Hi Bill,
> I believe you're recommending that I set up data replication using your mysqld on lsst-dev01 as the master and mine on lsst-dev03 as a slave. I don't believe this is trivial to set up (although, it doesn't look too hard):
> http://dev.mysql.com/doc/refman/5.0/en/replication-howto.html
This wasn't--I was trying to give you the simplest method that could
work, and formal mysql replication is not the simplest. I'm glad you
copied the data instead.
> Is there an obvious reason why I shouldn't simply copy the data over? Because, that's what I did. . .
No, that's exactly what I was hoping.
> I copied all object data over using mysqldump. I also copied the two qms databases. qserv and qms seem to startup okay, but I receive the following error in response to a simple query:
> mysql> select * from Object limit 1;
> ERROR 4120 (Proxy): Error during execution: 'open failed for chunk(s): 1234567890'
My understanding was that those databases were the only meta and data
state that mattered to qserv (aside from the objectId indexes, which are
in qservMeta for now). It seems that your emptychunks is incorrect, or
the partitioning settings are still in disagreement. You should see the
frontend print out its notion of table partitioning when it starts up.
Mine prints:
Metadata Cache in C++:
db: LSST: is partitioned (nStripes=60, nSubStripes=18, defOvF=0.0001,
defOvNN=0.25).
Tables:
Object: is partitioned (overlap=0.025, phiCol=ra_PS,
thetaCol=decl_PS, objIdCol=objectId, phiColNo=2, thetaColNo=4,
objIdColNo=0, logPart=2, physChunking=33).
Source: is partitioned (overlap=0, phiCol=raObject,
thetaCol=declObject, objIdCol=objectId, phiColNo=33, thetaColNo=34,
objIdColNo=3, logPart=1, physChunking=33).
> Any suggestions?
>
> Also, what does your empyChunks.txt look like? I believe I need a copy of this.
Mine is located in lsst-dev01:/u1/lsst/run/master
The frontend should printout what chunks are being rejected as 'empty'
as it prepares the query for execution, and you can verify whether the
correct chunks are being rejected (or whether the master believes the
table is partitioned at all).
I've updated the troubleshooting page to mention the qms, but feel free
to add/correct it:
https://dev.lsstcorp.org/trac/wiki/db/Qserv/Troubleshooting
Hope this helps,
-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
|