Re: New: squid-1.2-SQUIDFS

From: James H. Cloos Jr. <cloos@dont-contact.us>
Date: 21 May 1998 20:21:10 -0500

>>>>> "Eric" == Eric Stern <estern@packetstorm.on.ca> writes:

Eric> Intestingly, it IS log structured. Squid already keeps a log of
Eric> things written into the cache, so the whole thing just fell into
Eric> place. When you shutdown squid, it removes the current log and
Eric> writes a new one that is competely accurate.

Ah. Cool. But I trust you mean `writes a new log that is completely
accurate [and then moves that in place of the curretn log]', rather
than `removes the current log and [then?] ...'?

Eric> But, if you don't
Eric> get that chance (squid crash), there is a still a possibility
Eric> that you can use the old log file to recover objects from the
Eric> swap file. I'm working on this.

From reading the various lists when ext and ext2 were first being
developed, fsck(8) routines can be a PITA to do well. But since there
is that log, a backtrack until sane algorithm should be all that is
really required. (Subsequent replay of the log is, of course,
unnecessary, since it is only a cache. However, should someone find
such replay critical, it could be done by dumping the URIs of the lost
items to, eg, client(1). In any case the fsck(8)-equivalent program
needn't worry about it.)

-JimC

-- 
James H. Cloos, Jr.
<cloos@jhcloos.com>
Received on Thu May 21 1998 - 18:23:00 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:40:19 MST