.2? I'll see if I can test it out using Jiri's v2.2
test client.
5. Still trying to figure out what's going on with gPlazma (i.e. dCache
VOMs compatibility). I can copy files into and out of the system, but I
need to try using different VO roles. There are also some options set in
some gPlazma config files that are puzzling me. There _might_ be some
information in the documentation that I need to look at.
6. I've only updated the head node so far, and it's working with 1.6.6-5
door and pool nodes (so far, things might change if gPlazma config
changes).
7. For the pool configuration, there is a new link property called p2pPref
(to go along with writepref, readpref, cachepref). This is set to -1 by
default.
8. Couple of issues with the dCacheSetup file:
clientDataPortRange=50000:52000
even though I set range to 20000-25000 in site-info.def.
performanceMarkerPeriod=180
appears twice.
Assuming you check point 1 above the upgrade should not take you that long
for the head node. I'm just about to upgrade the pool/door nodes, but I
don't forsee any problems there. I'll let you know what else I find over
the next couple of days.
Cheers,
Greig
On Fri, 20 Oct 2006, Matt Doidge wrote:
> Heya,
> How's the testing going for the upgrade? If you think things are good
> to go I'm planning on putting a scheduled downtime for an hour on
> Tuesday morning and upgrading our dcache then. An hour should
> hopefully be more then enough time.
>
> Have a good weekend,
> Matt
>
--
=======================================================================
Dr Greig A Cowan http://www.ph.ed.ac.uk/~gcowan1
School of Physics, University of Edinburgh, James Clerk Maxwell Building
TIER-2 STORAGE SUPPORT PAGES: http://wiki.gridpp.ac.uk/wiki/Grid_Storage
=======================================================================
|