On 6/18/13 5:31 PM, Sage Weil wrote: >> 1) Remove the %ghost directive and allow RPM to install the directory. >> Potentially leaving orphaned pid/state files after the package is removed. >> >> 2) Or the directory needs to be created in the %post section. If it is >> created in the %post section and the %ghost directive stays then when >> the RPM is uninstalled any stale state pids/etc. will be automatically >> deleted and not left around. > > 3) create /var/run/ceph on daemon startup. This is needed anyway for > debian, where /var/run/ceph isn't supposed to be installed by the package > at all (and isn't). It's probably true for other distros too... I think > the goal is for this to be a tmpfs (and to also move to /run or > something). The init scripts mkdir -p /var/run/ceph, but sometimes the > daemons are run directly. Yeah so the ceph-deploy script runs a 'ceph-mon --mkfs' command before the init script ever gets called. This is what is emitting the error. The 3rd option is fine but if it is not something that can be done in the short term then I think the small change to the spec file is appropriate to make the RPMs work correctly with the new ceph-deploy method. > In any case, on my RHEL 6.3 box, I can 'rm -rf /var/run/ceph' and 'service > ceph start mon' without problems. What happens for you when you try this > locally? Yes so the init script does create the directory. Even if i manually create the directory before running the initial 'ceph-deploy mon create' I am still seeing the exception in the mon_create function that I originally posted about. Still trying to track down what the issue is. Thanks, derek -- --- Derek T. Yarnell University of Maryland Institute for Advanced Computer Studies _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |