Re: SEGV on syncRepl provider (ITS#3296)

[richton@nbcs.rutgers.edu]

On 22 Sep 2004 I had the chance to upgrade my production servers to 2.2.17
after experimenting with it in my testbed for a while. Both testbed and
production have been largely reliable, so I'm willing to declare this
closed unless:

(a) 'duplicated' above means this is affecting more than just me, in which
case I volunteer to try poking 2.2.17 harder (but as noted in Followup 9,
slapd is a LOT harder to crash as of late)

(b) my one 2.2.17 issue concerns any of the developers:

I wanted to modify an ACL on the master, so I edited slapd.conf and kill
-INT slapd. However, slapd SEGVd on shutdown. The curious part about the
stack trace was the apparent "deep" recursion:

# pstack core.slapd.22878 | grep avl_free | wc -l
      20

Once this unclean shutdown occurred, I was back in a very similar state,
with the provider SEGVing often. The "solution" was to clear out the
_slaves_ (ie, syncRepl to an empty slave database); since then things have
gone well.


Apart from that, I'd say close this out. If the new issue is particularly
interesting, let me know, we'll either follow through here or in a new ITS
as appropriate?