[omniORB] Interesting nameserver failure

[Duncan Grisby]

On Friday 15 December, "Dr Mark D Spiteri" wrote:

> Anyone else seen this problem? The nameserver stopped serving names and
> instead caused PERSIST_STORE exceptions to be thrown instead. When restarted
> the nameserver failed to recover details from its omninames-ferrari.log
> saying there was a parse error (see below). Copying
> omninames-ferrari.bak over omninames-ferrari.log did allow the nameserver to
> restart though and running nametop showed its old contents were still there.

If omniNames is unable to write its log file for any reason, it throws
PERSIST_STORE. That would happen if the disk was full, or an NFS share
went away, for example.

omniNames uses a simple redo log file, which it periodically
check-points. The .bak file contains the log file for the time before
the most recent check-point. If the disk problem occurs which
omniNames is check-pointing the log file, the .bak file contains the
complete state of the naming service. If the problem occurs at any
other time, the data in the .log file will be more up-to-date than the
last backup. It may be possible to restart the server with less loss
by manually editing the log file to remove any partial entries at the
end.

Cheers,

Duncan.

-- 
 -- Duncan Grisby  \  Research Engineer  --
  -- AT&T Laboratories Cambridge          --
   -- http://www.uk.research.att.com/~dpg1 --