cs/dns woesMIME-Version: 1.0

cs/dns woesMIME-Version: 1.0

[sl]

I'm consistently seeing some strange behavior from cs
and dns on cat-v.org. Tons of cs/dns procs are getting
(seemingly) stuck in Pwrite/Pread and never dying.
The number keeps building until the machine slows
to a crawl and I can no longer even cpu into it.

For example, right now:

	cpu% ps|grep -e ' cs'|wc -l
	     20

	cpu% ps|grep -e ' dns'|wc -l
	     60

The system is still useable but from recent experience
this will keep building until things become unuseably
slow. I did some traces in acid:

One of the cs procs stuck in Pwrite:

	http://okturing.com/src/1639/body

One of the dns procs stuck in Pread:

	http://okturing.com/src/1638/body

I think this is being triggered by mail messages, either
incoming or outgoing. I'm running another mlmgr list
that developed the same problem as what happened
with the 9front list when I was offline, moving -- people
were receiving numerous multiple copies of every message
sent to the list. After rebooting the flood of messages
seemed to stop. I suspect there is a connection here...
somewhere.

-sl

Re: cs/dns woesMIME-Version: 1.0

[sl]

The following commit fixes the problem:

http://code.google.com/p/plan9front/source/detail?r=a3579bbb5eddaca1df9c06662ff078dd281e1ccf

ndb/dns: dont override req->aborttime in udpquery()
 overriding aborttime in udpquery() makes no sense. it causes
 recursive queries to extend the timeout infinitely. nobody
 but the issuer of the request should modify aborttime.

Thanks, cinap.

-sl