[Xymon] Acknowledge issue continues with xymon 4.3.2

[Henrik Størner]

Den 07-04-2011 15:45, Clark, Sean skrev:
> But I can say, using the webpage default method produces the same error
> messages "Cookie not found" -- so I didn't think it would be my method of
> acknowledging

Ok, that would have been my next question :-)

It is quite possible that it's a bug in the xymond code. I don't know 
why it hits you so much, but that is kind of irrelevant.

Inside xymond, the cookies are stored in a datastructure called a 
"red-black tree" ("rbtree" for short). This uses some code that I picked 
up from someone else - it is used in lots of places, e.g. all of the 
hosts.cfg configuration is also stored in a similar datastructure.

However, the cookie-handling is special because it cookies are 
frequently deleted (hosts being removed happens much less frequently). I 
have had some crashes that I could never really explain when hosts were 
removed, and I really do suspect that particular bit of code that 
deletes an entry from the rbtree to be buggy. Therefore, it could very 
well be that there is a real problem here.

I've come up with a version of xymond.c that eliminates the rbtree code 
for the cookies. It uses a much less efficient way of looking up the 
cookies - basically, it will scan through all of the status-log entries 
that xymond has in memory - but since this only happens when a cookie 
needs to be renewed, or when xymond receives an ack, it should not put 
too much extra load on your system. It would be very interesting to hear 
if this patch on top of 4.3.2 solves the issue; if it does, then I 
surely know that there is a bug in the rbtree "delete node" code.

Regards,
Henrik


-------------- next part --------------
A non-text attachment was scrubbed...
Name: xymond-dumbcookie.diff
Type: text/x-patch
Size: 2886 bytes
Desc: not available
URL: <http://lists.xymon.com/pipermail/xymon/attachments/20110407/d8b2f9fd/attachment.bin>