[SERVER] restarting 1.7.9p2

• Date: Fri, 12 Jan 1996 12:48:36 PST
• From:
  ThwartedEfforts <abakun@scinc.com>
• Content-Type: text/plain; charset="us-ascii"

Recently, the machine E_MOO is on was adding another drive, and had to be
restarted.  I've allowed the sysadmin to shutdown the MOO, and he did,
before he hit the reset switch.  The log claims the db dumped fine.  When I
went to restart it, it loaded the db fine, but paniced with signal 11 as
soon as it finished (finished loading that is, it told me that 'will dump db
on e_moo.40.db.new', blah blah).  Anyway, promptly PANICed.

We had been running 1.7.9p2 (with MPL 1.4, FUP 1.7, and a custom builtin I
wrote).  I had recompiled the server with the regexp memory leak patch
posted a while ago, but never restarted, since I was just gonna use it the
next time it needed to be restarted.

I copied the version with the regexp memleak patch in it, and restarted that
one, and it loaded everything fine, but paniced as soon as was done loading
the db.  I reverted to the non-regexp memleak patch version, same thing.

Fortunately, the PANIC db seems fine, and was able to start up.  The only
real different I can make out between the two db files is that there were 2
open connections at the time of shutdown, both were outbound, and thus
negative.  The panic db doesn't contain any connection information ('0
active connections' is the last line in the db), whereas the one from the
good shutdown did.

Does anyone know anything about a possible bug with restarting a db that has
open outbound connections recorded in it, or should I start looking
elsewhere for the cause of this problem, or is this even an issue at all?

Andy

• Prev: [SERVER] restarting 1.7.9 (opps)
• Next: [SERVER] wishlist (was: logging)
• Index: MOO-cows
• Thread: MOO-cows