hi Tim
please could you post this into the ggus ticket
tia
Mario
On Oct 24, 2010, at 2:16 AM, Tim Dyce wrote:
> Hi Arnau,
>
> We have had some problems in the past at Melbourne, and I know there are issues seen by some sites running the BDII in a VM, but many run it in VMs and see no issues.
>
> We are running LDAP 2.4 now, using a repo by Buchan Milne, which he keeps pretty recent.
>
> [OpenLDAP-Buchan_Milne]
> proxy=http://wwwproxy.unimelb.edu.au:8000
> name=OpenLDAP-$releasever
> baseurl=http://staff.telkomsa.net/packages/rhel5/openldap/x86_64
> gpgcheck=1
> gpgkey=http://staff.telkomsa.net/packages/bgmilne.asc
> enabled=1
>
> I would suspect that running in a VM the problems lie somewhere in the disk or network realm.
>
> So I'd try ldap 2.4, it just seems to run better in my experience.
>
> The BDB database under LDAP can have some weirdness in locking with filesystems, so maybe try the old hack of mounting /var/run/bdii (new BDII package) as a ramdisk and seeing if that helps. By the way, you can also use the slapd_db_stat command to take a look at the state of the BDB database at any time, it might be interesting to take a look when the server is non-responsive.
>
> Last but definitely not least I'd increase the bunch settting in the info provider script
> /opt/glite/libexec/glite-info-provider-ldap
> Line 42:
> my $bunch=10; > my $bunch=30
>
> This means the info provider will perform a lot more simultaneous searches during update process, and mean that slow responding sites are more likely to be included, and potentially the amount of database delete operations. We found this really important withh the high latency at Melbourne.
>
> Cheers
>
> Tim
>
> On 22/10/10 17:31, Arnau Bria wrote:
>> Hi all,
>>
>> I did open a GGUS (https://gus.fzk.de/ws/ticket_info.php?ticket=63084) on the 13th. It's assigned to IS but I have no reply (no input).
>>
>> This problem is happening at pic since 22/10/10 (I did no tgive much importance that day) and still happening (we had 3 "glitches" last night).
>>
>> What are IS team recommendations? How do we have to proceed? downgrade? upgrade testing package?
>> Some input would be nice as this is an important issue.
>>
>> Cheers,
>> Arnau
>>
>
> --
> ----------
> Tim Dyce
> Research Computing
> EPP Group
> The School of Physics
> The University of Melbourne
> +61 3 8344 5462
> +61 431 485 166
>
|