Maarten Litmaath wrote:
> Emanouil Atanassov wrote:
>
>> [...]
>>
>> I think Replica Management worked a day ago, so what has happened now?
>> Some site polluting the BDIIs?
>
>
> In a sense; see below.
>
>> Something with the new glue schema?
>
>
> No. There has been the following bug in GFAL since a long time:
> if a record in the information system contains a "GlueServiceType",
> but _not_ a "GlueServiceAccessPointURL", or vice versa, the routine
> free() is called with an uninitialized variable as argument.
> Such can lead to an infinite loop or a segmentation fault.
> This bug has been fixed for LCG-2_6_0.
>
> What we will do now is search for the "bad" entry and either remove
> it from the BDII or preferrably have it "fixed". More news later.
OK, I found the entries that gave problems and removed their sites
from lcg-bdii.cern.ch:
#TU-Kosice ldap://ce.grid.tuke.sk:2170/Mds-Vo-name=TU-Kosice,o=grid
#AEGIS01-PHY-SCL ldap://ce.phy.bg.ac.yu:2170/mds-vo-name=AEGIS01-PHY-SCL,o=grid
Everybody, please ensure your site is not publishing entries that
only have either "GlueServiceType" or "GlueServiceAccessPointURL".
This probably means hand-editing /opt/lcg/var/gip/lcg-info-static.ldif
on Resource Brokers, GridICE nodes, MyProxy servers, ...
Just add the missing entry with some dummy value.
I have attached a script you can run to test your site BDII as follows:
ldapsearch -x -h your_bdii:2170 -b o=grid | perl bad-dn.pl
The output should be empty.
#!/usr/bin/env perl
use strict;
$/ = "";
while (<>) {
if (/GlueServiceType/) {
print unless /GlueServiceAccessPointURL/;
next;
}
if (/GlueServiceAccessPointURL/) {
print unless /GlueServiceType/;
next;
}
}
|