Heyup,
> I've just tried reinstalling EMI2 again but adding the pool nodes to
> /etc/hosts and I'm seeing similar (though possibly reduced) issues.
> Assuming that it won't right itself in the next couple of hours, would
> someone with a working EMI2 DPM head node please send me the output of:
How about having the head node in the pool nodes' hosts file? And maybe
have each other in their hosts files too?
>
> rpm -qa | sort
> ps -Af
Here you go (for an Sl6 node though):
http://www.hep.lancs.ac.uk/~msd/lancasteremi2dpmrpms.txt
http://www.hep.lancs.ac.uk/~msd/lancasteremi2dpmps.txt
Hope that helps!
Cheers,
Matt
>
> I can then see if I've got any issues with package versions...
>
> Thanks!
>
> Mark
>
> P.S. Matt's suggestion about exorcising the ghosts of DPM installs past
> is sounding more and more tempting...
>
> On 17/12/12 09:42, Sam Skipsey wrote:
>> What're the failure modes again? Several people found that adding
>> their disk server nodes to /etc/hosts helped to fix strange
>> performance issues with the newer releases of DPM... (we didn't notice
>> it at Glasgow, since we already have all of our servers in hosts).
>>
>> Sam
>>
>>
>> On 15 December 2012 12:42, Mark Slater <[log in to unmask]
>> <mailto:[log in to unmask]>> wrote:
>>
>> Spoke too soon. Though the shift of the MySQL DB did seem to
>> improve it a bit, it started failing in the same way a few hours
>> later. This morning I've tried EMI1 with the same result. The only
>> thing I have left to try is to install Glite from fresh and see if
>> the same thing happens. If it does then I can compare my working
>> head node with the bricked one to see what the differences are....
>>
>> If anyone fancies having a poke around the head node to try to
>> figure out what the problem is, send me your ssh public key!
>>
>> Thanks,
>>
>> Mark
>>
>> On 14/12/12 15:41, Mark Slater wrote:
>>
>> Hi All,
>>
>> Just to let anyone know who is interested, after a few months
>> (and quite a few mails!) I've finally got a working EMI2 head
>> node at Birmingham (yay!). It turns out that the problem was
>> having the MySQL DB on a separate machine. EMI2 DPM did not
>> like this at all. I've now shifted this back to the same
>> machine and it has been happily running for 2.5 hours and
>> counting (~1 hour longer than any previous attempt).
>>
>> I don't know specifically why the head node had problems with
>> a remote DB where the glite version didn't, but neither
>> machine ever got heavily loaded and the number of connections
>> on the DB side was never very big. I did notice that yaim does
>> do some different things depending on if the DB is on the same
>> machine or not, so maybe DPM did the same??
>>
>> Anyway, we now have a working EMI2 head node so I'm happy :)
>>
>> Many thanks to all those who offered help!
>>
>> Mark
>>
>>
>>
>
|