News

[Robert Dinse <nanook@eskimo.com>]

     The news server crashed today.  I don't know exactly at what time (will
be able to determine once it's back up) but I became aware of it around noon.

     It's a crash I haven't seen it do before, rolling "Drive becoming ready" 
errors with about the center third of the screen random graphics, like
something went ballistic and just randomly scribbled about 1/3rd of the frame
buffers memory. 

     Drives all seem to be responding fine after boot.  It's going through the
rather lengthy process of fsck on all the partitions, and the news spool being
27gb raw (about 24gb after file system overhead) and typically holding around a
million files, usually takes between 45 minutes to an hour and fifteen minutes
to fsck.  This is necessary because of the cached nature of the Unix file
system and when it's not halted properly, inconsistancies that can result.

     After that finishes and the machine comes back up into multi-user mode, it
will be necessary for INN to perform a re-number to bring the active file into
sync with news articles that are actually on disk.  That process adds about
another 20 minutes. 

     It will be possible to access news (read not post) during the re-number
using Unix clients on Eskimo, but not via NNTP.  Any posts will hang until the
re-number is complete.

     I estimate that news will be back in multi-user mode, and thus accessible
via Unix clients by about 1:30pm, and the re-number should complete by 2:00pm. 
This is contingent upon fsck not finding any errors during the first pass that
it can't fix and manual intervention to fix file system errors not being
required.