From: Michael Renzmann <madwifi@no...> - 2005-12-28 17:21:34
I've just committed HAL v0.9.16.13 to trunk (r1367). Refer to
http://article.gmane.org/gmane.linux.drivers.madwifi.devel/1847 for the
"announce" of this unofficial release. The other changes that are
mentioned there have not been incorporated yet - help with that would be
Sam asked for feedback, so please let me know about your experiences
with this HAL. I'll take care to forward the relevant information to Sam
Thanks for your attention.
From: Michael Renzmann <madwifi@no...> - 2005-12-29 07:44:06
On Wed, 2005-12-28 at 10:18 -0800, kelsey hudson wrote:
> Am I the only one to notice this?
No, it was a real problem. The source being most probably a high load of
Subversion updates and some people trying to browse the huge changeset
that imported the new HAL. This ate all memory (512MB plus the 1GB swap
space) and made the server running crazy.
In the monitoring I saw strange spikes in the network load, too. It was
said that the server received 110 MBit/s (!) of incoming traffic, while
the outgoing traffic just had about 20 MBit/s. First I thought this was
some kind of DoS attack, but it showed that this was a result of the
memory problems which led the VM to kill various processes (including
snmpd, which is used for the monitoring).
As a countermessure I reduced the number of allowed concurrent Apache2
processes for now (was 20, now it's just 10) and restricted access to
huge changesets (which looks like http://madwifi.org/changeset/1367 ).
I'll set the number of allowed concurrent Apache2 processes back to 20
This issue began at about 6pm and was solved at about 7:45pm CET. Now
everything should be working fine again. If you experience any further
problems, please let me know.