[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Wave_serverV recovery need




To:  Earthworm Community,

I have a suggestion for a modification to wave_serverV which I think would
benefit the Earthworm community.

Currently, if a hard drive containing waveserver tank files develops a bad
sector or sectors, and wave_serverV attempts to read the bad sector, the
wave_serverV process will exit.  Even though the bad sector may lie in only
one tank file, none of channels served by the wave_serverV process will
continue being served after this time.  It would be preferable if, instead
of exiting, wave_serverV quit serving data from the channel with the bad
tank file and continued to serve the rest of its channels.  The program
should also send out a pager and/or email message notifying someone of the
problem.

The reason this is important is that when wave_serverV decides to exit,
there may be no one available to fix the problem, and a lot of data could
be lost.  Currently, when a wave_server process dies in Menlo Park, it's a
pageing offense, which means that someone may be awakened at night or the
weekend (undesirable).  And, several times I have had to visit the office
myself on a weekend to repair or replace a bad disk drive.  If only one
channel were being lost, instead of all channels on a waveserver process, I
wouldn't send out pages.  I would just fix the problem on Monday morning.

Does anyone else think this is a problem?  Is anyone willing to modify the
code?

Thanks, Will

----- End Included Message -----