[H-GEN] Humbug machine problems and lost email

[Greg Black]

[ Humbug *General* list - semi-serious discussions about Humbug and     ]
[ Unix-related topics. Posts from non-subscribed addresses will vanish. ]

Bruce Campbell wrote:

| On Tue, 24 Sep 2002, Greg Black wrote:
| 
| > Robert Brockway wrote:
| >
| > | On Tue, 24 Sep 2002, Greg Black wrote:
| > |
| > | > The point of my message is that this has not happened.  Messages
| > | > that were sent to gjb at humbug.org.au have not been delivered
| > | > (yet).  I am wondering if they ever will.
| > |
| > | Right now I'm not up on the exact details of the outage but if there was a
| > | hardware problem any messages on disk at the time of the crash may have
| > | bitten the bullet :( This is conjecture only though.
| >
| > This seems like a long shot.  In general, a message makes its
| > way through various hosts until it reaches its destination.  In
| > this case, the destination host sees that the message is now to
| > be sent elsewhere and does so immediately.  So, unless the crash
| > happens in that couple of seconds, the message should not be
| > lost.
| 
| Well, not really.  Any MTA worth its salt only acknowledges successful
| receipt of a message to the sending host after it has successfully written
| the message to its spool directory.  If you have a slowly failing hard
| drive returning inconsistent information, the MTA may well think that the
| message has been successfully written, but when it comes to process the
| message (ie, forward it on), it suddenly cannot find it.
| 
| Hence, if the crash occurs over a period of hours, your MTA might be
| accepting messages (and complaining bitterly in its log files that its
| spool directory is now inconsistent) effectively into a black hole.

This is a valid point, although it applies mainly to systems
administered by unskilled admins -- normally, the admin would
monitor the logs and would notice disk problems and replace the
disk before much data would be at risk.  Nothing short of RAID
(or some other mirroring technology) can guarantee data, but my
experience is that data loss really should not happen.

| diskcheckd(8) is your friend.

Indeed.

Greg

--
* This is list (humbug) general handled by majordomo at lists.humbug.org.au .
* Postings to this list are only accepted from subscribed addresses of
* lists 'general' or 'general-post'.  See http://www.humbug.org.au/