On 2007-06-01T20:17:10, Dejan Muhamedagic <[EMAIL PROTECTED]> wrote:
> Last nights Alan's odd collection of machines, which, for some
> reason, he calls lab cluster, ran CTS. It was rather successful.
> My very symmetrical 3-node xen cluster also showed similar
> results.
With 3 nodes, my cluster has completed 100 iterations with a single
testcase showing 1 unconfirmed action.
With 5 nodes, my cluster completed fine except for 6 incidents of bug
1562.
I do admit that my local copy runs with a msgsize set to 1MB, because I
can't figure out the messaging code - it in some places compares the
uncompressed message to the maximum size etc - and that was the cheapest
path, and the one I'd probably suggest for the release. Yes, we consume
larger buffers, but we don't seem to have anyone who can fix it better.
Tangent ahead:
I _think_ the second cheapest path would be to move the compression out
of that layer (and either have the CRM compress the payload directly, or
directly compress all FT_COMPRESS data on _add, so that the lower levels
never see any uncompressed fields).
(I'm looking forward to the cleanup which the openAIS integration will
bring here, as making that module pluggable should help untangle it ;-)
Regards,
Lars
--
Teamlead Kernel, SuSE Labs, Research and Development
SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
"Experience is the name everyone gives to their mistakes." -- Oscar Wilde
_______________________________________________________
Linux-HA-Dev: Linux-HA-Dev@lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/
