On 12/10/09 21:12, Tundra Slosek wrote: > Log files are getting big enough that I'm going to strart attaching them as > files rather than inlining them - if this is poor practice, let me know. > > So after my last test failed to show exec for zfs, I changed my DTrace script > to simply log all exec calls, along with PID and PPID, to hopefully show a > history of who is calling what. Both the current script and a failure log are > attached. > > As I read these, it appears to me that zoneadmd is calling > /usr/lib/fs/zfs/umount (which becomes 'zfs' somehow?), and > hastorageplus_postnet_stop is getting called before the zfs unmount initiated > by zoneadmd is complete. So it seems to me that the HAStoragePlus is not > waiting for the zone stop to complete.
Hi Tundra, The HASP is controlled by RGM (resource group manager). It calls HASP stop method (to export the pool) after it met the dependencies, which is sczbt stop method. So it is sczbt stop method responsibility to ensure that zone is completely halted in synchronous manner by using zoneadm halt command. I am still doubtful that zoneadmd is calling zfs umount as parent pid doesn't seem to match. While discussing with Detlef on this, i came to know there is mount parameter for sczbt which allows you to mount required file systems? Are you using such parameter? Thanks -Venku >I'm not sure how to tell DTrace to log the sequencing of this except by >guessing based on timestamps. Is it possible for DTrace to show when the >smb1_zone resource actually thinks it is fully stopped, and when the >smb1_zpool resource thinks it should begin the 'stop' process? > > For reference: > > root at mltproc1:~# clrs show -p Resource_dependencies smb1_zone > > === Resources === > > Resource: smb1_zone > Resource_dependencies: smb1_lhname smb1_zpool > > --- Standard and extension properties --- > > > ------------------------------------------------------------------------ > > _______________________________________________ > ha-clusters-discuss mailing list > ha-clusters-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/ha-clusters-discuss
