From IT 237323: Description of problem: Qdisk now is started via cman init script. Starting qdiskd always fail with the following message in /var/log/messages: Oct 29 20:02:14 esblade1 qdiskd[6146]: <crit> Specified device /dev/mpath/sas1p1 does match kernel's reported sector size (0 != -1) The output of mkqdisk -L has looked quite normal, den multipath-device was listed in the first lines, even if as always the two qdisk devices are shown with parameters (Creation date and so on) are the both underlying sd-devices. Also reformatting the qdiskd device didn't work, also with zero-ing it before and recreating it using mkqdisk. cluster.conf configuration was done in this way: <quorumd device="/dev/mpath/sas1p1" interval="5" log_facility="local4" log_level="4" status_file="/tmp/quorum_state" tko="4" votes="2"> <heuristic interval="3" program="ping 172.16.128.30 -c1 -t3" score="1"/> </quorumd> How reproducible: Setup a RHEL5.3 BETA cluster. Start cman and have a look at the line when qdiskd starts. Also have a look on /var/log/messages. Steps to Reproduce: See above. Actual results: Qdiskd won't start and won't work. Expected results: Qdiskd should start and work. Additional info: Changing the binary to the one from RHEL5.2 seems to work, but in RHEL5.2 qdiskd only seems to work (does not cman_tool leave or reboot if actual scores fall below min_scores, even if paranoid=1 is defined).
This happens only when 'device' is used. On 5.3 beta, 'device' is no longer required in multipath configurations because we added intelligent scanning which avoids multipath slaves.
Created attachment 322876 [details] Fix Correct the logic around block size checking.
Oh - and remove status_file before going to production. :)
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on therefore solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHBA-2009-0189.html