CERN Accelerating science

Errors in logs : "Error: key=manager uses an empty value!" in MGM and "error: unable to resolve (null)" in MQ

Hi,

eos-server 4.8.10
I have some errors on MGM logs, near the MGM start :

...
###### mq messaging: starting thread 
200909 11:53:58 time=1599645238.103112 func=CreateObject             level=INFO  logid=6142b11c-f282-11ea-94bf-7845c4fc35a5 unit=mgm@lyoeosmgm1.in2p3.fr:1094 tid=00007f9efb19c780 source=PluginManager:
287              tident=<service> sec=      uid=0 gid=0 name= geo="" created plugin object type=NamespaceGroup
200909 11:53:58 time=1599645238.992691 func=Supervisor               level=INFO  logid=6146a34e-f282-11ea-94bf-7845c4fc35a5 unit=mgm@lyoeosmgm1.in2p3.fr:1094 tid=00007f9eb9ffd700 source=QdbMaster:245 
                 tident=<service> sec=      uid=0 gid=0 name= geo="" msg="waiting for namespace boot" mNamespaceState=booting
Error: key=manager uses an empty value!
Error: key=manager uses an empty value!
Error: key=manager uses an empty value!
Error: key=manager uses an empty value!
Error: key=manager uses an empty value!
...

And on the /var/log/eos/mq/xrdlog.mq logs (also on the MGM server) :

...
error: unable to resolve (null)
error: unable to resolve (null)
error: unable to resolve (null)
...

On the MGM /etc/xrd.cf.mq :

...
mgmofs.qdbcluster lyoeosqdb1.in2p3.fr:7777 lyoeosqdb2.in2p3.fr:7777 lyoeosqdb3.in2p3.fr:7777
...

Are these harmfull ? Does they point a misconfiguration ?
Cheers,
Denis

Hi,

At least for the xrd.cf.mq file I believe the line should be mqofs.qdbcluster <node1> <node2> <node3>.

Hi Crystal,

I had in my xrd.cf.mq (which seems to be wrong) :

mgmofs.qdbcluster lyoeosqdb1.in2p3.fr:7777 lyoeosqdb2.in2p3.fr:7777 lyoeosqdb3.in2p3.fr:7777
mgmofs.qdbpassword_file /etc/eos.keytab

I tried :

mqofs.qdbcluster lyoeosqdb1.in2p3.fr:7777 lyoeosqdb2.in2p3.fr:7777 lyoeosqdb3.in2p3.fr:7777
mqofs.qdbpassword_file /etc/eos.keytab

But the logs error: unable to resolve (null) are still there.

Then I tried mq.qdbcluster and mq.qdbpassword_file as Elvin wrote in Config in quarkDB for master/slave(s) then the problem goes away.
NB : I don’t know what is best, mq.qdbcluster or mqofs.qdbcluster as writen in http://eos-docs.web.cern.ch/eos-docs/quickstart/ns_quarkdb.html

Thanks a lot Crystal !

Now for the message Error: key=manager uses an empty value! in the MGM logs, I suspect a NS problem.

Denis

Hi Denis,

I fixed the documentation which was using the wrong configuration directive. Thanks for pointing it out!
For reference, the correct one is mq.qdbcluster/mq.qdbpassword_file.

Cheers,
Elvin