 |
 |
On 09/04/2013 08:06 PM, Bill Chickering wrote: > Nothing stands out to me in the log files immediately after starting mysql, two xrootd workers, xrootd manager, the corresponding cmsd processes, qms, and qserv. Upon running a simple query (after waiting several minutes for a response) waiting several minutes means that the redirector made an honest attempt to lookup the chunks, but never found them. This is good to know. > That is, all chunkQueries fail. Within the log file for the xrootd manager, I see entries like: > 130904 19:13:50 22220 qsmaster.22164:[log in to unmask] XrootdProtocol: 0600 req=3010 dlen=12 > 130904 19:13:50 22220 qsmaster.22164:[log in to unmask] XrootdProtocol: 0600 open ut /q/LSST/3360 > 130904 19:13:50 22220 qsmaster.22164:[log in to unmask] ofs_open: 2-600 fn=/q/LSST/3360 > 130904 19:13:50 22099 Receive lsst-dev03 48 bytes on 16383 > 130904 19:13:50 22099 Decode lsst-dev03 gave qsmaster.22164:[log in to unmask] err -2 'No servers are available to write the file.' /q/LSST/3360 > 130904 19:13:50 22220 qsmaster.22164:[log in to unmask] XrootdResponse: 0600 sending err 3011: No servers are available to write the file. > > Are you aware of changes in xrootd patch4 that might impact a multi-node setup like mine? Any thoughts? Well, no. You can see in commit 05d6d5 in our qserv branch of xrootd that it contains the same patch to prevent server selection from failing when there are multiple copies. I don't know if that is enough now because patch4 is based on a much newer branch of xrootd. Yours is the only case we have that exercises it. The new path exporting code is another potential problem, because that wasn't implemented in your older code tree (right?). Have a look at the cmsd logs of your workers, and make sure they are loading the libqserv_workerCmsd.so and look for the lookup requests to get a better idea of what's going on. > (Btw, I did have to create two new databases: qservw_worker1 and qservw_worker2, each containing the Dbs table with a single entry 'LSST' so that the xrootd workers would successfully initialize.) Right! I'm glad you figured that one out. Would you like to modify the Trac page that defines the pieces in a "healthy" worker instance? -Daniel ######################################################################## Use REPLY-ALL to reply to list To unsubscribe from the QSERV-L list, click the following link: https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1