[ome-users] var/registry/ corruption

[josh.moore at gmx.de]

Alex Sossick writes:
 > Hi Josh

Hi Alex,

 > Basically an ungraceful shutdown in this case was was caused by a faulty
 > harddisk in the server, basically freezing the server forcing a restart.
 > Restarting the server seems to cause the same issue quite often.

Ok. I could see how a power failure might cause the problem. Do you
still have the log files from the failed restart? If so, I'd like to
see if we can come up with a check for a corrupted registry. That
wouldn't be related to simultaneous restarts re:

  http://www.zeroc.com/forums/bug-reports/4237-sporadic-freeze-errors-concurrent-icegridnode-access.html#post18680

or is it possible that on system start bin/omero admin start could be
called multiple times?

 > Though I have been managing to reproduce this by just powering the
 > server down whilst its running. (effectively pulling the power on
 > it.)  This was all a bit brutal but started because the NAS server
 > there the omero data is on developed a problem and I had been
 > wanting to test what happened when the data was not available to
 > omero. Then one of the disks in the omero server raid failed.) So
 > far we have not lost any data which is really good and omero, once
 > restarted seems happy. (after deleting var/registry/)
 
 > I know we looked at the var/registry/ problem before, I am just adding a
 > script/cronjob and want to make sure there was no major issues deleting
 > var/registry/ 

None. Let us know if it works for you.
 
 > Thanks
 > Alex

And thank you,
~Josh.

 > josh.moore at gmx.de at josh.moore at gmx.de wrote on 4/9/09 10:58:
 > 
 > > Alex Sossick writes:
 > >> Hi
 > > 
 > > Hi Alex,
 > > 
 > >> In a bit of stress testing on my omero server, (maxOSX) I had
 > >> the problem that the omerodist var/registry/ gets corrupted
 > >> easily and needs deleting to be able to restart omero.
 > >> 
 > >> Is there any issue with in the unlike event omero is not running
 > >> to automatically delete var/registry/ and then restart the
 > >> server.
 > > 
 > > Deleting var/registry before the server starts is harmless. Doing
 > > so while the server is running will not compromise your data in
 > > any way, but may prevent you from logging in. Then you can
 > > shutdown and restart (creating a new var/registry) and all will
 > > work as before.
 > > 
 > >> An ungraceful shut down of the server corrupts var/registry/ quite often.
 > > 
 > > I know we've been through this before:
 > > 
 > >   http://lists.openmicroscopy.org.uk/pipermail/ome-users/2009-May/001767.html
 > > 
 > > but can you tell me again what an ungraceful shut down for you
 > > means. I need to take this issue up with ZeroC, and other than
 > > calling bin/omero admin start and stop in a loop, I've never been
 > > able to reproduce this.
 > > 
 > > Many thanks,
 > > ~Josh.
 > > 
 > >> Thanks
 > >> Alex