Alarm clock; cause?

Posted by Zeno on Wed 21 Dec 2005 05:59 PM — 9 posts, 29,367 views.

Zeno USA Wed 21 Dec 2005 05:59 PM #0

About 20 minutes ago, this error appeared:


Log: [*****] BUG:ALARM CLOCK!  In section new_descriptor: after accept
Alas, the hideous malevalent entity known only as 'Lag' rises once more!
Log: clearing newdesc

And it kept happening over and over. Doesn't seem to stop. I have no clue what is going on. Anyone have any idea?

[EDIT] A reboot didn't fix it.
Seems like when someone connects, this happens. But why it's happening or how to fix it, I have no idea.

[EDIT 2] I disabled DNS lookups, and it seems fine now. What would cause this all of a sudden? There have been no code changes. No server move.

Nick Gammon Australia Forum Administrator Wed 21 Dec 2005 10:11 PM #1

You have tracked it down as a DNS problem. I would try doing DNS lookups manually to see if that shows the problem. It is actually reverse DNS you are after (resolving an IP address to a name).

Zeno USA Wed 21 Dec 2005 11:56 PM #2

I'd have to do that from the host; how would I?

Nick Gammon Australia Forum Administrator Thu 22 Dec 2005 04:20 AM #3

Well you can type something like:

nslookup (address)

or

dig (address)

Nick Gammon Australia Forum Administrator Thu 22 Dec 2005 04:22 AM #4

Maybe a DNS server that should be up, is down, and the DNS system times out on the bad server.

If you have access, you could try traceroute to the address in question. Does it happen on all connections or certain ones?

Zeno USA Thu 22 Dec 2005 10:43 AM #5

Seems like all connections.

(nslookup edited out)

I noticed that in the shell after "Last login: Wed Dec 21 11:00:01 2005 from" now shows my IP when (I think) it used to show my hostname. I've already contacted the host though.

[EDIT] Well, it seems fine now. No more alarm clocks when I turned name resolving on.

Samson USA Fri 23 Dec 2005 02:45 AM #6

Just to put this baby to bed, there was in fact some kind of problem with the recursive DNS lookup process on the server. The kind that do just like muds - crash but leave no core, no log hints, nothing, and then just stay dead. Resulting in any DNS lookup process on the box to fail after the normal timeout period. Generally not fun :/

Zeno USA Sun 25 Dec 2005 09:35 PM #7

Well it's happening again. In case this helps to find out what the cause is Samson, here's the error when using nslookup:

;; connection timed out; no servers could be reached

Samson USA Mon 26 Dec 2005 12:49 AM #8

Yes, that's because it crashed on that box again. I still have not figured out why and there's really not alot of help to be found for PowerDNS, but at least when something breaks it's easier to fix. Trying to keep things running using Bind turned into more of a nightmare than it was worth.