Hi Fabio, Matthieu and Yvan,
Qserv install procedure can now be adapted to a cluster which is not connected to the Internet. So it can works on IN2P3 cluster.
Could you please help me to define a good install procedure?
Here’s my proposal:
- I have copied a distribution repository on /sps (it’s a bit like an apt or yum repository, but on /sps), this repository contains a given version of Qserv and dependencies packages (size: 324 MB).
- I can provide you a script to install a node. This script would retrieve the packages from the /sps distribution repository,
rebuild Qserv and its dependencies, update Qserv configuration file, run Qserv configuration procedure and then
start Qserv. This script could be used as a puppet post-install script.
One drawback of this simple solution is that /sps would have to support read requests
for all the nodes during install, and also that we would build the code on each node at each install (around 15 min assuming /sps supports load).
An other alternative is to build the Qserv binaries interactively on a machine (using the /sps distribution repository which contains Qserv package).
And then copy the binaries to /sps (1.1 GB),
and then, in the post-install scripts, copy/rsync the binaries on each machines, and then launch Qserv configuration
(here we avoid rebuilding Qserv on each machine but /sps will receive lots of read request at install time, maybe even morethant in the first solution)
Do you prefer one of these proposal, or would you suggest some other improvments?
A good news is that we may have soon some interesting data to put in this Qserv cluster.
Thanks,
Fabrice
########################################################################
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
|